marketing optimization

5月 112017
 

Multivariate testing (MVT) is another “decision helper” in SAS® Customer Intelligence 360 that is geared at empowering digital marketers to be smarter in their daily job. MVT is the way to go when you want to understand how multiple different 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 increases. 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 2.5 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 web marketer have lost interest – the test results will likely be irrelevant.

A/B testing: The current standard

Today, for expedience, web marketers often choose simpler, sequential A/B tests. Because an A/B test can only tell you about the impact of one element and its variations, it is a matter of intuition when deciding which elements to start with when running sequential tests.

Running a good A/B test requires consideration of any confounding factors that could bias the results. For example, someone changing another page element during a set of sequential A/B tests can invalidate the results. Changing the underlying conditions can also reduce reliability of one or more of the tests.

The SAS Customer Intelligence 360 approach

The approach SAS has developed is the opposite of this. First, you run an MVT across a set of spots on a page. Each spot has two or more candidate creatives available. Then you look to identify a small number of variants with good performance. These are then used for a subsequent A/B test to determine the true winner. The advantage is that underlying factors are better accounted for and, most importantly, interaction effects are measured.

But, of course, the combinatoric challenge is still there. This is not a new problem – experimental design has a history going back more than 100 years – and various methods were developed to overcome it. Among these, Taguchi designs are the best known. There are others as well, and most of these have strict requirements on the type of design. safety consideration.

SAS Customer Intelligence 360 provides a business-user interface which allows the marketing 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.

The analytic heavy lifting is done behind the scenes, and the marketer only needs to make choices for business relevant parameters.

MVT made easy

The immediate benefit is that that multivariate tests are now feasible. The chart below illustrates the reduction in sample size for a test on a page with four spots. The red line shows the number of variants required for a conventional test, and how this increase exponentially with the number of content items per spot.


In contrast, the blue line shows the number of variants required for the optimized version of the test. Even with three content items per spot, there is a 50 percent reduction in the number of unique variants, and this percentage grows larger as the number of items increase. We can translate these numbers into test duration by making reasonable assumptions about the required sample size per variant (10,000 visitors) and about the traffic volume for that page (50,000 visitors per day). The result is shown below.

A test that would have taken 50 days will only take18 days using SAS’ optimized multivariate testing feature. More impressively, a test that would take 120 days to complete can be completed in 25 days.

What about those missing variants?

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 for the tested variants and uses this to predict the outcomes for untested combinations. You can simulate the entire multivariate test and draw reliable conclusions in the process.

The Top Variant Performance report in the upper half of the results summary above indicates the lift for the best-performing variants relative to a champion variant (usually the business-as-usual version of the page). The lower half of the results summary (Variant Metrics) represents each variant as a point located according to a measured or predicted conversion rate. Each point also has a confidence interval associated with the measurement. In the above example, it’s easy to see that there is no clear winner for this test. In fact, the top five variants cannot reliably be separated. In this case, the marketer can use the results from this multi-variate test to automatically set up an A/B test. Unlike the A/B-first approach, narrowing down the field using an optimized multivariate test hones in on the best candidates while accounting for interaction effects.

Making MVT your go-to option

Until now, multivariate testing has been limited to small experiments for all but the busiest websites. SAS Customer Intelligence 360 brings the power of multi-variate testing to more users, without requiring them to have intimate knowledge of design of experiment theory. While multivariate testing will always require larger sample sizes than simple A/B testing, the capabilities presented here show how many more practical use cases can be addressed.

Multivariate Testing: Test more in less time was published on Customer Intelligence Blog.

7月 272016
 

Back in 2001, when I started working in the enterprise marketing software business, customer relationship management or CRM was seen as the cure all from a sales and marketing perspective.

“If only we could more quickly send direct mailers offering a buy one, get one video rental, we could corner the market” one executive told me. CRM deployments at that time were costly and resource intensive.

My how times have changed.  But one thing hasn’t changed – there remains three critical components to consider when standing up a solid customer intelligence software solution – data, insight and action.

Data meaning a centralized data repository – containing first and third-party data. Insight being intelligence derived from the use of analytics. And action being the ability to orchestrate interactions across sales, service, and marketing touch points.

Recently, SAS partnered with Forbes Insights to look in depth at the first component in this trilogy – data. This report, Data Elevates the Customer Experience, looks at how data insights can be customer experienceused to improve the overall customer experience with brands.

It’s clear that data management and integration are crucial components of delivering customer experiences that are relevant, satisfying and valued. If the foundational component of data is not a source that is trusted, reliable, and of high quality – marketing efforts will be subpar. Period. Some interesting findings from the report:

  • 20 percent report that individuals in their organization are able to take advantage of information and derive actionable insights from data being shared across their enterprises.
  • 14 percent of executives are able to report that their data is structured on a cross-functional, synchronized way.
  • 45 percent of respondents claim their data is “not yet fully integrated”.
  • 41 percent say their data is still siloed by departments.

Primary challenges to greater insight around customer experience management are familiar: siloed business units (34 percent), legacy applications (33 percent) and siloed applications and processes (28 percent).

The recurring message throughout the report was that a centralized and integrated data store, that contains all data sources and types, provides the most benefit to organizations. Siloes are still a huge challenge and integration is still difficult – especially when purchasers of data management, analytics, and marketing applications take a best-of-breed approach – without considering how those components work together. As Mike Flannagan of Cisco stated, “IT systems may take years to integrate, and now we’re seeing it with [Internet of Things] environments.

