Three implications of mathematics on the World Series

Sabermetrics provides useful advice for modern data science

Many of my friends who live in the American Midwest are captivated by the Major League Baseball World Series, not only for the drama of two teams with a rich history and many years of waiting for this opportunity, but also because it provides a welcome respite from news about our presidential election. While the Oakland Athletics, the baseball team for whom I root, will be waiting for a while before our next World Series appearance, I’ll share my enthusiasm for the Indians and the Cubs fans, and one of my favorite sports.

I remember, as a college student, being enthralled with reading the box scores every morning in the San Francisco Chronicle, watching the ERA of the pitching squad, and monitoring the amazing batting average of Carney Lansford, the A’s third baseman at the time. This experience of reading the box scores, and how they changed day to day, influenced the way I watched the game. I was thrilled when Michael Lewis published his book “Moneyball” in 2003, focused on the emerging role of analytics and how it shaped the strategy for the very same Oakland A’s that I loved and rooted for.

Read More

Three reasons why email marketing isn’t as free as you think

Contact strategy optimization for digital channels

I’ve recently interviewed retailers active in digital marketing to learn more about their analytics practices. In three distinct interviews this week, I learned that there is a consistent pattern of treating digital outbound communications as an essentially cost-free channel.
  • Specifically, these retailers act as though expected sales revenue will increase as the number of outbound promotional customer emails increases, without any boundary or penalty for over-contacting their customers.
  • These retailers also treat customers as largely fungible; in short, every customer gets the same contact policy, measured by the number of direct emails they are eligible to receive each week. I didn’t get the impression that these retailers actually believe that customers are fungible, only that in practice, there is no differentiation across customers given their historical behavior or any measurable performance data.
  • Finally, there are broadly-held and deeply-rooted perceptions amongst these retailers’ brand and product managers that any change in business as usual will dilute sales revenue and can’t be tolerated.
In contrast, we’ve found that retail customers are not fungible, that there are real penalties for over-contacting customers (some short-run and some long-run), and that practical solutions exist for tailoring the number and mix of promotional contacts per customer without sacrificing sales revenue. In fact, it’s likely that optimizing the mix of communications will grow sales revenue.
If your organization isn’t in the retail sector, there are plenty of analogous examples in other vertical industries. Consider the number of times that a brokerage financial advisor contacts their clients every month with portfolio rebalancing opportunities or active trading alerts, which in our experience, varies from daily, to once or multiple times weekly, to once monthly, and sometimes not at all.

Here is the way we’ve seen sales revenue dynamics actually work when augmented by measurable evidence from customer behavior. The first expectation (which is not usually supported by actual performance) is that sales revenue will increase as you continue to increase the number of contacts per customer per week. Some retailers place an upper bound on the number of contacts per customer per week, and others also place a floor on the minimum number of contacts for some classes of high-value customers. Nonetheless, the widely-held practice is that since the variable cost of an email is virtually zero, you may as well send as many emails as you can, to maximize total contacts and to reduce the average fixed costs of email communications.

cso-fig1

Expected sales given varying number of contacts per week

 

Working with one of the country’s largest apparel retailers, we found that there are very real costs to over-contacting customers. Short-run revenue losses can result from contact fatigue, by reducing the customer’s interest in opening or clicking through the email to a landing or promotional page. Long-run revenue losses can result from opt-out activity. Some enterprising retailers have introduced “opt-down”, which actively asks the customer to receive at least one email per week, but not to leave the retailer entirely. (Why don’t they ask this of every customer?)

Read More

Five Actionable Lessons Learned from Predictive Customer Journey Analytics

I’ve been actively researching customer behavior in the financial and retail industries for over 30 years now, and what fascinates me the most is determining the leading indicators of future customer choices and interactions. At Corios we call this predictive customer journey analysis.

journey noun_81349The most important and most elusive part of being responsive to customer journeys is identifying leading indicators. Leading indicators will tell you what each customer is likely to need or want, in advance of the customer making a decision, in a way that gives the firm enough time to prepare and respond with a relevant solution.

Predicting future customer activity is more important than merely observing past activity. If used correctly, history can be a wonderful predictor of future choices. However too often, marketing recommendations are based on customer averages, rather than on trends or events. The customer average is a good predictor if we have no other information, but in this era, we have much more information available to help predict the customer’s next actions. Being aware of, and utilizing this information effectively, is vital.  Often firms have one shot at delivering a relevant solution within a finite window of opportunity. Once that window closes, another opportunity may not present itself for days, months or years.

To underline the importance of effectively tackling the challenge of predictive customer journeys, I’ve researched the actions of our Corios clients, and catalogued the five practices that lead to the biggest differences between successful and less-than-successful efforts.

 

1) Set predictable goals, and monitor progress towards achieving them

Do you remember the era of web server logs and vendors like WebTrends? Amazingly, they are still in business, a few blocks away from Corios HQ in downtown Portland. When their software first emerged, it was a very solid way of reporting web server statistics: pages, hits and errors (the reports always reminded me of baseball scores.) However, what has always been missing from these reports was any indication of whether the firm using the web server was actually making any money from customers. A problem that is still relevant now. Business owners receive too many reports providing historical perspectives on activity, rather than whether these activities moved the needle by delivering meaningful value to their customers.

Today, these activity reports take the form of interactions across channels, hopefully integrated to the customer level, but the problem still remains that none of these reports inform decision-makers as to whether key objectives are being met, improved, or diluted, as the result of said activity.

Which brings me to my first “difference-making” action when considering customer journey analysis: Can the firm’s decision-makers trace a direct line from interactions to value-producing objectives? The path between interaction and valuable objective should go straight through an applied analytic model that identifies which interactions produce value, such as incremental closed sales or increased value per sale, as opposed to being completely irrelevant.

When this line of sight between activity and outcome is clear, unambiguous, and a leading indicator, decision-makers can redirect resources toward stimulating the most value-producing interactions with customers.

Read More

What does predictive customer journey analytics look like?

Walking through a predictive customer journey dashboard and results

When will a customer take action? How rapidly will they respond to a bank offer for credit, or to a retailer’s offer for a sales promotion? If you don’t know the answers to these questions, you can’t really develop a sound strategy for delivering the right offer to the right customer at the right time. This is what predictive customer journey analytics are all about.

Many of our clients are visual thinkers, so we wanted to provide everyone a visual walkthrough of what predictive customer journey analytics looks like to a business decision maker. We built an illustration and published a narrated video of Corios Veloce at work, using results from a recent engagement as our inspiration. It’s a short, 10-minute exploration of the new insights decision makers can use to determine when to engage with a customer at the best time, using leading indicators from transaction and interaction data.

 

Read More

Customer behavior is neither simple nor linear

Especially determining the onset of changing customer behavior

When will a customer take action? How rapidly will they respond to a bank offer for credit, or to a retailer’s offer for a sales promotion? If you don’t know the answers to these questions, you can’t really develop a sound strategy for delivering the right offer to the right customer at the right time.

As you might already know, at Corios, we tell the story numbers can’t. Telling the story of the consumers, using their transactions as the ink on the page, meant developing an entirely new way of analyzing data, and it stems from looking at consumer behavior differently than the traditional conventions of applied mathematics and big data.

Read More