You never know

One of the reasons I enjoy being a member of a book club is that I get introduced to books I would not have chosen myself.  The current book club selection is Flight Behavior by Barbara Kingsolver which is the story of what happens to the residents of a small Appalachian town when Monarch butterflies unexpectedly migrate to their town.  Imagine my surprise when I read an excellent definition of the old adage, “correlation is not causation” in this novel.

The book before this one was A God in Ruins by Kate Atkinson.  The term chi square automated interaction detection appears in this novel about a World War II Halifax Bomber pilot, Ted Todd.  It is usually abbreviated to CHAID.  For more information on the technique, click here.

I wonder what will be in the next book!

The direct marketing channel war revisited

These days the debate over direct mail versus email seems to be over.  The conventional wisdom is that direct mail is too expensive and takes too long.  If a retailer has had bad sales over the weekend, they want to take action now and not wait a month or two to get a direct mail piece delivered to their customers’ mailboxes.

However, this approach could ignore some valuable customers.  What about your customers who are not emailable either because you don’t have their email address or they have opted out of email communications?  Also, sending an email doesn’t mean it will actually be seen by the consumer.  Google’s use of the promotion inbox makes it easier for consumers to ignore marketing communications.  In addition, plenty of people have secondary email addresses that they use just for these types of communications and which they check only rarely.

In addition, there is the question of whether email is always the best channel for the message.  A recent study found that physical ads were better than digital ads in some respects.  See here: http://www.dmnews.com/postal/direct-mail-has-a-greater-effect-on-purchase-than-digital-ads/article/423292/

In the end, it may be a multichannel strategy that works best for you.  Through a test and learn approach you can determine what generates the best return on your marketing investment.

Take a minute to consider these questions

Big data has been a hot topic for several years and for good reason. There is value in analyzing unstructured, high volume and massive data sets. However, when I interview candidates that say they want to be data scientists, they focus on the technology and techniques.  They forget that the critical thinking and framework used for big data is also important and it is applicable to many types of analytic projects.

It comes down to some very fundamental questions:

  1. What problem am I trying to solve? Defining the problem up front will keep you grounded as interesting findings may lure you away from your goal.
  2. What data sources can I use? You want to consider multiple sources to triangulate your results and provided a richer picture of what is happening.
  3. Have I considered all the possible sources of bias? Bias of all sorts can skew results and must be considered and incorporated into your analysis plan.
  4. Do I need to use all the data available or will a sample be sufficient? There are times when it is not feasible or necessary to analyze all the data available. However, if you sample, you need to make sure that you are getting sufficient coverage and that your sampling is random.
  5. How can I validate my data? Validation must be part of your analytics plan, whether you validate one data set against another or at least compare your results to findings from other comparable projects.
  6. What analytic technique(s) are appropriate? Consider the pros and cons of various techniques and what would be most appropriate given the data and problem at hand.

While it is very tempting to dive straight into the data and analysis.  Spending time up front to answer these questions will help you be more efficient.

What is big data?

I am often asked, what is big data? It happens at holiday parties and even once after a funeral. Certainly there have been large data sets before. So what is different now? Big data commonly refers to data that is so large that you cannot use the typical environments to store and manage it or the typical software to analyze it. In addition to volume, big data is often defined by velocity and variety. Velocity refers to the speed at which the data is available and big data typically includes frequent inputs.  Variety refers to the diversity of sources and formats and big data typically contains unstructured data which is not easily categorized or organized.

The volume of big data requires new thinking about where to put the data.Traditionally, companies kept their data in house, in a data warehouse on an internal server.Now some companies are turning to the cloud, both private and public clouds, to house data because of its size.In addition, the cloud offers flexibility should the needed storage capacity grow.Similarly, the volume and variety of the data may make it impractical to load the data into a database for to do so would require assigning data elements to tables and fields.Some big data may not be easily structured.For example, it could be text messages from online customer service chats.In this case, companies might turn to a parallel programming framework such as MapReduce to capture the data.This enables them to load all the data and then parse the text of the on-line service chats to identify the frequency of words used.For example, how many customers reported a problem with a particular part or described themselves as frustrated.However, you can’t use SAS or SPSS to analyze the data in a MapReduce environment.Further, data mining techniques may be more useful than classical statistics because of the nature of the problem to be solved.Thus, almost everything about big data requires rethinking data and analytic tools.