I interpret that to mean – data and system integration has historically been difficult with backend environments – and that challenge isn't going away any time soon.

Additionally, a previous Forbes Insights report found that predictive analytics and the ability to deliver real-time insights across all channels are top of mind. However, in order to deliver these capabilities sound data management and data integration have to come first.

So what can you do at your organization if you see these same issues and want to tackle them head on? I would offer a few suggestions:

  • Consider a master data management solution. Organizations need the ability to manage multiple data domains, on-board source system data, match at the master data level and enable data governance – both at an analytical and operational level. Keeping data up to date, clean and relevant will drive better insight and action.
  • Use analytics to derive insight. If your data isn’t sound, your analytics won’t be great, but if your data house is in order, then analytics will uncover intelligence about your customers that will send your customer experiences skyrocketing.
  • Use data to inform marketing process orchestration. The better your data foundation is, the more insight you will have when setting up customer journeys and mapping them out across inbound and outbound channels and touchpoints.

These are just a few ways that you can take action to improve the customer experience. I would encourage you to look to a vendor, like SAS, that provides a truly integrated marketing solution – from data to insights to action. This will make your organization more efficient and help you to avoid the issues that our friends at Forbes found.

tags: Advanced Analytics, brand, customer experience, data management, Forbes Insights, marketing operations management, marketing optimization, master data management

Winning the customer experience war begins in the data trenches was published on Customer Intelligence.

11月 212015
 
Picture of a random teenager (my son) on his mobile phone.

My son not using email.

I have the hardest time getting my teenagers to use email. It's a generational thing that all parents probably contend with, and for marketers, it points to the growing importance of social media and mobile apps in the marketing mix. But unless teenagers comprise your largest target market, there's one enduring fact about email you should bear in mind:

Email still matters. A lot.

Email matters because it remains the most common form of business communications.

Picture of a random teenage girl (my daughter) taking a selfie.

My daughter definitely not using email.

It's very versatile, allowing you to attach files and embed links, and there are time-stamps and other important functions that give it enduring value. For my kids, they simply need to know that it's the preferred channel for their teachers, bosses and other adults (such as their parents). For communicating with most anyone over the age of 25, especially business decision-makers and influencers, email remains important.

With that thought in mind, I stopped by the booth of Emma - a company dedicated to email and survey communications - at a recent Content Marketing World. I was immediately taken by one of their hand-outs, enticingly titled: 18 Email Stats to Know, Love, and Quote at Parties. Their booklet includes the caveat, "Please don't quote these at parties," but nowhere did it warn me to refrain from blogging about these email stats.

So, after asking Content Marketing Strategist Jamie Bradley if I could blog about these statistics, I happily present them to you verbatim from their great handout:

  1. The average open rate for welcome emails is a whopping 50%, making them 86% more effetive then email newsletters. So, an automated welcome is a no-brainer. (Source: MarketingSherpa)
    .
  2. 51% of all email is opened on a mobile device. So, design for the smallest screen first, and use responsive design templates to ensure your emails look great on any screen size. (Source: Litmus)
    .
  3. 80% of people are only scanning your email. Capture the big idea of your email with a bold image and strong headline. (Source: Nielsen Norman Group)
    .
  4. Relevant emails drive 18X more revenue than broadcast emails. Use data you've collected to segment your subscribers, and create targeted, personalized emails. (Source: Jupiter Research)
    .
  5. Nurtured emails make 47% larger purchases than non-nurtured leads. Create an automated series that delivers your best content to subscribers over time, and it'll pay off. (Source: The Annultas Group)
    .
  6. The human fingertip is around 46px squared, so size your buttons accordingly to make tappipng easy for mobile readers. (Source: Apple)
    .
  7. 58% of adults check email first thing in the morning. Knowing that, try sending your emails early in the day - you might see an uptick in opens. (Source: Enzaga)
    .
  8. Our brains process images 60,000 times faster than text. So, change up your text-to-images ratio, and take the time to choose images that tell your story. (Source: 3M Corporation)
    .
  9. People check their mobile phone up to 150x a day. Use email and social together to reinforce your message in all the places people visit on their phones. (Source: kpcb.com)
    .
  10. iPhones will cut off a subject line over 32 characters. Put the most important words of your subject line first. (Source: Harland Clarke Digital)
    .
  11. Surrounding text with a more significant amount of white space improves comprehension by 20%, so don't pack your content too tightly. (Source: Crazy Egg)
    .
  12. Subscribers that receive a welcome note show 33% more long-term brand engagement. Extend the life of your list by making a warm, friendly first impression. (Source: Chiefmarketer.com)
    .
  13. Eye-tracking data shows that viewers look to the same part of the screen where images of people are looking. Seems like a pretty good place to put your call to action, eh? (Source: The Brain Lady Blog)
    .
  14. Receiving too many emails is the #1 reason people unsubscribe. Use automated emails that arrive just at the right time instead of blasting (ick) everyone with every piece of news. (Source: Chadwick Martin Bailey)
    .
  15. Adding video to your email campaigns can increase click rates by 300%. Host your video online and create a thumbnail play button for your email. (Source: Wistia)
    .
  16. Email conversion rates are 40x that of Facebook and Twitter. Promote email signup opportunities on social media, and then use email to convert. (Source: Kissmetrics)
    .
  17. Personalized emails improve clickthrough rates by 14% and conversion rates by 10%. But don't stop at the first-name greeting: Personalize the delivery of your emails with automation. (Source: Aberdeen Group)
    .
  18. 100% of people will high-five you for sharing a smart email marketing statistic. Take these stats on the go at myemma.com/stats - they look great on your phone. (Source: Emma)

