Chapter 15
Key Contextual Data Is Not in Your Company’s Databases
Oil is finite, but information is infinite.
—ERIC SCHMIDT, FORMER CEO OF GOOGLE
Life is infinitely stranger than anything the mind could invent.
—SIR ARTHUR CONAN DOYLE
A typical road to failure comes from a strategy that says, “Let’s build a really big data warehouse, then build our pricing capability and strategy from that base.” This plan simply does not work.
Why? Because no matter how large the database is, it will rarely capture the essential data required for pricing. The reasons for this are twofold:
The data in the data warehouse is captured for other purposes.
The likelihood that this data will also happen to support pricing strategy or pricing maintenance is quite low.
The former point doesn’t mean that information cannot serve multiple purposes. Rather, it reveals that if it is not gathered for pricing, the applicability will be coincidental.
Consider the purposes of different databases that are the typical building blocks of a data warehouse:
The bulk of data available in corporate systems is captured for accounting purposes, leading to corporate profitability reporting and dividends. Yet customers do not care about your profitability.
Some data is captured for administrative and human resources, which again is of little interest to buyers and potential buyers.
Customer support purposes seems like the right idea, but much of this data is centered on customers after the purchase and does not focus much on “Why did customers pay this price?”
Data may include market price information, but until recently pricing has been the poor relative at the marketing family reunion. Most of marketing is focused on product, promotion, and channel—not pricing.
Lately, some data has been gathered in systems for sales operations and sometimes for customer prospecting. Often this can be good, but rarely is it complete or compelling insight into customer-decision processes—that is generally left to the individual sales representative.
Companies cannot depend on coincidental overlap between pricing needs and other demands to ensure managers have necessary pricing inputs.
Pricing depends on understanding the customer and the customer’s decision process and decision context. That information, unlike, for example, cost information, is not usually systematically gathered (although it should be known to managers). One reason is that internal priorities like accounting take precedence over pricing requirements. Also, managers often do not develop specific requirements for essential pricing data because specificity is challenging, without prior investigation.
The Impact of Competition
Contextual pricing factors were discussed in Chapter 3, and you may have noticed that much of this data was not in your company’s database. As a reminder, however: one of the most important contexts is competition. What data does your company have systematically about its competitors? Can it be arranged or parsed by specific market? By product overlap? By channel? In most cases the answer is no. Even very large, smart, and well-resourced companies have limited competitive data.
Contextual pricing, unlike cost information (for example), is based on data of which management is aware, but it is usually not systematically gathered. Managers need to develop concrete requirements for essential pricing data rather than random “grabs” of data in hopes that it will suffice.
The impact of competition on pricing is quite pervasive. Having robust information on context means you have available extensive and systematic information on competitors. For instance, frequently even the number of competitors that operate in each geography is not tracked. That is unfortunate because often the number of competitors can be a good approximation of “supply” to a market, and hence one half of the supply-and-demand equation.
Sadly, most companies tend to not keep systematic records of competitor price initiatives, price levels, product coverage, or geographic or competitive segmentation. In fact, one large consumer goods company specifically prohibits its competitive intelligence unit to keep on file anything that is not public—which of course is a small subset of the total information available.
The key point here is that the mere fact that your company “has a lot of data” is rarely adequate for pricing purposes. The amount of data in the world is literally infinite. That means that on a coincidental level, no matter what you have in your data warehouse, the chances of you having what is needed is virtually zero. To put it graphically, many managers believe the image below represents their likelihood of having the pricing data they need. This assumption is wrong. If drawn to scale, each of the circles would be invisibly small, and the chance of the data in each coincidentally overlapping with required pricing is literally zero. Thus, do not visualize your data as represented in Figure 15-1.
Figure 15-1 Mistaken conception of information: Venn diagram of how many managers view information as finite, and how they wrongly think building a big database will address contextual data needs. Planning is necessary.
A company database can have many terabytes of data but still not inform any of the key market-price drivers. Big divided by infinity is zero.
A Balanced View of the Information Needed
Of course, there is more than mere coincidence driving availability of data. Yet while management is no doubt working to see that the best information is captured, specifying the right pricing information is sometimes fiendishly difficult, and that difficulty is often dismissed when resources are being allocated. Market forces are often obscured without careful study. The trouble is that pricing information is different from that organically supplied by a company’s operations.
What adds to the imperative for gathering of the right categories of data is that the same data can be viewed in different ways. For instance, college test results can be viewed in absolute terms (A, B, C, etc.) or they can be viewed as comparisons (top 5 percent, top 25 percent, etc.). Very different analysis results from these different views of the same data: one is subject to “grade inflation”; the other is immune. In the same way, pricing is often based not only on product effectiveness but also on the relative effectiveness of the product compared with some baseline, such as competitors or a previous generation of product.
Knowing what are the likely contextual comparison points helps. This reinforces the need for a focused gathering of price data. Hypothesizing the likely contextual needs is necessary because information can be examined in many ways, possibly an infinite number of ways.1 Let’s take a look at another nonpricing example: recent presidential election results are susceptible to different analysis, depending on whether they are viewed at a voter, state, ethnic group, or national level.
At a state level, linkages between results and demographics (e.g., education, income) are clear, but at another level those relationships may not be clear or not show reasonable confidence levels. For instance, a recent Republican candidate may have systematically captured the votes of wealthy age groups (ages 40 to 50), yet systematically lost the vote in wealthy states (Connecticut, California, etc.). So is wealth a good determinant of voter trends? That depends on your intended use.