However, in the end, big data is like all data.   It must generate value. Big data is meaningless unless it enables companies to increase revenue and/or reduce costs by enabling them to identify insights that were previously unavailable. The power of big data is that analysts can explore larger data sets that were impossible to analyze before and delve into unstructured data that was typically ignored because of its non-conforming format.

What I am thankful for

Thanksgiving Day is a time that I reflect on all the things I am thankful for and the increased emphasis on analytics is one of them.  With a struggling economy businesses want to know the value they are receiving from their marketing dollars.  Analysis can help them determine their ROI and optimize marketing efforts.  Increasingly, companies are looking at their wealth of data as a valuable asset that can drive revenue growth.  With data mining techniques, companies can identify hidden trends and insights that can lead to new customer segments or new product offerings.  Lastly, increases in technology have made it easier and easier to delve into data and display the results visually.  Thus, making it easier to uncover value in your data.

At a recent road race someone asked me what I did for work.  When I told her, with a smile, that I optimize marketing efforts through analysis, she remarked that it was nice to meet someone who enjoyed their job.  I enjoy what I do, in part, because I believe in the value of analysis.  It is very rewarding that others are recognizing that value as well.

Don’t forget the intangible skills

While everyone seems to be looking for data scientists these days, I have been interviewing Analyst and Sr Analyst candidates.  The focus lately has been on technical skills and rightly so.  Data scientists, for example, need to be able to transform raw data into analysis and actionable insights.  That may require experience with Java as well as data mining techniques, an unusual combination.

However, my interviews have reinforced the value of the intangible skills:  creativity, commitment, and curiosity.  The best candidates provide examples of how they have solved a problem creatively.  One candidate described a model he created to deal with missing data.  It is rare to have perfectly clean, comprehensive data.  Being able to overcome data issues is an important analytic skill.

I also look for candidates who are committed to providing the best possible analysis.  Clients pay me to solve difficult problems they can’t solve themselves.  They entrust me with their data and I take that trust very seriously.  Thus, I want everyone I work with to do the same.  While it can be difficult to assess commitment during the interview process, there are telling cues.  One candidate asked in advance about her interviewers, came prepared to her interviews with excellent questions that indicated she had researched the company and quickly sent thank you notes after every interview.  Her attention to detail and follow through made an excellent impression and spoke louder than words.

I always start an analysis with a hypothesis and a project plan.  However, sometimes during the course of an analysis, you find something interesting that changes your plan or analysis.  At other times, you may have a hypothesis that turns out to be false when pilot tested.  One candidate gave a presentation and at the end admitted that the marketing program she had developed did not generate incremental revenue as expected.  I liked that she presented on a program that did not perform as expected.  The only way to innovate is to remain curious and willing to test hypotheses.

While technical skills are important, don’t underestimate the intangibles.  I can teach someone SAS.  It is much harder to teach someone to be curious about data.

Happy New Year

The new year has begun. Now is the time to measure the success of your holiday campaigns. How did your campaigns perform? This is an opportunity to look at their effectiveness in terms of building awareness, generating revenue, increasing retention and aiding customer acquisition? How do your metrics compare to industry benchmarks as well as internal benchmarks? How much revenue did they generate and were they profitable? In addition, what worked and what didn’t? Now is the time to evaluate any tests that were done – date/time, subject line, creative, etc. Finally, compare the results of this past holiday campaign to the one before and analyze the differences. The insights from the holidays can inform your strategy for 2012.

I am dreaming of a white Christmas

Even though it still feels like summer outside, now is the time to start planning for the holidays.

