Digital Analytics

8月 122017
 

Optimization is a core competency for digital marketers. As customer interactions spread across fragmented touch points and consumers demand seamless and relevant experiences, content-oriented marketers have been forced to re-evaluate their strategies for engagement. But the complexity, pace and volume of modern digital marketing easily overwhelms traditional planning and design approaches that rely on historical conventions, myopic single-channel perspectives and sequential act-and-learn iteration.

SAS Customer Intelligence 360 Engage was released last year to address our client needs for a variety of modern marketing challenges. Part of the software's capabilities revolve around:

Regardless of the method, testing is attractive because it is efficient, measurable and serves as a machete cutting through the noise and assumptions associated with delivering effective experiences. The question is: How does a marketer know what to test?

There are so many possibilities. Let's be honest - if it's one thing marketers are good at, it's being creative. Ideas flow out of brainstorming meetings, bright minds flourish with motivation and campaign concepts are born. As a data and analytics geek, I've worked with ad agencies and client-side marketing teams on the importance of connecting the dots between the world of predictive analytics (and more recently machine learning) with the creative process. Take a moment to reflect on the concept of ideation.

Is it feasible to have too many ideas to practically try them all? How do you prioritize? Wouldn't it be awesome if a statistical model could help?

Let's break this down:

  • Predictive analytic or machine learning projects always begin with data. Specifically training data which is fed to algorithms to address an important business question.
  • Ultimately, at the end of this exercise, a recommendation can be made prescriptively to a marketer to take action. This is what we refer to as a hypothesis. It is ready to be tested in-market.
  • This is the connection point between analytics and testing. Just because a statistical model informs us to do something slightly different, it still needs to be tested before we can celebrate.

Here is the really sweet part. The space of visual analytics has matured dramatically. Creative minds dreaming of the next digital experience cannot be held back by hard-to-understand statistical greek. Nor can I condone the idea that if a magical analytic easy-button is accessible in your marketing cloud, one doesn't need to understand what's going on behind the scene.That last sentence is my personal opinion, and feel free to dive into my mind here.

Want a simple example? Of course you do. I'm sitting in a meeting with a bunch of creatives. They are debating on which pages should they run optimization tests on their website. Should it be on one of the top 10 most visited pages? That's an easy web analytic report to run. However, are those the 10 most important pages with respect to a conversion goal? That's where the analyst can step up and help. Here's a snapshot of a gradient boosting machine learning model I built in a few clicks with SAS Visual Data Mining and Machine Learning leveraging sas.com website data collected by SAS Customer Intelligence 360 Discover on what drives conversions.

I know what you're thinking. Cool data viz picture. So what? Take a closer look at this...

The model prioritizes what is important. This is critical, as I have transparently highlighted (with statistical vigor I might add) that site visitor interest in our SAS Customer Intelligence product page is popping as an important predictor in what drives conversions. Now what?

The creative masterminds and I agree we should test various ideas on how to optimize the performance of this important web page. A/B test? Multivariate test? As my SAS colleague Malcolm Lightbody stated:

"Multivariate testing is the way to go when you want to understand how multiple web page elements interact with each other to influence goal conversion rate. A web page is a complex assortment of content and it is intuitive to expect that the whole is greater than the sum of the parts. So, why is MVT less prominent in the web marketer’s toolkit?

One major reason – cost. In terms of traffic and opportunity cost, there is a combinatoric explosion in unique versions of a page as the number of elements and their associated levels increase. For example, a page with four content spots, each of which have four possible creatives, leads to a total of 256 distinct versions of that page to test.

If you want to be confident in the test results, then you need each combination, or variant, to be shown to a reasonable sample size of visitors. In this case, assume this to be 10,000 visitors per variant, leading to 2.56 million visitors for the entire test. That might take 100 or more days on a reasonably busy site. But by that time, not only will the marketer have lost interest – the test results will likely be irrelevant."

SAS Customer Intelligence 360 provides a business-user interface which allows the user to:

  • Set up a multivariate test.
  • Define exclusion and inclusion rules for specific variants.
  • Optimize the design.
  • Place it into production.
  • Examine the results and take action.

Continuing with my story, we decide to set up a test on the sas.com customer intelligence product page with four content spots, and three creatives per spot. This results in 81 total variants and an estimated sample size of 1,073,000 visits to get a significant read at a 90 percent confidence level.

Notice that Optimize button in the image? Let's talk about the amazing special sauce beneath it. Methodical experimentation has many applications for efficient and effective information gathering. To reveal or model relationships between an input, or factor, and an output, or response, the best approach is to deliberately change the former and see whether the latter changes, too. Actively manipulating factors according to a pre-specified design is the best way to gain useful, new understanding.

However, whenever there is more than one factor – that is, in almost all real-world situations – a design that changes just one factor at a time is inefficient. To properly uncover how factors jointly affect the response, marketers have numerous flavors of multivariate test designs to consider. Factorial experimental designs are more common, such as full factorial, fractional factorial, and mixed-level factorial. The challenge here is each method has strict requirements.

This leads to designs that, for example, are not orthogonal or that have irregular design spaces. Over a number of years SAS has developed a solution to this problem. This is contained within the OPTEX procedure, and allows testing of designs for which:

  • Not all combinations of the factor levels are feasible.
  • The region of experimentation is irregularly shaped.
  • Resource limitations restrict the number of experiments that can be performed.
  • There is a nonstandard linear or a nonlinear model.

The OPTEX procedure can generate an efficient experimental design for any of these situations and website (or mobile app) multivariate testing is an ideal candidate because it applies:

  • Constraints on the number of variants that are practical to test.
  • Constraints on required or forbidden combinations of content.

The OPTEX procedure is highly flexible and has many input parameters and options. This means that it can cover different digital marketing scenarios, and it’s use can be tuned as circumstances demand. Customer Intelligence 360 provides the analytic heavy lifting behind the scenes, and the marketer only needs to make choices for business relevant parameters. Watch what happens when I press that Optimize button:

Suddenly that scary sample size of 1,070,000 has reduced to 142,502 visits to perform my test. The immediate benefit is the impractical multivariate test has become feasible. However, if only a subset of the combinations are being shown, how can the marketer understand what would happen for an untested variant? Simple! SAS Customer Intelligence 360 fits a model using the results of the tested variants and uses them to predict the outcomes for untested combinations. In this way, the marketer can simulate the entire multivariate test and draw reliable conclusions in the process.

So you're telling me we can dream big in the creative process and unleash our superpowers? That's right my friends, you can even preview as many variants of the test's recipe as you desire.

The majority of today’s technologies for digital personalization have generally failed to effectively use predictive analytics to offer customers a contextualized digital experience. Many of today’s offerings are based on simple rules-based recommendations, segmentation and targeting that are usually limited to a single customer touch point. Despite some use of predictive techniques, digital experience delivery platforms are behind in incorporating machine learning to contextualize digital customer experiences.

At the end of the day, connecting the dots between data science and testing, no matter which flavor you select, is a method I advocate. The challenge I pose to every marketing analyst reading this:

Can you tell a good enough data story to inspire the creative minded?

How does a marketer know what to test? was published on Customer Intelligence Blog.

8月 112017
 

How can you tell if your marketing is working? How can you determine the cost and return of your campaigns? How can you decide what to do next? An effective way to answer these questions is to monitor a set of key performance indicators, or KPIs.

KPIs are the basic statistics that give you a clear idea of how your website (or app) is performing. KPIs vary by predetermined business objectives, and measure progress towards those specific objectives. In the famous words of Avinash Kaushik, KPIs should be:

  • Uncomplex.
  • Relevant.
  • Timely.
  • Instantly useful.

An example that fits this description, with applicability to profit, nonprofit, and e-commerce business models, would be the almighty conversion rate.  In digital analytics this metric is interpreted as the proportion of visitors to a website or app who take action to go beyond a casual content view or site visit, as a result of subtle or direct requests from marketers, advertisers, and content creators.