Collecting contextual data is subject to the twin requirements that it cannot rely on coincidental collection and that it cannot be specified too narrowly. This is because necessary information cannot easily be anticipated a priori. As Sir Arthur Conan Doyle commented, “Life is stranger than anything the mind could invent.” In a similar vein, advertising creative guru Patrick Thiede once commented, “If the photo exists, it cannot be a great idea.”2 Similarly, architect Dan Jansenson said that there must always be a site visit, and you must be open to what you see at the building site. No building can ever be designed from topological maps.
Thus the recommendation in gathering data to support contextual price is one of balance. If you accept any old source of data, you are likely to find you have no usable data. If you specify too narrowly, you may find you have missed the right way of looking at the market. In line with Jansenson’s architectural requirement of a site visit, learn something about the likely pricing dynamics of the market before building a data warehouse. Being open to the dynamics of their market led accounts payable processor Tymetrics to discover its pricing did not just depend on payable through-put rates but also on the size of its network of electronic payees and how rapidly that network evolved. Again, this moves the information required for pricing from the internal and familiar, to the external and less familiar.
Build In a Systematic Look at Competition
Directly addressing the need for the right market data is not standard practice in building pricing data warehouses, but it can be. GE Patient Care recently integrated its competitive information sourcing with its pricing function sourcing. This was in recognition that competitive information was a key foundation for pricing. Internal information, however plentiful, would not suffice. Many companies would agree that competitors have an impact on their pricing, but few companies undertake the work of integrating this known factor into their pricing.3
Yet the rewards of finding what the right data are and then making sure that this data is part of the data warehouse are great. Many companies have employed talented statistical analysts to use the vast trove of internal company data to discover what drives customer defections and what drives price levels. The results have been modest. Typically such efforts yield relationships that explain about 25 to 35 percent of the results. In statistical terms, the R2 is typically from 0.25 to 0.35. In other words, not very complete.4
When the inquiry begins with a query of management and customers, asking what matters to price and loyalty, the resulting list of candidate drivers is generally quite complete. When such a list is tested against market demand proxies and specially gathered data, the results generally explain 80 percent or more of the results. This level of explanation (R2 = 0.82, in one case) leaves little to guesswork and simplifies price management. Further, when deployed in the field, the sales force rapidly gains confidence in the results.
In financial terms we have found this added degree of precision (explaining 80 percent of the price results, rather than 35 percent of the price outcomes) is worth an increase of 7 to 12 percent in revenues. For instance, typically there is a lot of unnecessary discounting, and understanding discount drivers might well cut the level of discounting in half—a major boost. This kind of gain will often justify the nontrivial investment in research programs and systems efforts to gather and deploy the needed pricing data. It also compounds the return from CRM and pricing engines.
Here’s a test, in case you can’t help but believe that the information needed for strong contextual pricing must be lurking somewhere in a company database. Suppose all your company’s managers and sales reps were replaced with new managers and sales reps tomorrow. Would they get what they needed from the existing databases? Probably not. They would have a starting point, but they would sorely miss the insights of today’s marketing and sales team. Without those insights, you suspect the company’s performance would sag. This tells you that adequate contextual information is not in your company’s databases.
Summary
Cold internal statistics will rarely suggest the contexts. Context requires more outwardly focused data: an understanding of how the customer will judge and compare your price offer. Thus, extensive interviews are required—both of one’s own customers and those of competitors. We find that focus groups often help in this process. No more information should be required to initially develop contextual pricing scenarios than to create a good segmentation, but also no less.
Routinizing the process means asking sales to capture new data in their sales-information system or asking market research to ask the right questions. To fully capture the benefit of an evolutionary segmentation approach, sales management should dynamically adjust its account strategies to capture the value of favorable sales contexts and opportunities. One digital device manufacturer adopted an approach of scouting for signals of an impending contextual change at potential customers, such as changes in operating and pricing authority. If the manufacturer found such signs of impending migration, it dramatically increased account resources—aimed at likely future decision makers.
Successful institutionalization of contextual pricing will generate a unique store of knowledge of customer trends that won’t be easily duplicated by competitors. In contrast to SIC codes, for example, contextual understanding can be redefined and refined over time. In a few years, the company will have a unique, proprietary insight into the history, transformation, and future of the market while competitors will have to constantly start anew with yet another snapshot to be discarded just as the market changes again.
Notes
1. John Matson, “Strange but True: Infinities Come in Different Sizes,” Scientific American, July 19, 2007, p. 21 ff. The article notes that even finite sets of natural numbers, such as those between zero and one, have infinite decimal natural numbers packed in between.
2. From a speech when accepting the “Euro-best” award for creative advertising, in 1990.
3. Galileo Galilei, in his struggle to find acceptance for his astronomical findings, found a flat refusal by his contemporaries to even consider the facts. In a letter he wrote, “In spite of my oft-repeated efforts and invitations, they have refused, with the obstinacy of a glutted addict, to look at the planets or the moon through my glass [telescope].”
4. To be clear: the factors that constitute the 25 to 35 percent explanation are not in doubt—often confidence levels top more than 80 percent. However, this is like getting a cooking recipe and finding that it lists only 35 percent of the ingredients. While there may be no doubt as to the ingredients actually listed, this leaves a lot to guesswork.