 

In addition to being a nice mix of fun and helpful, what struck me about these email facts was how some of the recurring themes across many posts in this blog showed up in the advice they served up with each one. The most prominent one is the idea that we need to stop annoying our customers with email. We do that by paying attention to the relevance of our message, but also taking care to optimize our interactions so channel, frequency and other factors make sense to our customers and we're also confident we have the best combination of offers and recipients in our marketing.

I hope you find this useful and thought-provoking. One other fun little something I'll share from Emma is the Our People page on their website. Move your cursor around on that page over each person's picture and you'll get a fun surprise and a little glimpse into their corporate culture.

Let me know what you think - and as always, thank you for following!

tags: email marketing, marketing optimization, Mobile Marketing

18 fun and handy facts about email marketing was published on Customer Analytics.

10月 232015
 

In our wonderfully complex world of digital media, with on-line all-the-time connectivity, social media, big data opportunities (and challenges), it's never been more challenging (and at the same time easy) for marketers to use money wisely. Technology has accelerated the pace of doing business and one consequence is the common perception of saturation - and no medium is more glaring in this regard than email.

Will he open your email? Will he convert? Opt-out?

Will he open your email? Will he convert? Opt-out?

We all get too many emails, and in many places regulation drives the need to have easy opt-outs, so the price to pay for an overly active outbound marketing campaign can be quite steep.

So let me start with a confession.  As a marketing analytics manager, I have never challenged a contact policy.  By contact policy, I am talking about the rules such as ‘Don’t send more than one email a week’ or'Don’t contact a customer by telephone more than once a quarter’.  I like to describe these as the ‘X in Y’ type of contact policies.  They always seemed reasonable as the high touch and high value products had bigger windows than the low touch and low value products, so were probably in line with what customers wanted.

I support a lot of optimisation projects now, and contact policy is a central piece to the optimisation jigsaw, so I get to see how it varies across territories and industries.  But whilst the X and Y in the ‘X in Y rule’ varies significantly, the one invariable is that none of them are analytically driven.  Yet in the increasingly digital world, with its new channels, with more products to be sold and communications made, with more organisations making them, to customers with more devices…..but with less time, it seems appropriate to manage this more efficiently, doesn’t it?

Now some may say that in the digital world that contact policy is not required – after all everything can be trigger- or rules-based, and whilst that is clearly debatable, even in the digital world, I think it is beyond debate that if a customer abandons a cart 20 times in one day – they don’t need 20 SMSs or emails of phone calls to remind them of this fact….and the twentieth SMS is unlikely to get a better response than the previous nineteen messages did.

But what is the limit – and how should we derive it?

It seems to me there are two core drivers behind what we might try to achieve with a contact policy:

  1. Reduce cost / increase response
    That is, don’t sent communications that, because there were other communications sent recently, are less likely to get a response, making it uneconomical.
  2. Increase customer satisfaction
    That is, don’t upset the customer with too many communications.

Reduce Cost/Increase Response

To achieve the first objective, there may be natural tests that have occurred – for example some customers were contacted one week apart, others two weeks apart.  This can help to build a degradation curve, and the definition of degradation here being the drop in actual response versus the expected response.  So for example if a model is telling you to expect 5% response, but you get 4% response, then there is 20% degradation.

Adrian Carr OptimisationI’ve used a bit of artistic licence with the curve to the right, but I am pretty comfortable that the end points are correct.  That is, if you contact a customer with the same offer, an infinitesimal time after the first offer, there is probably going to be 100% degradation of response for that second communication.  Similarly, if you leave it a long enough time between offers, response will return to what was expected, and so there will be no degradation.

The dotted line can then be established so that an appropriate ROI or other relevant KPI can be achieved, and the contact rule is set up appropriately.  For example if response degrading 50% is acceptable, and this happens at 2 weeks, then a ‘1 in 2 week’ contact rule probably makes sense.

Of course this curve could vary by product, channel, customer segment etc, and it may not be extremely accurate, but it would be interesting to make a start on finding out what it looks like, right?

Increase Customer Satisfaction

To achieve the second objective – to increase customer satisfaction - I would first of all spin it around and describe it as reducing customer dissatisfaction.  This can be easier to measure using KPIs such as attrition (churn) rates, or opt out rates.

Most marketers have the tools to achieve this – it’s similar to building a response model, but the thing we are trying to predict is ‘dissatisfaction’ (i.e. the customer attrited, or the customer opted out) the inputs are the communications that a customer received (and some additional variables built up on these).

So it may be the case that a particular email caused the opt outs, but equally, it may be the case that customers that received 10 emails in a month were more likely to opt out, but those that received less than 5 didn’t opt out more than the average.  A traditional logistic regression model will highlight these negative trigger points for the customer, and so a contact rule can be set up to avoid it – e.g. no more than 5 emails per month.