{\mathrm {Conversion\ rate}}={\frac {{\mathrm {Number\ of\ Goal\ Achievements}}}{{\mathrm {Visitors}}}}

Although successful conversions can be defined differently based on your use case, it is easy to see why this KPI is uncomplex, relevant, timely, and useful. We can even splinter this metric into two types:

Macro conversion – Someone completes an action that is important to your business (like making you some money).

Micro conversion – An indicator that a visitor is moving towards a macro conversion (like progressing through a multi-step sales funnel to eventually make you some money)

Regardless of the conversion type, I have always found that reporting on this KPI is a popular request for analysts from middle management and executives. However, it isn't difficult to anticipate what is coming next from the most important person in your business world:

"How can we improve our conversion rate going forward?"

You can report, slice, dice, and segment away in your web analytics platform, but needles in haystacks are not easily discovered unless we adapt. I know change can be difficult, but allow me to make the case for machine learning and hyperparameters within the discipline of digital analytics. A trendy subject for some, a scary subject for others, but my intent is to lend a practitioner's viewpoint. Analytical decision trees are an excellent way to begin because of their frequent usage within marketing applications, primarily due to their approachability, and ease of interpretation.

Whether your use case is for supervised segmentation, or propensity scoring, this form of predictive analytics can be labeled as machine learning due to algorithm's approach to analyzing data. Have you ever researched how trees actually learn before arriving to a final result? It's beautiful math. However, it doesn't end there. We are living in a moment where more sophisticated machine learning algorithms have emerged that can comparatively increase predictive accuracy, precision, and most importantly – marketing-centric KPIs, while being just as easy to construct.

Using the same data inputs across different analysis types like Forests, Gradient Boosting, and Neural Networks, analysts can compare model fit statistics to determine which approach will have the most meaningful impact on your organization's objectives. Terms like cumulative lift or misclassification may not mean much to you, but they are the keys to selecting the math that best answers how conversion rate can be improved by transparently disclosing accurate views of variable importance.

So is that it? I can just drag and drop my way through the world of visual analytics to optimize against KPIs. Well, there is a tradeoff to discuss here. For some organizations, simply using a machine learning algorithm enabled by an easy-to-use software interface will help improve conversion rate tactics on a mobile app screen experience as compared to not using an analytic method. But an algorithm cannot be expected to perform well as a one size fits all approach for every type of business problem. It is a reasonable question to ask oneself if opportunity is being left on the table to motivate analysts to refine the math to the use case. Learning to improve how an algorithm arrives at a final result should not be scary because it can get a little technical. It's actually quite the opposite, and I love learning how machine learning can be elegant. This is why I want to talk about hyperparameters!

Anyone who has ever built a predictive model understands the iterative nature of adjusting various property settings of an algorithm in an effort to optimize the analysis results. As we endlessly try to improve the predictive accuracy, the process becomes painfully repetitive and manual. Due to the typical length of time an analyst can spend on this task alone - from hours, days, or longer - the approach defies our ability as humans to practically arrive at an optimized final solution. Sometimes referred to as auto tuning, hyperparameters address this issue by exploring different combinations of algorithm options, training a model for each option in an effort to find the best model. Imagine running 1000s of iterations of a website conversion propensity model across different property threshold ranges in a single execution. As a result, these models can improve significantly across important fit statistics that relate directly to your KPIs.

At the end of running an analysis with hyperparameters, the best recipe will be identified. Just like any other modeling project, the ability to action off of the insight is no different, from traditional model score code to next-best-action recommendations infused into your mobile app's personalization technology. That's genuinely exciting, courtesy of recent innovations in distributed analytical engines with feature-rich building blocks for machine-learning activities.

If the subject of hyperparameters is new to you, I encourage you to watch this short video.

This will be one of the main themes of my presentations at Analytics Experience 2017 in Washington DC. Using digital data collected by SAS Customer Intelligence 360 and analyzing it with SAS Visual Data Mining & Machine Learning on VIYA, I want to share the excitement I am feeling about digital intelligence and predictive personalization. I hope you'll consider joining the SAS family for an awesome agenda between September 18th-20th in our nation's capital.

Hyperparameters, digital analytics, and key performance indicators was published on Customer Intelligence Blog.

8月 042016
 

SAS Customer Intelligence 360 is a new digital marketing hub offering that enables users to plan, analyze, manage, and track customer journeys. It includes SAS 360 Discover for digital intelligence and SAS 360 Engage for execution capabilities that enable marketers to dynamically create, manage, and place digital content across a variety of channels. These new enhancements to the our customer decision hub extends the capabilities of an organization to orchestrate omnichannel customer activity. Our intent behind this new offering? To enable our clients to take predictive action through their customer-preferred channels, and deliver a desirable, personalized experience.

Woman using tablet on table with coffee

Now that I have delivered the trendy marketing description, let's get down to business on the intersection of digital marketing, analytics, optimization, and personalization. No matter who you chat with (insert any marketing cloud vendor who recently delivered you a sales pitch), or what 2016 marketing conference you attended, it's safe to say this is a popular topic. Let's review the popular buzzwords at the moment:

  • Predictive personalization
  • Data science
  • Machine learning
  • Self-learning algorithms
  • Segment of one
  • Context-aware
  • Real time
  • Automation

It's quite possible you heard these words at such a high frequency, you could have made a drinking game out of it.

What concerns me as I scan the marketing landscape of offerings is the amount of overlap in product descriptions. Does every company do data science equally well? Will the wizardry of automation solve all personalization challenges? While we can all agree that personalization is important, it’s apparent that different organizations can have distinct views of what it actually means, and what it takes to be successful. At the center of the debate is the role of data and sophisticated analytics in personalization best practices. For the past few years, I'm repeatedly asked by clients one specific question:

Do you NEED a data scientist to be successful with data-driven personalization?

At first, I attributed this concern to the perceived lack of analytical talent available for hire. However, there are more than 150 universities offering analytic and data science programs, and talent is flooding into industry. So I placed that stereotype aside and began digging deeper, only to observe that numerous marketing cloud brands were positioning their technology to assume data scientists are NOT needed. This supports the trend of vendors highlighting "easy-button" solutions that address every imaginable obstacle related to scalable personalization. I came across lots of cute descriptions like the "just for them" algorithm, or the "magical machine learning" algorithm. Even imagery of hip marketers wearing augmented reality eyewear while deciphering targeting strategies. Are you kidding me?

When it comes to personalization, I'm uncomfortable with black-box automation, lack of statistical best practices, and the removal of the human analyst. Although 360 Engage Image 3
improvements can be achieved in using automated optimization algorithms, they can’t glean the why behind the what. Machine learning doesn't fully replace real learning. Real learning happens when analysts and marketers partner to dig into data, connect the dots and unearth insights that help brands interact with customers. Embedding those insights into hypotheses, then testing and validating completes the process.

In some ways, automating personalization ignores issues that statisticians and data scientists traditionally think about: sampling populations, confounders, model stability, bias and overfitting. In the rush to take advantage of the hype around big data, these ideas seem to be ignored or not given sufficient attention. Marketers need to pay careful consideration to the nuances of consumer behavior, brand management, and the impact of the facilitated experiences they are delivering.

Remember, SAS has passionately loved everything about analytics for 40 years. We do not care for black-box solutions, so we embedded transparent analytical features into SAS 360 Engage. It was developed with this in mind. We are the leader in advanced customer analytics, and deployed our prowess in data science for digital personalization through "blue-box" analytical decision helpers. These decision helpers are specifically built to be integrated and easy to turn on for marketers that prefer automation and efficiency. However, supporting analysts will have the option to dig deeper into the why behind the decisions. Analysts can exploit their own targeting model creations in conjunction with SAS 360 Engage's automated helpers by leveraging our deep library of algorithms and approachable tools. SAS will not limit you to a single algorithm to solve for complex consumer behavior. Compare, contrast, and most importantly, truly optimize!

