Sudoku got bigger last month. You may not have been able to carry on your toothpaste or drinks on your last flight, but many tried out sudoku puzzles or helped the person next to them work the solutions. (One reader asked me to tell everyone to please get his own sudoku book or software.) I can’t believe how many people, in the month since my previous column, ”Lessons from Sudoku,” related stories of their brushes with sudoku.
In that article I discussed the eerie coincidence between customer and marketing analytics and sudoku. One of the successful techniques for quickly solving sudoku puzzles is to test the relevance and appropriate usage of logical solutions and then project the “impact” on the entire puzzle. That’s simply how you get customer and data-driven analysis with the numbers you have today.
Relevance
The key to ensuring that analytics is relevant is to fully align the techniques, tools, and data you use with the business problem you are trying to solve. Applying a state-of-the-art technique to an issue that would be better addressed using a more straightforward approach is making the means more important than the end. It’s ineffectual.
One leading national bank was grappling with customer-retention challenges as well as potentially big opportunities in customer acquisition. Its internal analytics team was advocating the application of multiple advanced techniques, some of which had not been proven in that environment before. The approach could best be described as “this is a big, hairy, important problem requiring impressive analytics.” The executive was skeptical and overwhelmed. Would he ultimately get an answer that was technically elegant but of little practical value?
Our approach was to 1) quickly understand the business challenges and opportunities, 2) use simple and proven predictive modeling techniques, 3) add a dashboard depiction of results that the executive’s team could use to track progress and learning, and 4) recommend specific strategy and tactics to target the right customers for retention and acquisition. By proposing a simpler but more relevant and aligned approach, we ensured that relevance drove results:
• assisted in identifying an $89 million yearly direct mail “mover” retention opportunity for the bank
• doubled direct mail response rates
• enabled testing and learning for more-focused customer and prospect targeting.
Impact
At its heart, predictive modeling is about enabling prioritization among poor, marginal, and superior marketing investment opportunities for topics such as customer relationship development. Key to making that happen is the ability to effectively deploy analytic techniques that produce measurable impact.
A multinational hotels company had historically engaged in a one-size-fits-all approach to marketing even though it had very rich data on customer stay patterns. It was considering many analytic techniques and direct marketing tactics to unlock the potential in stay patterns. We conducted a very tightly scoped and rapidly executed analysis that involved highly focused predictive modeling techniques to put attention only on the most important insights from its customer data. The impact was quick and sharp:
• The stay-propensity models ranked the hotel’s guests in terms of following-year stay potential.
• The models predicted that focusing on approximately 20% of hotel guests would address almost 90% of the five-plus stay opportunities in the following year.
This kind of actionable and high-impact insight represented a revolution for our client, which acted quickly to use our models to capitalize on this opportunity. Marketing spend was shifted and performance was better than ever.
Usage
Appropriate usage of analytics techniques is a key part of achieving the kind of impact demonstrated above. Used properly–meaning based on a thorough understanding of the business opportunity combined with the right data selection and preparation–analytic techniques can create huge opportunities. Used inappropriately, these sophisticated approaches lead to a lot of wasted time and, worse, potentially misleading business direction.
We recently worked with a travel organization whose analytics-based approach to customer retention had, it thought, worked very well over the years. But returns were diminishing, and it didn’t know how to improve them. Upon our closer examination of the company’s use of analysis in marketing decisions, we found a key off-target definition of attrition combined with an inappropriately executed program of predictive modeling was masking a significantly larger attrition problem.
We quickly applied proven and relevant techniques encompassing a new definition of attrition behavior and defined new attrition metrics. The client is now implementing new marketing programs that are stemming attrition rates.
My colleague Niall Budds, the vice president heading up our analytics practice, collaborated with me this month. He and I purposely avoided formulas or technical definitions of the statistical and modeling techniques used in all three of these examples. We’d be happy to share that information with you, though. Believe me, Niall has it all in gory detail.
Here’s what we’d like you to ponder:
• Conducting analysis does not always mean getting or waiting for more data.
• Know that the appropriate, simplest possible analysis and modeling techniques are being applied so that you can make timely, sound, and predictable decisions.
• Institute the discipline to use analysis in critical stages of marketing workflow and in some way with every marketing team member. This may just be your key source of competitive advantage in the very near future.
Anthony Power, whom I referenced in my column last month, has a new article entitled “Analytics Without Numbers: Perspectives Learned from Sudoku” on www.MultichannelMerchant.com (a sister Website to www.ChiefMarketer.com). The article is a good read on putting analytics in perspective—and it should keep the puzzle craze accelerating!
Lane Michel is executive vice president/managing director of the Marketing Performance Management business unit of Quaero Corp., a marketing and technology services provider based in Charlotte, NC. You can reach him at [email protected].
Network