Clearly there are many caveats around both analytical approaches, and there are other ways of achieving a good outcome (e.g. market research, or including past contacts in existing models), but it’s worth highlighting that the data and the toolset to do this more analytically is probably sitting there in many organisations right now, and with the current plethora of communications that customers are receiving, now may be a good time to work out when is the right time.

Whilst having a set of customer level contact policies is going to be much more customer focused, and will lead to fewer opt outs and reduced cost….it will raise a few more challenges.  Multiple contact rules are hard to manage, especially across customers and channels and product sets.  Add the continued responsibility to stay focused on maximising sales, increasing revenue, and meeting operational lead volume requirements, and you have a situation where you cannot use money wisely by managing with spreadsheets.

An analytically-driven process, such as SAS Marketing Optimization can step up to manage the complexity.  With easily configurable customer level contact rules, constraint management and multiple goal maximisation, to name just a few capabilities, it’s the perfect way to achieve a more customer focused approach, that still meets business requirements. You're welcome to explore more at www.sas.com/customerjourney or contact us and ask for me or any of my colleagues.

tags: email marketing, marketing optimization

Want to use money wisely and reduce opt-outs? Read this. was published on Customer Analytics.

10月 162015
 

We have finally arrived to our last stop in this Journey Towards Direct Marketing Optimization. If you have read the entire series of articles, you are now ready to understand what optimization means and how to perform it. For those who have not read the previous articles, I suggest you to do so before reading this last one:

In the first article, I explained what Campaign Optimization means in a direct marketing context. Then I talked about how to differentiate customers through propensity scores or response rates. Segmenting customers by any of these measures is essential in order to make decisions about the best customers to target. In the second piece, I explained how to plan campaigns in order to optimize them all together. It is very important to understand that our optimization scenario should include all the eligible customers, before current prioritization rules. After selecting the targets, we stopped to analyze what kind of goals we can set when optimizing. Having the right information to calculate the optimized value is critical. As optimization would not be necessary without constraints, I described which the most common ones are and how to handle them. So now it is time to put everything together and arrive to our Optimized destination.MO_Journey_step5

 

No need for complex coding

As said in the first article, to calculate an optimized scenario considering all the factors we have discussed, it is necessary to use complex optimization algorithms. However, there is no need to program them. SAS Marketing Optimization is a solution with a user-friendly interface to simulate and evaluate different scenarios. Our solution is aimed at marketing analysts that need to spend more time analyzing possibilities rather than building complex queries or decision flows. The user only needs to input data and then he/she will be ready to simulate scenarios. This input data is no other than planned campaigns/offers and eligible customers for them.  Additional data can be used in order to evaluate scenarios that are more complex.

Business constraints and contact policies can be loaded as a table or written in the same user interface. This is very flexible as users can evaluate different possibilities without changing any code.

MO_UI

Simulate and analyze scenarios

Optimization is an analytical process, whereas prioritization is usually an automated procedure. Before arriving to an optimized final solution, several scenarios are usually analyzed. A scenario is a set of constraints and contact policies restrictions over an eligible universe. Each scenario is an optimized solution, considering the constraints we are setting.

Going back to our very simple example in the first article, we could work on 3 different scenarios:

  1. No restrictions at all: all the eligible customers can receive all the offers (no contact policy or business restrictions in place). This is usually called “Base Scenario” and we use it just to have an idea of the potential value of all our offers. Of course, this scenario is unrealistic.
  2. Contact Policy Constraints: we use the same information as above, but add as a constraint that we can contact at most one time each customer. There is no need to input the data again, we just save the previous scenario with a different name and add the restrictions we need. We will get the optimal solution, given the new contact policy restriction.
  3. Contact Policy and Business Constraints: to the previous scenario, we add a new business rule “at least one customer per offer”. As before, we use the “save as” capability and testing scenarios becomes just a matter of selecting or deselecting constraints. Again, this is another optimized solution for the constraints set.

If we extrapolate this to a real world problem, we could have many different constraints (budget, channel capacity, contacts per segments, etc.) and analyze what would happen if we loosen or tighten the restrictions.

Once that we discovered the best solution, the result is a list of customers and the offers we should send to each of them.

Optimization needs commitment

Probably while reading the series of articles you have identified more than one obstacle to apply optimization in your organization. Most organizations that embrace optimization get on board to change processes at the same time. Although this might sound discouraging, in our experience, we have found that it is indeed a very useful exercise. Many clients have discovered during assessment sessions that most of the prioritization rules they have in place are outdated. They just keep on using them because they are used to it and are afraid of changing the automated processes. However, when asking them about the value they are missing they have no possibility to answer.

In order to make the transition smoother, we recommend starting by optimizing only a small group of campaigns. Usually, the most critical campaigns are the ones that use the call center as channel, as they are the most restrictive ones. After simulating several scenarios, we arrive to the most convenient solution and customers are targeted.

Probably the optimized solution is not the one that a Product or Segment Manager was expecting. For example, the amount of customers to contact in each offer might not match the expected by the offer owner. It is useful to bear in mind that optimizing is about finding the best overall results, not for each product in particular. This is why optimization needs the commitment of all the participants and from the management above all.

I hope you have enjoyed our journey towards direct marketing optimization and that you had taken a souvenir from each of our stops. I will be glad to be your Captain if you decide to start this trip - or any of my colleagues. Do not hesitate to contact us if you would like to discuss how to get on board.


