Guest Post: Kevin Hillstrom
Published by Eric T. Peterson on July 19, 2010.
|« Back to all posts||Share, Save or Email|
Kevin Hillstrom is one smart dude. President of MineThatData, author of Online Marketing Simulations, and prolific contributor to the Twitter #measure channel. Kevin spends a huge amount of time in Twitter challenging web analysts to think and work harder on behalf of their “clients,” 140 characters at a time.
A few weeks ago I asked Kevin “what five practices learned in the offline data analytics world would you like to see web analytics professionals adopt?” The following contributed blog post has Kevin’s answers which are, unsurprisingly, awesome. Near the end Kevin says “The Web Analyst has the keys to the future of the business, so it is a manner of getting the Web Analyst to figure out how to use keys to unlock the future potential of a business.”
Brilliant. We are the future of business … so what future will we be helping to create?
Kevin Hillstrom, President, MineThatData
In 1998, I became the Circulation Director at Eddie Bauer. Back in those days, Eddie Bauer printed money, generating more than a hundred million dollars of pre-tax profit on an annual basis.
One of the ways that Eddie Bauer generated profit was through the use of discounts and promotions. If a customer failed to purchase over a six month period of time, Eddie Bauer applied a “20% off your order” offer. The customer had to use a special promotion code, in order to receive discounted merchandise.
We analyzed each promotion code, using “A/B” test panels. Customers were randomly selected from the population, and then assigned to one of two test panels. The first test panel received the promotion, the second test panel did not receive the promotion. We subtracted the difference between the promotion segment and the control segment, and ran a profit and loss statement against the difference.
In almost all cases, the segment receiving the promotion generated more profit than the control segment. In other words, it became a “best practice” to offer customers promotions and incentives at Eddie Bauer. Over the course of a five year period of time, the marketing calendar became saturated with promotions. In fact, it became hard to find an open window where we could add promotions!
Being a huge fan of “A/B” testing, I decided to try something different. I asked my circulation team to choose two customer groups at random from our housefile. One group would receive promotions for the next six months, if the customer was eligible to receive the promotion. The other group would not receive a single promotion for the next six months. At the end of the six month test period, we would determine which strategy yielded the most profit.
At the end of six months, we observed a surprising outcome. The test group that received no promotions spent the exact same amount of money that the group receiving all promotions spent. After calculating the profitability of each test group, it was obvious that Eddie Bauer was making a significant mistake. It appeared that we would lose, at most, five percent of total annual sales, if we backed off of our promotional strategy. Eddie Bauer would be significantly more profitable by minimizing the existing promotional strategy.
In 1999, we backed off of almost all of our housefile promotions. At the end of 1999, the website/catalog division enjoyed the most profitable year in the history of the business.
This experience shaped all of my subsequent analytical work.
Just because we have the tools to measure our activities in real-time doesn’t mean we are truly optimizing business results. In the Eddie Bauer example, we had the analytical tools to measure every single promotion we offered the customer, and we used existing best practices and “A/B” testing strategies. All of it, however, was wrong, costing us $26,000,000 of profit on an annual basis. Simply put, we were measuring “conversion rate”. What actually happened was that we “shifted conversions” out of non-promotional windows, into promotional windows! Had we measured non-promotional windows, we would have noticed that demand decreased.
So, by measuring customer behavior across a six month period of time, we made a significant change to business strategy, one that dramatically increased annual profit.
What does this have to do with Web Analytics?
The overwhelming majority of Web Analytics activity is focused on improving “conversion rate”. Our software tools are calibrated for easy analysis of events. Did a visitor do what we wanted the visitor to do? Did a promotion work? Did a search visitor from a long-tail keyword buy merchandise when they visited the website? All of these questions are easily answered by the Web Analytics expert, the expert simply analyzes an event to determine if the event yielded a favorable outcome.
Offline analytics experts (often called “Business Intelligence” professionals or “SAS Programmers” if they use SAS software to analyze data) frequently analyze business problems from a different perspective. They use whatever data is available, incomplete or comprehensive, to determine if the individual actions taken by a business over time cause a customer to become more loyal.
With that in mind, here are five offline practices I wish online analytics experts would adopt.
Practice #1 = Extend the Conversion Window: Instead of analyzing whether a customer converted within a single visit or session, it makes sense to extend the conversion window and learn whether the customer converted across a period of time. For instance, when I ran Database Marketing at Nordstrom, we learned that our best customers had a 5% conversion rate, when measured on the basis of individual visits, but our best customers nearly achieved a 100% conversion rate when combining website visits and store visits during a month. By extending the conversion window, we realized that we didn’t have website problems, instead, we had loyal customers who used our website as a tool in a multi-channel process.