The first step is to evaluate all of the tests that have been done throughout the year in order to put your best foot forward.  In addition, it involves reviewing the results from the prior holiday season.  That means determining the most effective:

  • communication method (e.g., email, direct mail, multi-channel) by customer segment
  • timing (both day of the week and time of day)
  • creative (hero images, placement of links, etc.)
  • subject lines (when and where to mention free shipping offers, brand or product offers, etc.)
  • offers (discount percentages, dollars off, buy one get one free)

Next step is to evaluate any implementation issues from the prior holiday season.  Before coming up with your holiday strategy it is important to determine any limitations or challenges with respect to execution.  Your strategy cannot be developed in a vacuum.  Thus, I recommend that you review what has worked and what did not work with the entire team.

Once all of this information has been gathered, you can develop a holiday strategy.  It should incorporate the lessons from past tests and holiday campaigns as well as encompass:

1.  Start Date. The average holiday campaign begins in October.  Some retailers hold pre-holiday clearance sales and send informational emails to start their holiday campaigns.

2.  Black Friday. For Marketers, the holiday campaigns have been starting earlier and earlier on the calendar.  The same is true for Black Friday.  It is now beginning on Thanksgiving Day for some retailers.  When will yours start?

3.  Cyber Monday. While many digital sales are made on the Monday after Thanksgiving, digital sales are occurring earlier as consumer shop from home.  Will you wait for Cyber Monday or start earlier?

4.  Sequence. If you are using email, you can easily send at least an email a day.  It is important to determine the contact frequency and cadence.  Will all or a segment of your customers receive an email a day, every other day, every third day, etc.?  Will emails be sent only on weekdays or only weekends or a mix?  Will there be a resting period or a maximum number of emails that can be received?

5.  Free Shipping.  Many consumers expect to get free shipping online, especially during the holidays, and will not pay for shipping.

6.  Social Sharing.  Consider how to tie in Facebook, Twitter and other social sites with your campaign.

7.  After Christmas. Lastly, there is also the opportunity for follow on sales after Christmas.  It is the time to promote use of gift cards and purchases of parts or refills.

The value of a loyalty program

As I mentioned in my prior post, loyalty programs are a valuable tool.  They can help retain customers and companies can win greater share of wallet as a result.  If a customer can buy the same goods or services from multiple sellers, a loyalty program encourages customers to consolidate their purchases.  It might also create additional demand.  For example, a reward certificate can spur an incremental trip or customers may splurge in order to meet a spending threshold.

Another benefit of loyalty programs is the insight into customer behavior.  This has far reaching benefits.  Take the example of a retailer.  This customer insight can help both marketing and merchandising.  Using the data collected, a retailer can segment their customers based on past behavior so that they can tailor their messages and offers appropriately.  For example, marketers can use this information to personalize product promotions, cross-sell products and identify new customers that have the potential to become to best customers.

Further, this data will provide insight into what products bring new customers into the store, what products drive repeat purchases and what products are typically purchased together. Merchandisers can use this information to plan promotions and make buying decisions.

To be valuable, the data must drive actionable insights and be used to continually improve the loyalty program.  I will write about using data to evaluate the health of a loyalty program in my next post.

How do you prioritize new customers?

I was asked recently how to prioritize new customers if you do not have demographic or firmographic data available.  In other words, what can you do with just the data from the first purchase with which to work?

To make this more concrete, let’s consider the following situation.  You are asked to call each and every new customer who has made a purchase.  The question is, how do you prioritize the calls?  You want to make the first calls to those with the greatest potential to become loyal and valuable customers.  The only data available relates to the first purchase:  total revenue generated, products purchased,  product revenue, etc. 

In this case, a linear regression could be used to help you identify the factors that predict lifetime value.  (Other types of models can be used depending on the independent and dependent variables available.)  Using your existing customer base, build a model that leverages data about the first purchase to predict lifetime spending.  You can identify the best and worst new customers using the resulting model equation.  Armed with this insight, you can test your model by calling on new customers with the best predicted lifetime revenue and a random selection of new customers regardless of predicted lifetime revenue.  In addition, you can test call back timing to determine if there is an optimal call back window. 

Even with limited data, analysis can lead to insight.  Further, there is always an opportunity to incorporate testing.  In this case, testing can validate initial findings and help you learn more about the purchase cycle.