This concept of "blue-box" addresses an opportunity for SAS to solve for customer journey challenges, while meeting the needs of your team's preferences. We recognize that some departments don't have data scientists available, and have real needs for automation. SAS 360 Engage offers analytical automation without challenging your team to hire new analytic resources immediately. However, as your organization begins to experience success through data-driven methods, your needs will mature. SAS 360 Engage will support that growth trajectory, and offer attractive features for analysts, statisticians, and data scientists to insert their value-add into your personalization mission.

Do I believe a data scientist is a requirement for data-driven personalization? I believe in a future where approachable technology and analytically-curious people come together to deliver intelligent customer interactions. Analytically-curious people can be data scientists, citizen data scientists, statisticians, marketing analysts, digital marketers, creative superforces, and more. Building teams of these individuals will help you differentiate and compete in today's marketing ecosystem, and SAS 360 Engage was built for the data-driven.

360 Engage Image 4

In the coming weeks, I will be releasing follow-up posts to drill into demonstrations and use cases of SAS 360 Engage. Keep an eye on this space!

 

tags: analytics, customer decision hub, customer intelligence, customer journey, data science, Digital Analytics, digital marketing, Digital Personalization, Integrated Marketing, machine learning, Marketing Personalization, Omnichannel Marketing, Predictive Personalization, SAS 360 Engage

SAS 360 Engage: Delivering blue-box predictive personalization was published on Customer Intelligence.

5月 252016
 

In April, SAS 360 Discover was introduced at SAS Global Forum 2016. Since my career started at SAS over five years ago, I have been anticipating this important announcement. In my opinion, this is a major breakthrough for the space of digital intelligence.

In my first year working at SAS, I learned of research and development to address industry needs for digital marketers. Although technologies from Google, Adobe and others address web analytics with measurement reporting, there was a shortcoming.

Historically, web analytics has always had a huge data challenge to cope with since its inception. And when the use case for analysts is to run summary reports, clickstream data is normalized:

Data Aggregation for Web Analytics

It nicely organizes raw clickstream into small, relevant data for reporting. However, this approach presents challenges when performing customer-centric analysis. Why? Holistic customer analysis requires the collection and normalization of digital data at an individual level. This is one of the most important value props of SAS 360 Discover.

Multi-source data stitching and predictive analytics require a data collection methodology that summarizes clickstream:

Data Aggregation for Advanced Analytics

The data is prepared to contextualize all click activity across a customer's digital journey in one table row, including a primary key to map to all visits across browsers and devices. The data table view shifts from being tall and thin to short and wide. The beauty of this is it enables sophisticated analysis to prioritize what is important, and what isn't. This concept of data collection and management is considered a best practice for advanced customer analytics.

How many marketers do you know who wake up in the morning and claim they can't wait to hear about how analysts are spending 80 percent of their time preparing raw web behavior data, rather than focusing on analysis and actionable insights? None, you say? Exactly! Wouldn't you rather hear your marketing analysts spend their time doing this?

20-80 Rule

I have always appreciated SAS for what it can do with structured, semi-structured, and unstructured information, but there has always been one dependency – where do I point SAS to obtain the originating data? SAS 360 Discover eliminates this requirement, and provides data collection mechanisms for your brand's website(s) and mobile apps.

SAS-Tag

 

In addition, the raw semi-structured data streams SAS natively collects are run through a pre-built relational data model using SAS Data Management for various forms of contextualization that stretch far beyond traditional web analytic use cases.

Data Model

The output of this data model schema summarizes all digital visitor behavior at this level of detail:

  • Customers.
  • Anonymous visitors.
  • Sessions (or visits).
  • Interactions (or clicks/hits).

Complete View

The data model schema will allow for additional configurations and introduction of other digital data sources to accommodate your organization's evolving needs. More importantly, the benefits of the output are profound, and listed below is a summary of SAS 360 Discover benefits:

  • Digital data normalization to support online and offline data stitching of customers.
    • When offline data is residing in your organization's data warehouse, information is available at the customer level (not a click or hit level). That's a problem when you want to link it with web or app data. The amount of time analysts spend reshaping raw HIT extracts from their web analytics solution is astonishing, and quite difficult. Customer analysis requires online/offline data stitching, and overcoming this obstacle was a problem SAS set out to solve.
  • Measurement reporting and visualization of customers and segments.
    • The reporting remains critical as an entry stage for analytics. SAS believes there should be no limit to how many reports and dashboards can be produced to meet business objectives. In other words, unlimited ad hoc reports using SAS Visual Analytics, which is the analysis tool that is packaged with SAS 360 Discover
  • Predictive analyticsmachine learning, and data science  of customers and anonymous traffic.
  • Fueling the SAS customer decision hub
    • Brands gain a competitive edge if they stop perceiving customer engagement as a series of discrete interactions and instead see it as customers do: a set of interrelated interactions that, when combined, make up the customer experience. By folding in all known customer level information into a common hub, SAS can analyze, score and take intelligent, contextual actions across channels.

SAS CDH

The path to digital intelligence from traditional web analytics covers the diversity of data, advanced analytic techniques, and injection of prescriptive insights to support decision-making and marketing orchestration. Digital intelligence is a transformation — making it a competitive differentiator. It aims to convert brands to become:

  1. Customer-centric rather than channel-centric
  2. Focused on enterprise goals as opposed to departmental
  3. Enabled for audience activation and optimization
  4. Analytical workhorses

I suspect you would love to see demonstrations of the data that SAS 360 Discover collects from websites and mobile apps in action:

  1. Decision Trees
  2. Clustering
  3. Forecasting
  4. Logistic Regression

In addition, here is the on-demand video of the SAS Global Forum 2016 keynote presentation of SAS Customer Intelligence 360.

As a marketing analyst at heart, it is extremely gratifying to share my excitement for SAS 360 Discover.  The time for predictive customer marketing in the digital ecosystem is here, and the 800-pound gorilla in advanced analytics has just unleashed your new secret weapon.

tags: 360 Discover, Data Driven Marketing, data science, Digital Analytics, Digital Intelligence, digital marketing, Integrated Marketing, marketing analytics, predictive analytics, Predictive Marketing, SAS Customer Intelligence 360

SAS 360 Discover: Predictive marketing's new secret weapon was published on Customer Intelligence.

4月 152016
 

As promised a couple of weeks ago, I am very happy to share Part 2 of a webcast series highlighting how SAS participates in the space of digital analytics for data-driven marketing with applications for personalization and attribution. Before launching the video, let me set some context for what you are about to see.

Why do we care about the intersection of digital analytics and personalization? Honestly, it is increasingly important to predict how customers will behave so you can personalize experiences with relevance. The deeper your understanding of customer behavior and lifestyle preferences, the more impactful personalization can be. However, digital personalization at the individual level remains elusive for most enterprises who face challenges in data management, analytics, measurement, and execution. As customer interactions spread across fragmented touch points and consumers demand seamless and relevant experiences, content-oriented marketers have been forced to re-evaluate their strategies for engagement. But the complexity, pace and volume of modern marketing easily overwhelms traditional planning and design approaches that rely on historical conventions, myopic single-channel perspectives and sequential act-and-learn iteration.

The majority of technologies in use today for digital personalization have generally failed to effectively use predictive analytics to offer customers a contextualized digital experience. Most are based on simple rules-based recommendations, segmentation and targeting that are usually limited to a single customer touch point. Predictive MarketingDespite some use of predictive techniques, digital experience delivery platforms are behind in incorporating predictive analytics to contextualize experiences using 1st-, 2nd- and 3rd-party customer data. In my opinion, I believe the usage of digital data mining and predictive analytics to prioritize and inform the marketing teams on what to test, and to analytically define segment audiences prior to assigning test cells, is a massive opportunity. Marketers are very creative, and can imagine hundreds of different testing ideas – which tests do we prioritize if we cannot run them all? This is where advanced analytics can help inform our strategies in support of content optimization, as it allows the data to prioritize our strategy, and help us focus on what is important.