Editor’s note:

If you did not read the previous posts in this series, I encourage you to do so since Luciana planned them as a step-by-step journey. Marketing optimization is a very effective way to tie overall business objectives (often profitability) to marketing campaign activity because it mathematically calculates the best aggregate outcomes based on how you define them. If you'd like to dig a little deeper into how marketing optimization could work for you, I suggest you download this whitepaper, Improving Multichannel Marketing with Optimization. Among other useful content, it includes a practical checklist of seven steps to optimize your marketing.

tags: analytics, marketing optimization, Optimization Journey

Optimization step 5: putting everything together was published on Customer Analytics.

10月 092015
 

Over this blog series: “The Journey toward direct marketing optimization” I have covered all the topics that are part of the optimization process. As said before, however, the most impactful part in an optimization problem is to set the constraints.  If a company does not have any type of boundaries when selecting and contacting customers, then there is no optimization problem at all.

As a customer intelligence advisor, I have never worked with a company that does not have some contact policy restrictions. Most marketing teams do manage customer saturation and choose not to contact their customers more than certain number of times during a period. Another frequent constraint is channel capacity, particularly call center capacity. Most companies choose this channel as it is usually more effective, but it is also the more restrictive.

In the example used in the first article “The Journey toward direct marketing optimization” I already introduced two different type of constraints: contact policy and amount of targets per offer. In this last stop before we reach the destination, I will go deeper into restrictions and how they influence the optimized results.  As in SAS Marketing Optimization restrictions are classified as business constraints and contact policies constraints, I will use the same categories.MO_Journey_step4

 

Observing Contact Policies

In order to cover this type of restrictions, I will use the same simple example used at the beginning of these series of articles.

Two customers and two offers - not as easy to optimize as you'd expect.

Remember that if there were no constraints, we could have an expected profit of 265, the sum of all the possible offers for all the customers. However, the company in the example indeed has a contact policy in place. The company cannot contact the customer more than once, so we cannot select the same customer for the two available offers. In this simple example, having this restriction made us lose 105 of profit as we selected the most valuable offer for each customer (offer 1 in both cases).

It is important to differentiate between two different types of contact policy restrictions. The first one makes a customer eligible for optimization. This is relative to the target selection process and not optimization, as we discussed in the second stop of our journey, “Designing Campaigns for Optimization”.

When setting the contact policy in the optimization scenario, we are considering the contact for the period we are optimizing, not the previous ones. This is necessary because a client might be eligible for more than one campaign, so we have to decide how many times we are going to contact him/her during the period under analysis.

Contact policies can also be different by channel. Many companies we work with don't count email campaigns as a contact, so if they have different campaigns that are communicated through different channels, they can give their client more than one offer in the period optimized. The most restrictive channel is usually the call center. Most of the companies set that a customer can only be contacted once by an agent in the period.

Setting the contact policy in the scenario is very important in the optimization process. Knowing the company’s boundaries is essential as it affects the optimization results. The contact policy can be as complicated as necessary. SAS Marketing Optimization will take all the information in order to calculate the best results, with no need of complicated processes.

Managing Business Constraints

Contact Policies are not the only restrictions companies have. Many other business constraints have to be considered in an optimization problem. Continuing with our example, the next restriction we had was the minimum amount of customers per offer. Even for our very simple example, setting this restriction makes the problem more difficult to solve. By defining that each offer must have at least one customer, we have to change the solution and we are now able to get 150 as a result (offer 2 for client 1 and offer 1 for client 2), 10 less than the previous scenario.

Constraints can include, among others, budget limits, channel capacity, risk limitations, amount of offers, etc. These constraints can apply to all the customers or we could set constraints by segments, cities, or any other characteristic of the customer. For example a bank might need to limit the amount of loans offered to customers with a higher risk profile, even if they might probably the most likely to accept them.

Constraints can be “at least” or “at most”, giving the users the possibility to analyze different combinations and possibilities. For example, a Product Manager may want a minimum of customers for his offer, but also a maximum in order to control expenses. SAS Marketing Optimization will find the best number of offers between these boundaries.

In order to set certain constraints we will need particular data to input to the tool. For example if we have a budget limitation, we need to know the cost of each offer. This information can also be used when setting the optimization goal. If we need to set restrictions on channels, we need the capacity of each of them. The more information we have, the more complex we can make the scenario. However, be careful, complexity doesn't always mean better results.

All these kinds of factors are usually not considered in prioritization strategies as they are very difficult to implement. Implementing SAS Marketing Optimization can help marketers to target the best customers, with the best offer for each, observing all the business restrictions, and getting the best results overall.

We are coming to the end of this exciting journey. In the next and final stop I will explain how to put all these pieces together. However, as you might have noticed, Marketing Optimization is not only about having the best solution in place, but also about rethinking the campaign process. I will then cover this topic as well in my next article.  Join me!


Editor’s note:

If you did not read the previous posts in this series, I encourage you to do so since Luciana planned them as a step-by-step journey. Marketing optimization is a very effective way to tie overall business objectives (often profitability) to marketing campaign activity because it mathematically calculates the best aggregate outcomes based on how you define them. If you'd like to dig a little deeper into how marketing optimization could work for you, I suggest you download this whitepaper, Improving Multichannel Marketing with Optimization. Among other useful content, it includes a practical checklist of seven steps to optimize your marketing.

tags: analytics, marketing optimization

Optimization step 4: setting the right boundaries was published on Customer Analytics.