Practice #2 = Measure Long-Term Value: Offline analytics practitioners want to know if a series of actions results in long-term profit. In other words, individual conversions are relatively meaningless if, over the course of a year, individual conversions do not yield incremental profit. This is essentially the “Eddie Bauer” example I mentioned at the start of this paper, we learned that individual conversions (customers purchasing via a promo code) yielded increased profit during the promotional period, but generated a loss when measured across a six month timeframe. A generation of Web Analytics experts were trained, largely because of software limitations, to analyze short-term business results, and have not developed the discipline to do what is right for a business across a six month or one year timeframe. Fortunately, Web Analytics practitioners are exceptionally bright, and are easily able to adapt to longer conversion windows.
Practice #3 = Comfort with Incomplete Data: I recently analyzed data for a retailer that was able to tie 70% of store transactions to a name/address. During my presentation, an Executive mentioned that my results must be inaccurate, because I was leaving 30% of the transactions out of my analysis. When I asked the Executive if it would be better to make decisions on incomplete data, or to simply not make any decisions at all until all data is complete and accurate, the Executive acknowledged that inferences from incomplete data are better than inaction caused by data uncertainty. Offline analysts have been dealing with incomplete multi-channel data for decades, and have become good at communicating the benefits and limitations of incomplete data to business leaders. The same opportunity exists for Web Analytics practitioners. Don’t hide from incomplete data! Instead, make confident decisions based on the data that is available, simply communicating what one can and cannot infer from incomplete data.
Practice #4 = Demonstrate What Happens to a Business Five Years From Now Based on Today’s Actions: Believe it or not, this is how I make a living. I use conditional probabilities to show what happens if customers evolve a certain way. Pretend a business had 100 customers in 2009, and 44 of the 100 customers purchase again during 2010. This business must find 56 new customers in 2010 to replace the customers lost during 2010. I can demonstrate what the business will look like in 2015, based on how well the business can retain existing customers or acquire new customers. This type of analysis is the exact opposite of “conversion rate analysis”, because we are looking at the long-term retention/acquisition dynamics that impact every single business. I find that CEOs and CFOs love this type of analysis, because for the first time, they have a window into the future, they actually get to see where the business is heading if things remain as they are today. Better yet, the CEO/CFO can go through “scenario planning” to identify ways to mitigate problems or to capitalize on favorable business trends. The Web Analytics practitioner has the data to do this type of analysis, it is simply a matter of tagging customers or shaping queries in a way that allows the analyst to make inferences that impact long-term customer value.
Practice #5 = Communicate Better: This probably applies to all analysts, not just Web Analytics experts. Executives are frequently called “HiPPOs” by the Web Analytics community, a term that refers to “Highest Paid Person’s Opinion”. The term can be used in a negative manner, suggesting that the Executive is choosing to not make decisions based on data but rather on opinion or gut feel or instinct or internal politics. I was a member of the Executive team at Nordstrom for more than six years, and I can honestly say that I made far more decisions based on opinion than I made based on sound data and analytics … and I am an analyst by trade!! Too often, the analytics community tells an incomplete story. Once, I witnessed an analytically minded individual who made a compelling argument, demonstrating that e-mail marketing had a better return on investment than catalog marketing. This analyst used the argument to suggest that the company shut down the catalog marketing division. On the surface, the argument made sense. Upon digging into the data a bit more, we learned that 75% of all e-mail addresses were acquired when a catalog shopper was placing an online order, so if we discontinued catalog marketing, we would cut off the source of future e-mail addresses. This is a case where the analyst failed to communicate in an appropriate manner, causing the Executive to not heed the advice of the analyst. Too often, analysts fail to put data and customer findings into a larger context. Total company profit, long-term customer profitability, total company staffing strategies and politics, multi-channel customer dynamics, and Executive goals and objectives all need to be taken into account by the analyst when communicating a data-driven story. When this is done well, the analyst becomes a surrogate member of the Executive team. When this is not done well, the analyst sometimes perceives the Executive to be a “HiPPO”.
These are the five areas I’d like to see Web Analytics experts evolve into. The Web Analyst has the keys to the future of the business, so it is a manner of getting the Web Analyst to figure out how to use keys to unlock the future potential of a business. Based on what I have witnessed during the past forty months of multi-channel consulting, I am very confident that Web Analytics practitioners can combine offline techniques with online analytics. The combination of offline techniques and online analytics yields a highly-valued analyst that Executives depend upon to make good business decisions!
About Eric T. Peterson
Eric T. Peterson is the founder of Web Analytics Demystified, Inc. and the author of Web Analytics Demystified, Web Site Measurement Hacks, and The Big Book of Key Performance Indicators. Mr. Peterson frequently presents on web analytics, is often cited in articles about digital measurement, and has been blogging on the subject since 2004.
Want to speak with Eric? Contact Web Analytics Demystified