Moving on to our second subject of interest, we transition to the wonderful world of marketing attribution. At the very core of this topic, modern marketers recognize that customers expect brands to deliver relevant conversations across all channels at any given moment. The challenge is to uncover the interactions that drive conversions through integrated measurement and insights. However, organizations struggle to employ a holistic measurement approach because:

  1. It's confusing to distinguish among the measurement approaches available.
  2. Marketers bombard customers with extraneous content.
  3. Today's misaligned data makes customer level measurement a very difficult task.

It seems like attribution has been a problem for marketers for a very long time. According to a popular quote by Avinash Kaushik of Google:

“There are few things more complicated in analytics (all analytics, big data and huge data!) than multichannel attribution modeling."

The question is: Why is it challenging? SAS strongly believes three years later that we are living in a game-changing moment within digital analytics. Marketers are being enabled with approachable and self-service analytic capabilities, and this trend directly impacts our ability to improve our approaches to problems like attribution analysis. However, rules-based methods of attribution channel weighting continue to be far more popular in the industry to date, which contradicts the recent analytic approachability trend. The time has arrived for algorithmic attribution . . . Attribution

 

Did I whet your appetite? I hope so...please enjoy episode two of our two-part webcast series, now available for on demand viewing:

 

SAS for Digital Analytics: Personalization and Attribution [Part 2]

 

SAS Customer Intelligence offers a one-stop modern marketing platform to comprehensively support the objectives of predictive personalization and algorithmic attribution - from digital data collection, management, predictive analytics, omnichannel journey orchestration, delivery across online and offline channels, and measurement. On April 19 at SAS Global Forum 2016, SAS Customer Intelligence 360 will make its debut, and subjects like digital intelligence and predictive personalization will be primary topics. This new offering will drive unprecedented innovation in customer analytics and data-driven marketing, putting predictive analytical intelligence directly in the hands of digital and integrated marketers responsible for the customer experience.

If you enjoyed this article, be sure to check out my other work here. Lastly, if you would like to connect on social media, link with me on Twitter or LinkedIn.

tags: Advanced Analytics, customer intelligence, Data Mining, data science, Digital Analytics, Digital Attribution, Digital Intelligence, digital marketing, Digital Personalization, marketing analytics, Predictive Marketing, Predictive Personalization, segmentation

Introduction to SAS for digital personalization and attribution was published on Customer Intelligence.

3月 312016
 

Digital analytics primarily supports functions of customer and prospect marketing. When it comes to the goals of digital analysis, it literally mirrors the mission of modern marketing. But what exactly is today's version of marketing all about?Modern Marketing

Honestly, we've been talking about this for years. And years. We ALL know it's what we should be doing and conceptually it's very simple, but practically, it has been very hard to achieve. Why?

Even with great web analytics, there have always been critical missing insights, which meant we didn't know for certain what the next-best-interaction for each customer was at any point in time. In addition, the development of insights and the use of analytics to define high-propensity audience segments has been distinctly slow and batch-driven in nature, delaying relevant delivery of targeted interactions. So we may get the message right, but we probably don't deliver it in a timely, consistent way, which has a dramatic impact on customer responsiveness and marketing effectiveness.

So in today's connected, always-on, highly opinionated world, we need to be a little sharper in meeting our customer's basic expectations, never mind surprising, delighting, and impressing them. While the concept of customer-centricity continues to increase in importance, improving our analytical approach to support this premise is vital.

SAS recognizes today's modern marketing challenges with digital and customer analytics. It is our mission to enable marketers to benefit from approachable and actionable advanced analytics to make more powerful decisions within today’s complex and interconnected business environments. That sounds great, right? I sense some of you reading this are raising an eyebrow of suspicion at this very moment.

Practically speaking, we want to show you exactly what that means. On March 29th, 2016, we aired episode one of a two-part webcast series, and it is now available for on demand viewing:

SAS for Digital Analytics: Introduction & Advancing Segmentation [Part 1]

We genuinely hope the webcast provided a proper introduction to how SAS participates in the space of digital analytics for data-driven marketing, and please come back in a couple of weeks when we will post Part 2 in this series entitled: SAS for Digital Analytics: Personalization & Attribution [Part 2]

If you enjoyed this article, be sure to check out my other work here. Lastly, if you would like to connect on social media, link with me on Twitter or LinkedIn.

tags: Advanced Analytics, customer analytics, customer intelligence, data integration, data management, Data Mining, data science, Digital Analytics, Digital Intelligence, digital marketing, Integrated Marketing, marketing analytics, predictive analytics, Predictive Marketing, segmentation, web analytics, webcast

Introduction to SAS for digital analytics and segmentation was published on Customer Intelligence.

3月 232016
 

The business opportunity to intelligently manage customer journeys across their lifecycle with your brand has never been greater, but so is the danger of not meeting their expectations and losing out to savvier competitors. In my opinion, the current state of most digital analytic practices continue to be siloed, tactical, and narrowly fixated on channel-obsessed dashboard reporting. That might come across as presumptuous, but keep this in mind - customer-centricity is a hot topic at the C-Suite level, and your CMO has  stated (or will very soon) that your organization is transforming into a personalization super force that will be marketing to the segment of one. If that is the case, the category of digital analytics has got to step up its game!

The antidote is digital intelligence which represents a strategic shift in approach to marketing analysis that uses insights from traditional and modern channels (we're talking online AND offline) to enable actionable, customer-obsessed analytical brilliance.

The era of the empowered customer is unraveling itself — trends in which consumers, not brands, own influence, backed by the rapid rise of digital. I strongly believe that no matter how important a company's products or services are with my life, the majority of brands I do business with continue to perform channel-centric analysis, and remain unaware of the different interactions I have with them across ALL channels. I don't care about your email or search marketing KPIs. What I care about is how you treat Suneel, no matter what device, channel, or platform I select to interact with you on.

Meanwhile, digital marketing spend continues to grow at a tenacious pace, cementing the importance of digital channels in managing the customer journey. Digital marketing is effective in all phases of the customer life cycle, ranging from acquisition, upsell/cross-sell, retention, and winback, proven by the ongoing shift of wallet share to online channels. While these are exciting times for omnichannel marketers, these more holistic approaches bring challenges. In today's fragmented digital landscape, long-established methods focused on web analytics and aggregated customer views are ill-equipped to keep pace with:

Digital interaction bread crumb trails

Customers (and prospects) interact with brands across an array of online channels and devices, creating new paths to generate incremental value associated with marketing-centric KPIs. However, customers expect personalized relevance in moments of truth, raising the bar for analytics and marketing execution. A brand's digital presence is much more than a website, such as social media, mobile applications, and wearable technologies. Conventional web analytics only track onsite behavior and lack the ability to comprehend tech-savvy customers in 2016.

The collapse of the digital silo

Brands typically construct offline and online interaction channels confined from one another, so let's reflect on that for a moment. Isn't it time we recognize that customer data is customer data, regardless of where the ingredients are collected? To deliver comprehensive customer insights, brands seek to merge digital and offline data sources together. Digital & customer analytics teams are attempting to work together, but their projects struggle due to a clash of approaches & culture. Some of the main drivers are:

  1. Data — Customers leave trails of information for marketers to chew on, and are available in structured, semistructured, and unstructured formats. There's no excuse anymore for brands to not be able to work with all three. Approachable technology exists to integrate multiple sources of online and offline customer data in meaningful ways to analyze and take action on.
  2. Skills — Have you ever sat in a meeting with data scientists and web analytic ninjas? It's like they speak two different languages, and communication between these two segments is critical for an organization to innovate in its commitment to customer analytics.
  3. Analysis — There is a reason why there is so much discussion around the application of advanced analytics. In many ways, digital marketing is ripe for analytical maturity, ranging across segmentation, attribution, and personalization. The discipline has proven its value to help differentiate a brand from its competition. When are the days of Data-Scientist“good enough” analytics going to end? Let's keep the science in data science, and stop succumbing to the false hype that sophisticated predictive marketing can be accomplished through black box, easy-button solutions.