10月 022015
 

We are now in one of the most interesting stops in our journey towards marketing optimization. We have been talking about optimization on the assumption that we have an optimization goal, however, as strange as it might sound, companies do not always know the overall goal of their direct marketing activities, much less how to measure the results. Although the ultimate objective is usually to sell more, doing that does not always deliver the highest profit for the organization. In fact, each campaign usually has its own objective and when combined together they might even have contrasting objectives.

In the first post, “The Journey Toward Direct Marketing Optimization” I explained, through a very simple case, how to maximize the expected profit, where:

Profit = probability of response * offer value

Then I covered how to measure the probability of response in my second post, “First Stop: Identifying the best customers through analytics.” Now we'll address what we can optimize and what information we need in order to do so for the second part of the equation.

MO_Journey_step3

 

Maximization

Almost every company does direct marketing campaigns in order to increase sales, profit or revenue. The optimization goal in these cases is to maximize one of these measures. However, most of these companies don't actually know the income that they will get from a customer if he/she accepts the offer.

The optimization process will make that determination better if we can measure each customer/prospect we are targeting. Exactly what that means depends on the industry we work in and the type of products/services we are offering:

  • In telecommunications, incremental ARPU (average revenue per user) is a common measure that shows the extra value a current customer will bring if the offer is accepted. Incremental ARPU is relatively easy to calculate for plan upgrade campaigns, but more difficult for other type of offers.
  • At some banks we have worked with, we have used the Net Operating Income (NOI) of a customer as a measure to maximize. However, measuring the incremental NOI as a result of a certain product is a very difficult task, so we use the current NOI.
  • In the Insurance sector, one of the typical ways of valuing a customer is by doing a Customer Lifetime Value (CLTV) model. Then we can use this number in order to set a value for the campaign we are doing.
  • In Retail optimization, it is easier to calculate the value of a campaign if the offer is a specific product, because in that case it's the product's price x units sold to calculate the impact. In other industries or other scenarios, offer values are a bit more complicated to calculate.

Minimization

Although is more common to set maximization goals, it's also possible to have minimization as a goal. One of the most frequent measures to minimize is the cost of the campaigns. If we have several campaigns and different channels to use, we could choose to minimize the overall cost of all campaigns. We all know that an email campaign is much cheaper than a call center campaign, but probably contacting the customers by emails is less effective. So we may want to have a balance between cost and expected response, which is possible by setting constraints. We will come back to this topic at the next stop in our journey.

Like for the maximization goals, the type of measure we need to minimize varies by industry. For instance, it is very frequent in the Banking industry to try to minimize the overall risk. If we are offering risk products, like loans or credit cards, we can measure the risk of the customer and try to minimize the overall risk of these campaigns as a goal. In the Retail industry, many campaigns can be discount offers, so we could need to minimize the overall discount we are giving to our customers.

Combined Goals

Sometimes one single optimization goal is not enough and we need to set secondary objectives. Going back to the risk example in banking, our main objective could be to maximize profit, but to minimize the risk of the offers. In most industries, a very common goal is to maximize revenue and minimize cost, and it's important to note that by maximizing profit instead of revenue, the outcome minimizes the applicable costs.

Controlling this in a prioritization process is certainly very complicated. SAS Marketing Optimization can work with two optimization goals in order to calculate the best result for both. This is of course more time consuming as a simple optimization, but very easy to perform through our software.

Whichever the goal is, we need information to use as an input in the tool. Yet, as said before, it is not imperative to have perfect information in order to optimize campaigns. Like with probability scores, we have alternative ways to measure the results. With the customers we are working with, generally the value is set by offer and not by customer. This is already a step forward as most prioritization processes do not consider this when selecting customers.

If choosing any customer as a target will produce the same results (they all have the same value), then the probability of response becomes more important to differentiate among the customers. If the probability of response is not a predictive model but an average response rate for all the customers, then constraints and contact policies will be more relevant to define the optimized solution.

Constraints and Contact Policies are in fact our next stop in this journey.  If we could do whatever we wanted, without restrictions, then this journey towards marketing optimization would not be necessary. Therefore, our next stop is the key to understanding optimization. I invite you to keep on travelling together. We are very close to our final destination: Optimized Campaigns!


Editor’s note:

If you did not read the previous posts in this series, I encourage you to do so since Luciana planned them as a step-by-step journey. Marketing optimization is a very effective way to tie overall business objectives (often profitability) to marketing campaign activity because it mathematically calculates the best aggregate outcomes based on how you define them. If you'd like to dig a little deeper into how marketing optimization could work for you, I suggest you download this whitepaper, Improving Multichannel Marketing with Optimization. Among other useful content, it includes a practical checklist of seven steps to optimize your marketing.

tags: analytics, marketing optimization

Optimization step 3: setting the goal was published on Customer Analytics.

9月 252015
 

In this journey to direct marketing optimization, we have already gone through two important concepts: understanding what optimization means and differentiating customers through analytics for optimization.

It is now time to make a stop to think about the campaign design process. We will describe how we need to plan our marketing activities in order to make the most out of the optimization process.Image depicting a step 2 on a 5-step journey.

Typical Target Selection Process

In order to select the targets for campaigns, most companies use campaign management solutions. In this kind of solutions, users can build campaign decision flows to select those customers that are eligible for a certain offer. The decision flows will typically have several selection rules to check basic descriptive variables like gender, product holding, city, segment, etc. More analytical driven companies will also use predictive scores in order to select the best customers for their campaigns.