Dynamic interaction management

Brands seek to react intelligently to shifts in consumer behavior in milliseconds, which makes the intersection of predictive analytics and data-driven marketing vital for orchestrating the customer journey. To reach your target audience in opportunistic micro-moments, the requirement of real-time actionable analytics with direct connections to personalization and marketing automation systems is the queen bee. The sole dependence on isolated, retrospective reports and dashboards of aging web analytic solutions has serious limitations in modern marketing.

Given the investment and revenue at stake for most brands, it is increasingly important to champion support of the development and continuous optimization of digital channels. Simply put, analytical sophistication lives at the center of that process. Yet most organizations continue to approach digital analytics focused on discerning traffic sources and aggregated website user behaviors. Given the intricate complications and aspirational promise of digital marketing, brands should consider modernizing and maturing their approaches to customer analytics because:

  • CX matters: Customers don't care about the challenges related to identity management across multiple visits (or sessions), browsers, channels, and devices. Does your web analytic platform support your team's abilities to recognize and track customers, not clicks or hits, across the fragmentation of touch points? With careful consideration towards the areas of data management, data integration, and data quality, analyzing customer-centric (or visitor-centric) digital activity on their journeys to making (or not making) a purchase with your brand is absolutely feasible.
  • "Good enough" analytics must end: Digital analytic teams must graduate from machine gunning their organizations with traffic-based reports that summarize the past to producing predictive insights that marketers can interpret, and take action with. I'm always impressed by web analytic teams that produce an array of historical reports with beautiful visualizations, segmenting and slicing away at their tsunami of clickstream data. However, how much impact and relevance to the business can this approach have? Customer-centricity demands that we re-engineer our thinking, and make the shift from reactive to predictive marketing analytics.
  • There's nothing exciting about siloed channel analysis: To deliver the elusive and mythical 360 degree view of customer insights, it turns out you don't need magical wizards like Gandalf or Albus Dumbledore by your side. Have you ever wondered why web analytic software doesn't allow you to perform data stitching with offline data sources? How about data mining and predictive analytic capabilities? Well, it boils down to how digital data is collected, aggregated, and prepared for downstream use cases.

Web analytics has always had a BIG data challenge to cope with since it's inception in the mid 1990's, and when the use case for analysts is to run historical summary reports and visual dashboards, clickstream data is collected and normalized in a structured format as shown in this schematic:

Data Aggregation for Web Analytics

This format does a very nice job of organizing clickstream data in such a way that we go from big data to small, more relevant data for reporting. However, this approach presents challenges when performing customer-centric analysis which requires data stitching across online and offline data sources. Why you ask? Because you cannot de-aggregate data that was designed for channel and campaign performance summarizations. Holistic customer analysis, from a digital viewpoint, requires the collection and normalization of granular, detailed data at an individual level. Can it be done? Of course it can.

Multi-source data stitching, data mining and predictive analytics require a specific digital data collection methodology that summarizes clickstream data to look like this:

Data Aggregation for Advanced Analytics

Ultimately, the data is collected and prepared to contextually summarize all click activity across a customer's digital journey in one table row, including a primary customer key to map to all visits across channels and devices. The data table view shifts from being tall and thin, to short and wide. The more attributes or predictors an analyst adds, the wider the table gets. The beauty of this approach is it allows marketers and analysts to be curious, add more data sources, and allow algorithmic analysis to prioritize what is important, and what isn't. This concept is considered a best practice for advanced customer analytics.

  • Beware of blind spots: As time passes, customers in every industry are progressively sharing more data about themselves through existing and emerging digital outlets, such as mobile applications, wearables, and other connected technology. The opportunity to ingest and analyze these new sources should excite any marketer who claims to be data-driven. However, does your web analytics platform allow you to analyze these new digital touchpoints? A brand's ability to absorb, integrate, analyze, and derive marketable insights from emerging data sources is key in this new paradigm to avoid being blindsided by customers and the competition.

The path to digital intelligence from traditional web analytics needs to cover the diversity of data, advanced analytic techniques, and injection of prescriptive insights to support decision-making and marketing orchestration. Digital intelligence is a transformation for web analytic teams — making it a competitive differentiator if executed well. It aims to transform brands to become:

  1. Customer-centric rather than channel-centric: As customers and prospects weave across an ocean of marketing channels and connected devices, digital intelligence supports the integrated analysis of interactions in concert, rather than with disconnected channel views. In addition to visibility across all channels, analysis is highly granular to identify, track, and prioritize next-best-actions for individuals. In other words, hyper-personalization to the segment of one!
  2. Focused on enterprise goals as opposed to departmental: To enable omnichannel analytics, digital intelligence is highly dependent on customer data management capabilities across all data types – structured, semistructured, and unstructured. This includes fusing interaction and behavioral data across all digital channels with first-party offline customer data, as well as second- and third-party data (if available). This enriched potpourri of data must be prepared to feed the analytical ninjas that sit within the marketing organization, line of business or centralized customer intelligence team, because it is their job to exploit this stream of information and generate insights for the organization as a whole.
  3. Enabled for audience activation and optimization. The mission of digital intelligence is the direct application of analytics to generate data-driven evidence that helps business stakeholders make clearer decisions. The potential of data mining exponentially increases with richer customer data to support segmentation, personalization, optimization, and targeting - in other words, connecting data and analytics to the delivery of relevant content, offers, and awesome experiences.
  4. Analytical workhorses: The incredibly fast-moving world of digital interactions and campaigns mean that marketers desperately need quicker analysis. Waiting days or weeks for reports and research equates to failure. Digital intelligence delivers efficiency at a pace that more nearly matches users' decision-making schedules.

SAS Customer Intelligence offers a one-stop modern marketing platform to comprehensively support the mission of digital intelligence - from digital data collection, management, predictive analytics, and marketing delivery across online and offline channels. On April 19 at SAS Global Forum 2016, SAS Customer Intelligence 360 will make its debut, and digital intelligence will be a primary topic. This new offering will drive unprecedented innovation in customer analytics, putting predictive analytical intelligence directly in the hands of digital marketers, business analysts, and data scientists. In the last few months, industry analysts have previewed and validated our abilities in advanced and customer analytics.

We are very excited for the future and potential of digital intelligence. The question is...

Are you excited?

 

If you enjoyed this article, be sure to check out my other work here. Lastly, if you would like to connect on social media, link with me on Twitter or LinkedIn.

tags: customer intelligence, Data Mining, data science, Digital Analytics, Digital Intelligence, marketing analytics, personalization, predictive analytics, Predictive Marketing, segment of one, web analytics

Web analytics vs. digital intelligence - what's the difference? was published on Customer Intelligence.

3月 102016
 

The age of the customer is upon us. As data-driven marketers, we are now challenged by senior leaders to take a laser-focus on the customer journey, and optimize the path of consumer interactions with your brand. Within that journey, there are a number of trends (or challenges) to focus on:

  1. Deeply understanding your target audience to anticipate their needs and desires
  2. At a minimum, meeting their expectations (although aiming higher can help differentiate your brand from the pack)
  3. Address their pain points to increase your brand's relevance

According to Forrester Research in their Forrester Wave™: Customer Analytics Solutions, Q1 2016 report, no matter where customer insight professionals sit within the marketing organization, line of business or centralized customer intelligence team, it is their job to employ advanced customer analytics to generate this insight. In the study, Forrester employed a 33-criteria evaluation of customer analytics vendors, and identified the 11 most significant software providers in this category. Forrester researched, analyzed, and scored them. This evaluation details their findings about how well each vendor fulfills their criteria and where each vendor stands in relation to each other. Click here to see the results! ForresterAs I imagine many of you reading this blog posting are users of SAS technology, let's say it together:

That's right baby!!! #ILoveSAS

Okay, after completing my little happy dance, let's get serious again. The report begins with a very interesting headline, that some of you might find debatable, that reads as:

You don’t need a Ph.D. for advanced customer analytics

How does that make you feel? I would imagine for some, it might be inspiring, motivating, and aspirational. For others, you might be raising an eyebrow in suspicion. Personally, based on my interpretation, I believe it speaks to the exciting era of approachable analytic technology and innovation that we are living in. The authors of the report proceed to share observations on the current bottleneck of available data science talent, and the market demand for their talent far exceeding supply. Due to this phenomenon, the job of deriving valuable insights is increasingly becoming the responsibility of data-driven marketers who need friendly tools for data management and analytics.