Another important selection filter is to check contact policy rules. Customers that have already been contacted last month with a similar offer might not want to be bothered with the same offer, so they are not included in the target selection.

During any given month, companies can run dozens or even hundreds of campaigns. As a result, a customer can be selected for more than one campaign. In order to solve this issue, companies usually include prioritization filters in their campaign flows. By doing this, they select a unique offer for each customer. As these rules are part of the decision flows in the campaigns, the targets are already unique and there is no room for optimization.Image showing a funnel with eligible customers coming out the small end.

Setting the Eligible Universe

One of the premises in Marketing Optimization is that there are customers that are eligible for more than one campaign and we cannot offer all of them due to different restrictions.

The word eligible is very important in the formulation of an optimization scenario. Only customers that meet the basic requirements for campaigns (eligibility) should be included in the optimization exercise. It is very important to set these basic requirements or filters. For example if our contact policy is not to contact a customer that has already been contacted last month, then this customer is not eligible for any campaign and cannot be part of the eligible universe. However if we would like to understand how our optimized result might change if we include them, then we need to incorporate these customers in the optimization exercise. The same criteria applies to customers with high risk profiles, with high probability of churn, etc.

As most companies have automatic prioritization processes in place that control the customers that are assigned to each campaign, thinking about the eligible universe is not always straightforward. It implies thinking about rules we would like to challenge and then making the changes in the automated process already in place.

Campaign Planning

Another important aspect of campaign design for optimization is planning. We have worked with many customers and we have found there are different approaches. In some companies, 80% of the campaigns are planned at the beginning of the month. During the rest of the month, some extra marketing campaigns are run, but the majority of the activities occur as planned. In other companies, where the marketing dynamic is not that controlled, most campaigns are not planned and targets are selected ad-hoc as requested by product or segment managers.

When optimizing, planning is very important. By planning the campaigns that are going to be delivered during certain period, we can select all the eligible customers at once and run the optimization scenarios with all the possible targets. Most often, the optimization period is monthly or bi weekly. For example if the optimization is done monthly, at the beginning of each month the campaign designers can select the targets, build the eligible universe, evaluate different scenarios and set which customer receives which offer. If there is no planning and campaigns are delivered just as they come, a customer that could have been selected for a more profitable campaign might be selected for an earlier campaign that month. This would be more likely to be a prioritization exercise than an optimization one.

As you can see, the optimization journey is a very exciting one. It's full of adventures and new discoveries. Our next stop in is the land of objectives. We will explore what our goals should be when optimizing direct marketing campaigns and which data we need to collect in order to set clear objectives. Stay tuned for our next stop on our journey - same time, same place next week.


Editor’s note:

If you did not read the previous posts in this series, I encourage you to do so since Luciana planned them as a step-by-step journey. Marketing optimization is a very effective way to tie overall business objectives (often profitability) to marketing campaign activity because it mathematically calculates the best aggregate outcomes based on how you define them. If you'd like to dig a little deeper into how marketing optimization could work for you, I suggest you download this whitepaper, Improving Multichannel Marketing with Optimization. Among other useful content, it includes a practical checklist of seven steps to optimize your marketing.

tags: analytics, marketing optimization

Optimization step 2: designing campaigns was published on Customer Analytics.

9月 172015
 

Welcome back to our journey toward marketing optimization! If you missed the introduction, I previously explained the concept of direct marketing optimization in my last post, "The journey toward direct marketing optimization.” From that departure point, the topics in this post will take us to the first stop in the journey where we'll explore how to optimize the target selection.

MO_Luciana_step1

I'll start by explaining how to understand the value of each customer for each offer using basic and advanced analytics. Even if you are a campaign manager and already familiar with this approach, let's quickly review it because it's vital in order to feed the optimization algorithm with appropriate data.

Differentiating customers through analytics

When selecting targets for campaigns, we need to know which customers are more suitable for each of the campaigns. Think of "suitability" as the probability of the customer to accept the offer. This probability will be more or less accurate, depending on the analytical methods we use.

In order to explain this, I will use the same example I used before. We have a scenario with two offers and two clients where:

Profit = probability of response * offer value

Two customers and two offers - not as easy to optimize as you'd expect.

A customer can be more suitable for a campaign because he/she has a higher propensity to accept, or because the value it can generate is greater. The value we need to assign to a customer depends on the optimization objective we set. We'll will cover that idea in a future post, and for now let's focus on the probability of response.

Predictive Analytics

The ideal situation would be to have propensity scores for each of the offers that a customer is eligible for, as this would help you improve how you select your targets. In that regard, it's not only about calculating the scores, but also understanding how these scores were calculated. It's not the same to calculate the probability to accept an offer as it is to calculate the probability of buying a certain product. The modelling approach is different and the results might be different as well. Let me explain it with an example:

In order to improve targeting, a Telecommunications company wants to know the probability of a customer to buy a new Smartphone next month. To calculate this, they take the last month's campaign data where they offered the new Smartphone at a special price. With the past targets, their descriptive data and the responses, they run a data mining algorithm and build a predictive model. This predictive model allows them to score the new targets for the new offer. This approach is called response modeling.

Another way of calculating the propensity of a customer to buy a new Smartphone without campaign response data could be to identify customers that changed their mobile model from one month to the following and those who did not. This flag will be similar to the response flag we could have from a campaign. With this information and the customers’ descriptive data, they could run a data mining algorithm and build a predictive model. New targets then would be scored with the model.

The key is to understand if conditions were similar from one period to another. Predictive models are based on historical data.  If the company offers the new Smartphone at very different prices each month, then predictive scores might not be that precise. For instance, a customer probably would be more willing to buy the last model at half-price this month, than only getting a 10 per cent discount last month. Analysts should inform campaign managers about the approaches followed in order to let campaign managers make better target selections. These considerations are also very important in the optimization process.

Descriptive Analytics

Predictive modeling is not the only way to categorize customers through analytics. Another valid technique is to use past response rates, without building predictive models.

Using the same example as above, we could calculate the response rate by segments or another descriptive variable (data plan, account aging, etc.) and use this information to assign “probabilities” to the new targets. This could be done with all the offers and we could assign a value to each of the offers we want to optimize.

Although this technique might not be the most precise one, in many situations it is far better than existing prioritization methods and much easier to achieve.

Choose Carefully

Whichever analytical method is used, the important thing is to be able to differentiate among customer sub-sets with meaningful variables that you choose carefully. Without that valid differentiation, all customers will essentially be the same in terms of probabilities and value. The end result would be an optimization algorithm with limited information to pick one customer over the other for each of the campaigns. Think of it as offering a group of people 39 tubs of vanilla ice cream and then asking them to pick their favorite one - the answers you get may be correct, but they won't be very meaningful.

Our next stop in this journey toward marketing optimization will be to explore designing campaigns for optimization. Thanks for joining us on this journey! Please stay on board for the whole trip - I promise it will be worth it!

tags: analytics, marketing optimization

Optimization step 1: identify the best customers was published on Customer Analytics.

9月 092015
 

Today most companies engage in direct marketing to communicate with their customers. If you work on a campaign management team, you are probably sending emails, SMS, letters (you can admit it -  you are still doing it!) and also using call centers in order to extend offers to your customers.

Some companies use analytics to choose their targets in a data driven way, while others rely their business knowledge and mostly use rules to do so. Furthermore, in the same company the strategy can vary among Product Lines or Departments. Whichever your case is, all companies face the same problem: too many offers for too few customers and limited resources.

The need for prioritization

When selecting targets for the different offers, it can happen than a client can be eligible for more than one campaign. In a "perfect" world we'd have unlimited call center agents, an unlimited marketing budget and customers would not mind being contacted many times with multiple offers or even the same offer over and over. As we all know, that kind of "perfect" world doesn't exist.

Most of the companies we work with develop prioritization strategies in order to decide which offer the customer receives. Prioritization is a hierarchical approach that usually involves creating rules to pick customers for each campaign.   Although prioritization mostly handles contact policy restrictions, it usually does not take into account overall marketing results. We can illustrate this situation with a very simple example.

Consider the scenario where there are two offers that can be extended to two customers. We know the value we could get from each offer if the customer accepts the offer and we also know the probability or response rate for this offer.Two customers and two offers - not as easy to optimize as you'd expect.

Observing business restrictions

In a perfect world, we could get a value of 265 (the sum of all the offers). But as said before, the world is not perfect. Companies have some business constraints that campaign managers should consider when selecting the audience. The most usual constraint is the contact policy. Most of the companies have in place contact strategies so as not to saturate their clients. In this example, we will consider we can contact the customer only once in the period. The solution in this case would be to choose for each client the most valuable offer. This can be considered a Maximization strategy.  By doing so, Customer 1 and 2 would receive Offer 1 and we could get 160 profit. However, the product manager of Offer 2 will need some prospects in order to reach his sales goals, so he will claim a minimum amount of customers for his campaign. To fulfill these new requirements, we follow a hierarchical process were we choose for each offer, the best customer. If we start by offer 1, then we would get 115 profit, as each customer can only receive one offer.

The optimization algorithm

None of these approaches is the optimal one. By working with an optimization algorithm that considers all the possible combinations regarding the constraints, we could get 150 profit (offer 2 for customer 1 and offer 1 for customer 2).

Extrapolate now this problem to your real life situation with millions of customers, dozens of campaigns, several channels, and limited marketing budget. The possible combinations are huge. We can only solve it using advanced algorithms.

Did your eyes glaze over after reading the word algorithm above? Mentioning the word algorithm in a marketing environment can be discouraging. Have no fear - marketers do not need to have programming skills or mathematical knowledge in order to optimize their campaigns. By using SAS Marketing Optimization, campaign managers can easily find the best solution for their problem without needing to code.

Optimization is a journey

 

The journey toward direct marketing optimization.This is the first of a series of blog posts where I will cover all the necessary steps to reach an optimal campaign solution. I will start by explaining how to understand the potential of each customer for each offer using basic and advanced analytics. Then I will talk about how to design campaigns for an optimization process, as this will probably mean some cultural change. After, I will discuss about the optimization objective. Should we maximize revenue? Should we minimize costs? Or should we do both? We only need to optimize if we have restrictions, so I will write about the kind of constraints we can handle and the most common ones we see in different industries. With all this, the final article will be about how to put everything together.

Stay tuned for the next step in this journey toward optimization - it's the way to better marketing.

tags: analytics, marketing optimization

The journey toward direct marketing optimization was published on Customer Analytics.