Marketing analytic ninjas - please stand up!

The next big headline stirs the geeky juices of data science enthusiasts by stating:

Don’t assume you can use traditional analytics to uncover deep customer insights

Calling for the usage of a new breed of algorithms, methods, and advanced analytical techniques, leveraging all your data (vs. statistical sampling), and moving past the reliance of reports (because let's be honest, many organizations are still drowning in endless dashboards). Additional differentiators that are called out include:

  1. Delivering output at the customer level - We are living in the age of the customer, not the age of the aggregated, siloed marketing channel. Customer-centric analysis is a transformation of approach for many organizations, but the benefits can be massive.
  2. Predicting the future - Across the entire customer journey, predictive marketing is all the rage!
  3. Leveraging new data types and sources - Semistructured and unstructured data continue to be under-exploited by marketing organizations, and this needs to come to an end. The technology and tools are in place and available today to enable analysis of these data sources

Jumping ahead past the evaluation's criteria scores (link available below for those yummy details), here is what Forrester had to state about SAS:

"Consider SAS if you are looking to drive innovation in customer analytics. As the only company in this evaluation whose sole focus is on analytics, SAS excels at analytics production — that is, turning data into insights. It invests heavily in its customer intelligence suite to ensure that its solutions anticipate and even drive innovation in this space for marketers. With real-time data ingestion and integration, regular addition of new algorithms, and a robust selection of model management and deployment options, the limitations you will encounter with using SAS for customer analytics are less likely to be with the core functionality than with your team’s ability to use them."Innovator

Wow! We love this viewpoint, and are extremely pleased with the assessment. SAS is strongly committed (maybe obsessed) to innovating customer analytics by adapting solutions to meet our clients’ ever-changing data, business and deployment needs.

  For those of you interested in reading the entire press release, it is available here.

When you're ready to explore more, please start with these technologies:

We look forward to beginning a customer analytic journey with you soon.Forrester2

The Forrester Wave™ is copyrighted by Forrester Research, Inc. Forrester and Forrester Wave™ are trademarks of Forrester Research, Inc. The Forrester Wave™ is a graphical representation of Forrester's call on a market and is plotted using a detailed spreadsheet with exposed scores, weightings, and comments. Forrester does not endorse any vendor, product, or service depicted in the Forrester Wave. Information is based on best available resources. Opinions reflect judgment at the time and are subject to change.

If you enjoyed this article, be sure to check out my other work here. Lastly, if you would like to connect on social media, link with me on Twitter or LinkedIn.

tags: Approachable Analytics, customer analytics, customer intelligence, Data Driven Marketing, Digital Analytics, Forrester Research, marketing analytics

Who's the king of advanced customer analytics? was published on Customer Intelligence.

12月 222015
 

Although the title of this blog posting has all the ingredients to attract the eyes of an analyst, the content is targeted for all personalities of a digital marketing organization. Before we jump into the marketing analytic use case regarding forecasting, scenario analysis, and goal-seeking  for digital analytics, let's spend some time on the magic of stories. As Tom Davenport stated in his fantastic article titled, Telling a Story with Data:

"The essence of analytical communication is describing the problem and the story behind it, the model, the data employed, and the relationships among the variables in the analysis. When the relationships among variables are identified, the meaning of the relationships should be interpreted, stated, and presented relevant to the problem. The clearer the results presentation, the more likely that the quantitative analysis will lead to decisions and actions—which are, after all, usually the point of doing the analysis in the first place."

While creative visionaries and data scientists are both tremendous organizational assets within a team, it is the alliance between these two segments that will push marketing forward. Although aspirational, this is a difficult challenge to overcome. Let me begin by sharing a bit of my story - one that began with a four year career start in graphic design and creative marketing communications, and then taking making a leap to the quantitative side of marketing. I've seen and listened to how DIFFERENT these two segments of the marketing world are, and now as a preacher for the potential of marketing analytics, one's ability to make analysis interpretable and approachable is critical.

Google recently published a nice article titled, Staffing Your Marketing Measurement Team: Why You Need Data Storytellers, and one takeaway that I love from this piece is:

"The true value of data emerges when marketers are able to use it to tell a meaningful story. Enter the data storyteller, or marketing measurement analyst. This is the person who can push the tools, translate insights across the business, and motivate stakeholders to participate."

This quote nails the crux of the issue - if we don't take ACTION on the insights of analytics, it was nothing but a school project. Influencing decision-makers within an organization isn't easy, and if they do not understand the analysis, nothing will ever change. There are people who are good at creative marketing strategy, and there are people who are good at marketing analytics. However, there aren't many people who can toggle between the two, and serve as the translator who inspires both sides.

In my personal opinion, the recent surge in analytic technologies becoming more approachable is key. The special ingredient in that trend is visualization and analytics joining forces in ways we have never seen before. Why is this happening? Seeing and understanding data is richer than creating a collection of queries, dashboards, and workbooks. According to the infamous American mathematician John W. Tukey:

"The greatest value of a picture is when it forces us to notice what we never expected to see.”

The "ah-ha" moment. The best part of my work day!

In addition, when analytics becomes approachable, interpretable, and transparent to the entire marketing organization, the behavioral change of how we work together highlighted in this video becomes a reality:

Visual Analytics represents a new category of interactive and collaborative technology to provide a path to be curious and innovative. Marketers are imaginative, and are constantly pushing to analyze new and exciting data sources (i.e. clickstream, social, IoT wearables, etc.), which require the ability to scale to very large amounts of information. However, what is different here is the ability to perform sophisticated analysis, and produce visualizations to support data-driven storytelling.

Finally, we arrive at the digital analytic use case. The intention is to highlight my personal approach to tip-toeing that fine line of producing meaningful analysis, while narrating the marketing storyline. Here is the description of the business case, and my demonstration video.

Business Challenge:

How do I allocate digital media spend to drive more traffic to my website in a future time period?

Marketing Applications:

  1. Identify the most important acquisition channels (i.e. attribution)
  2. Simulate & optimize ad spend to acquire incremental traffic and meet business objective

Let me know what you think in the comments section below. If you enjoyed this article, be sure to check out my other work here. Lastly, if you would like to connect on social media, link with me on Twitter or LinkedIn.

tags: data visualization, Digital Analytics, Digital Intelligence, digital marketing, Forecasting, Goal-seeking, marketing analytics, predictive analytics, Predictive Marketing, Scenario Analysis, visual analytics, visual statistics, web analytics

Forecasting, goal-seeking, and magical stories for digital analytics was published on Customer Analytics.

12月 092015
 

Broadly speaking, the holy grail of digital media measurement is to analyze the impact and business value of all company-generated marketing interactions across the complex customer journey. In this blog post, my goal is to take a transparent approach in discussing how data-driven marketers can progress past rules-based attribution methods, and make the business case for leveraging algorithmic applications.

Let's begin with a video example that pokes humor at the common problems related to multi-channel marketing attribution. The business challenge is that everybody believes they should have more marketing budget because their tactics are supposedly responsible for driving sales revenue.  The video suggests that challenges arise rapidly when supporting analysis to justify these claims isn't sound. While the video is fictional, the problems are very real. With that said, there are three main drivers to getting digital attribution analysis right:

  1. Allocating credit across marketing channels more accurately
  2. Providing invaluable insights to channel interactions
  3. Empowering marketers to spend more wisely in future media activity

Have you ever given thought to the many ways that a customer can find your brand's digital properties? Organic results on a search engine, display media campaigns, social media links, re-targeting on external sites, and the list goes on in today's fragmented digital ecosystem. One thing is for certain - consumer digital journeys are far from linear. They can occur across multiple platforms, devices and sessions, and organizations are challenged with gaining an accurate understanding of how:

  • External referral clicks (or hits) are mapped to channels and visits
  • Visits are mapped to anonymous visitors
  • Anonymous multi-channel visitor journeys are mapped to identifiable individuals across different browsers and devices

With careful consideration towards the areas of data management, data integration, and data quality, analyzing customer-centric (or visitor-centric) channel activity on their journeys to making a purchase with your brand can have immense benefits. Ultimately, marketers desire to apply a percentage value that can be attached to each channel's contribution to the purchasing event (or revenue). This is critical, as it allows the organization to determine how important each channel was in the customer journey, and subsequently, influence how future media spend should be allocated.

Sounds fairly easy, right? Well, as Avinash Kaushik (digital analytics thought leader at Google) stated in his influential blog post on multi-channel attribution modeling:

"There are few things more complicated in analytics
(all analytics, big data and huge data!)
than multi-channel attribution modeling."

The question is...why is it challenging? Avinash's blog post was written in the summer of 2013, and I strongly believe 2.5 years later we are living in a game-changing moment within digital analytics. Marketers are being enabled by technology companies with approachable and self-service analytic capabilities, and this trend directly impacts our ability to improve our approaches to problems like attribution analysis. However, rules-based methods of attribution channel weighting continue to be far more popular in the industry to date, which contradicts the recent analytic approachability trend. Before we dive into algorithmic attribution, let's review the family of approaches commonly applied in rules-based attribution:

Last Touch & First Touch Attribution

Suneel 1Allocates 100% of the credit to either the last or first touch of the customer journey. This approach has genuine weaknesses, and ignores all other interactions with your brand across a multi-touch journey. It is stunning, in my opinion, that web/digital analytic technologies have traditionally defaulted to this approach in enabling their users to perform attribution analysis. The reason for this was last/first touch attribution was easy, and could claim ownership of the converting visit (although that is only partially true). Thankfully, times are changing for the better, and this rudimentary approach has proven ineffective, guiding marketers (for the sake of job security) to try more intelligent methods.

Suneel 2Linear Attribution

Arbitrarily allocates an equal credit weighting to every interaction along the customer journey. Although slightly better than the last and first touch approaches, linear attribution will under-credit and over-credit specific interactions. In a nutshell, it over-simplifies the complex customer journey with your brand.
.

Time Decay & Position Based Attribution

Suneel 3Time decay attribution arbitrarily biases the channel weighting based on the recency of the channel touches across the customer journey. If you are bought into the concept of recency within RFM analysis, there is some merit to approach, but only when comparing with other rules-based methods. Position based attribution is another example of arbitrary biasing, but this time we place higher weights on the first and last touches, and provide less value to the interactions in-between. As Gary Angel (partner & principal of the digital analytics center of excellence at Ernst & Young) stated in his recent blog posting:

"There’s really no reason to believe that any single weighting system somehow captures accurately the right credit for any given sequence of campaigns and there’s every reason to think that the credit should vary depending on the order, time and nature of the individual campaigns."

Although there are some other minor variants to the rules-based method approaches, highlighted above are the majority of approaches that the digital marketing industry commonly uses. As a principal solutions architect at SAS, I have the opportunity to meet with clients across multiple industries to discuss and assist in solving their marketing challenges. When it comes to attribution, here is a summary of what I have seen clients doing in 2015:

Buying Web/Digital Analytics Software That Includes Rules-Based Attribution Measurement

This is typically when an organization invests in a premium (or more expensive) software package from their web/digital analytics technology partner, which includes out-of-the-box attribution capabilities. Here is a video example discussing how one of the most popular web analytic platforms in the world includes capabilities for various methods of rules-based attribution.

Two takeaways from this video that I love are:

  1. Suneel 4Comparing the attribution problem to soccer (or futbol), and accepting that we cannot give 100% credit to the goal scorer. There is a build up of passes to set up the goal (i.e. purchase), and each of these events (i.e. marketing channel touches) contribute value. Even though names like Messi, Ronaldo, and Neymar are commonly known in soccer, ignoring names like Iniesta, James Rodríguez, or Schweinsteiger would be a travesty.
    .
  2. Focus on the journey, and performing visitor-centric analysis as compared to visit-centric analysis

The difficulty I possess with the video is leveraging the term "data-driven attribution" when rules-based methods are the only approaches highlighted. In my opinion, we are only grazing the surface of what is possible. Algorithmic attribution, on the other hand, assigns data-driven conversion credit across all touch points preceding the conversion, using data science to dictate where credit is due. It begins at the event level and analyzes both converting and non-converting paths across all channels. Most importantly, it allows the data to point out the hidden correlations and insights within marketing efforts.

Have you ever wondered why web/digital analytic software doesn't include data mining and predictive analytic capabilities? It has to do with how digital data is collected, aggregated, and prepared for the downstream analysis use case. Suneel 5

Web/digital analytics has always had a BIG data challenge to cope with since it's inception in the mid 1990's, and when the use case for users is to run historical summary reports and visual dashboards, clickstream data is collected and normalized in a structured format as shown in the schematic to the right.

This format does a very nice job of organizing clickstream data in such a way that we go from big data to small, relevant data for reporting. However, this approach has limited analytical value when it comes to attribution analysis, and digital marketers are only offered rules-based methods and capabilities.

Data mining and predictive analytics for algorithmic attribution require a different digital data collection methodology that summarizes clickstream data to look like this:

Suneel 6

Ultimately, the data is collected and prepared to summarize all click activity across a customer's digital journey in one table row. The data table view shifts from being tall and thin, to short and wide. The more attributes or predictors an analyst adds, the wider the table gets. This concept is referred to as preparing data for the analytic base table (or input modeling table). This is the best practice for advanced algorithms to be used to fit the data. Shhhhhhhh! We'll keep that insightful secret between us.

More importantly, don't let this intimidate you if you're new to these concepts. It boils down to the ability to reshape granular, HIT-level digital data for the best practices associated with data mining. Can it be done? Absolutely, and algorithmic digital attribution is a prime example of big data analytics for modern marketing. The question I challenge my clients with is to consider the arbitrary (or subjective) nature of rules-based methods, and associated limitations. Although they are easy to apply and understand, how do you know you aren't leaving opportunity on the table? This leads me to the next recent trend of what I observe clients doing.

Buying Algorithmic Attribution Consultative Services

The best way to kick this section off is to share how 3rd party marketing attribution vendors introduce themselves. Here are two video examples to consider:

  1. Video #1
  2. Video #2

How do you feel after watching these videos? If you are raising your hands to the sky thanking the higher forces of the marketing universe, I completely understand. Many of my clients describe their marketing organization's culture as unprepared for algorithmic attribution, ranging from lack of subject-matter expertise, big data hurdles, or employee analytical skills. There can be tremendous value in selecting an external partner to handle analysis and actionability, and accelerate your ability to make better digital media investment decisions. 3rd party attribution vendors have the domain knowledge, technology, and a track record, right?

In addition, this segment of my clientbase seem less concerned with transparently understanding how to analytically arrive to their decision strategy, as long as the financial results of their attribution vendor's services look good compared to baseline KPIs. Although these vendors will never reveal their analytical secret sauce (i.e. intellectual property), digital marketing is an overwhelming ecosystem, and who has the time to discuss analytical model diagnostics, misclassification rates, ROC plots, lift curves, and that silly confusion matrix...

That's one trend I see. The other trend is when a marketing organization is analytically mature, and this leads me to the next section.

Building Algorithmic Attribution Models In-House

Do you want to perform algorithmic attribution analysis yourself and maintain a transparent (white-box) understanding of how your analytic approaches are influencing your digital media strategies? If you answered yes, I believe the best way to take you on this journey is through a case study. I'm a strong advocate of this approach, and believe this is a cutting edge application of marketing analytics:

Case Study: Hospitality Industry

The Challenge:
Unable to scale digital analytics for algorithmic attribution to measure drivers of conversion and advertising effectiveness.

Business needs to understand:
- Drivers of resort hotel bookings online,
- Marketing channel attribution to bookings with statistical validation,
- Insights to allocate future digital media ad spend.

Current Limitations:
- Clickstream and display ad-serving data very large in size,
- Rules-based attribution methods largely inaccurate.

Technical Summary:
- 90 day est. file size for extracted Adobe HIT data: 3.0 TB,
- 90 day est. file size for extracted Google DoubleClick (display media) data: 4.0 TB,
- Analytical data prep, modeling, and scoring workflow must be capable of processing on Hadoop platform (i.e. big data lake).

Digital Data Preparation Summary:
In this exercise, the hospitality brand was extracting raw data from their relationships with these digital marketing technologies into an internal Hadoop data landing zone. Their goal is to start stitching various digital marketing data sources together to gain a more complete view of how consumers interact with their brand. Analytically speaking, this is very exciting because we can gain a better understanding of the value of channel touches, onsite click activity, media impressions, viewability, creative content, ad formats, and other factors that we do not have comprehensive visibility into with traditional web/digital analytics.

One valuable insight I would like to share is if you have never worked with raw clickstream data or display media data before, it would be advantageous to obtain a data dictionary and channel processing documentation from your digital marketing solution vendor(s). For example, every website that has installed web analytic tracking has an array of unique goals, interactions, segments, and other attributes that were configured for that specific business model. Analysts will not understand what eVar 47 is without a translation document. Guess what? eVar 47 is going to have a completely different definition from Brand #1 to Brand #2 to Brand #3. Sorry - there is no easy button for this.

Your analysts will thank you sincerely for taking these steps, and it will improve their ability to succeed. Since this is a SAS Blog, I imagine many of you will want to understand how we worked with the raw digital data in this case study.

1. Data access: SAS Data Loader for Hadoop
2. Visual data exploration to assess data quality issues: SAS Visual Analytics
3. Reshaping the data for analytic modeling (i.e. recoding, transformations, joins, summarization, transpositions): SAS Enterprise Guide

Analytic Model Development Summary:

Now we move on to the fun and sexy step of the process...

Our methodology of approach was to address the digital attribution challenge as a predictive modeling problem. This involves three key goals:

  1. Produce a predictive model that computes the probability of conversion given a set of visitor journey predictors.
  2. Determine the incremental lift in probability of conversion for each channel in a visitor journey, and use this to compute attribution.
  3. Provide insight into the relationship between conversion and the predictor variables (marketing channels, onsite click activity, digital demographics, etc.).

To drill into the details a bit further, I'll break this down in three steps through a hypothetical example:

Step 1 - Create Predictors and Target

  • Convert visitor journeys into a table with rows of channel impression counts and conversion information.
  • This data is used to train (and validate) the predictive model.

Suneel 7

 

 

.
.
.
.
.
.
.
.
Step 2 - Compute Incremental Lift

  • Use the predictive model to compute the incremental lift in probability of conversion by adding one channel at a time in each visitor journey.
  • Example for one visitor journey: Display > Email > Search > Display > $

Suneel 8

Step 3 - Compute Attribution

  • Process all conversion journeys & accumulate channel credit to compute channel attribution.
    .

Suneel 9

Analytic Modeling Results:

Now we can get to the fun and sexy stuff...

This analysis included 17 marketing channels, over 1,000 predictors, ~24,000,000 digital visitor journeys, and a rare conversion event occurrence of less than 1%. Oh my!

Suneel 10

Now let me ask you a question - do you believe there is one piece of math that will solve all of our attribution challenges?

Suneel 11

Absolutely not. The game of digital media investment is all about precision, precision, PRECISION! To maximize precision, in the field of data mining, we employ the use of champion-challenger modeling. Simply put, we throw a bunch of math at the data, and the algorithm that does the best job of fitting the data (i.e. minimizing error) is selected.

Suneel 12

Scaling to large digital data with champion-challenger modeling is not trivial, but through the modernization of analytical processing in recent years, the time has arrived to dream bigger. Random forests, neural networks, regressions, decision trees, support vector machines, and more are all fair game, which means we can produce accurate assessments of marketing channel importance using the power of advanced analytics. Here is a snapshot of our modeling results within this project:

Suneel 13

For those of you unfamiliar with misclassification rates, it's nothing more than a metric to summarize how many mistakes our analytical model is making. The lower the value, the better, and in this exercise, the random forest algorithm did the best job in analyzing and fitting our attribution data. There's your champion!

Next, let's share a lift chart visualization to help us get our heads around what we've accomplished here:

Suneel 14

The beautiful takeaway in this example is we have identified an attractive segment (top decile with highest probability scores) that is 8.5 times more likely to convert as compared to randomly targeting the entire marketable population. Secondly, if we alter that segment view to the top two deciles, they are 4.7 times more likely to convert.

BOOM! This is awesome because we can now profile these segments, and proactively hunt for look-a-likes. In addition, be imaginative in how you might use these segments in other forms of digital marketing activities. For example, A/B testing in web personalization efforts.

But what about the marketing channels themselves? Which ones ended up being more (or less) important)? Well, here is a great visualization for channel weighting interpretation:

Suneel 15

The odds ratio plot clearly highlights these insights in a non-technical manner. Channels above the horizontal line have a positive impact in increasing the probability of a visitor conversion, and channels below the line have a negative impact. For those of you who are unfamiliar with odds ratio plots, they serve as an ingredient to feed into a marketing dashboard that can explain market channel attribution performance.Suneel 16

So how accurate were we? Was this model any good?

True positive rate simply means how accurate was our ability to correctly predict conversions. True negative rate summarizes our ability to accurately predict non-conversions. Given that our original event of conversion behavior was below 1% across a three month time window, our ability to predict conversions based on the modeling insights is a MASSIVE improvement (86.67 times more accurate) versus the mass marketing approach (or pure random targeting). Even though there is still room for improvement, these are very promising results.

To deploy or activate on these insights, this will vary based on your organization's approach to taking action. It may be the scoring of an internal database, or it might be passing the model score code to your digital data management platform to improve their ability to deliver media more intelligently. There are a number of use cases for marketing activation, but by doing this analysis in-house, you will have flexibility to conform to a variety of downstream process options.

Again, I suspect many of you will want to understand how we analytically modeled the digital data in this case study.

  1. Algorithmic modeling: SAS Enterprise Miner (High Performance Data Mining)
  2. Analytic scoring: SAS Scoring Accelerator for Hadoop
  3. Marketing channel performance dashboarding: SAS Visual Analytics

Why Aren’t More Organizations Doing This?

From my experiences in 2015, I believe there are three reasons:

  1. Large data volumes require the use of modern big data platforms
  2. The talent required to unlock the marketing value in that data is scarce, but the climate is improving - if you're searching for talent, please consider the future analysts, data miners, and data scientists we are training at the GWU MSBA program in Washington DC
  3. Organizations are rethinking how they collect, analyze, and take action on important digital data sources

If you made it this far in the blog posting, I applaud your commitment, desire, and time sacrifice to go on this written journey with me. We discussed the current landscape of digital marketing attribution, from methods of approach to providing a real case study in support of making the justification for algorithmic attribution (i.e. it's not a mythical creature from another universe). Digital data mining is on the rise, becoming more approachable, and will provide organizations competitive advantage within their industries for years to come.

Marketing analytics matter!

Let me know what you think in the comments section below. Lastly, if you would like to connect on social media, link with me on Twitter or LinkedIn.

tags: Digital Analytics, Digital Attribution, Digital Data Mining, Digital Data Science, digital marketing, marketing analytics, Marketing Attribution, Multi-channel Attribution

Making the case for algorithmic digital attribution was published on Customer Analytics.