How marketers can avoid those big data blind spots

By Dr. Jesko Perrey, Dr. Dennis Spillecke and Dr. Andris Umblijs

Just because the data and analytical techniques are available doesn’t mean they provide complete insights.

If you were looking for a theme song that captures marketing today, you could do worse than pick Queen’s anthem “Under Pressure.” Marketing is under pressure to show results, cut costs, and drive growth. Marketers should welcome it. That’s because marketing has a big opportunity to drive above-market growth and demonstrate its value to the C-suite and the boardroom. In our experience, marketing can increase marketing ROI (MROI) by 15 – 20 percent. That kind of value can turn plenty of heads in the C-suite.

How? The explosion of data about consumers and the analytics techniques now available have made marketing a much more precise science.

All this wonderful Big Data created by the digital revolution, however, has created a serious problem for marketers. Just because the data and analytical techniques are available doesn’t mean they provide complete insights. This is very much what Albert Einstein meant when he said, “Not everything that can be counted counts.”

Blind spots

Consumer decisions are driven also by many stimuli outside the digital realm (e.g. TV ads). This results in a number of issues, not the least of which is misattribution of cause and effect based on a tendency to measure what is easy to measure, ie. giving credit where credit isn’t necessarily due.

One energy company, for example, observed that their customer losses were closely correlated to the intensity of customers’ Google searches for an energy supplier. They built a customer churn econometric model in which search was responsible for 65% of churn. However, in-depth analytics revealed that customers’ decisions to switch energy providers were driven by their and competitors’ prices, advertising and company’s position in social media, TV, print and other mass media. When all these additional explanatory factors were included in the customer churn model, search was not the cause of customer churn since people had already made up their minds by the time they were searching.

Most recently we’ve seen a lot of companies with claims about the ability of their analytical tools to provide a silver bullet set of answers to any marketing question. In our experience, those claims are hard to back up in the real world. What we see that’s most effective is having the right combination of tools and capabilities with as clear a sense of what they cannot do as what they can.

Overcoming “short-term-ism”

One major blind spot for marketers to be aware of is “short-term-ism” that most analytics engender. The reality is that the majority of marketing activities have both a short- and long-term impact on sales. The short-term impact is typically responsible for 10-30 percent of total sales while the long-term impact (called also the “base” or brand-building impact) is 1-3 times greater than the short-term effect. Big data-based analytical approaches, however, such as econometric and digital attribution modeling, for example, can detect only the short-term impact of marketing. What this means in practice is that the majority of data and analytics provide marketers with a short-term picture and can lead to short-term decisions that are detrimental to the long-term sales performance.

Given this reality, marketers need to overlay their Big Data models with analysis of the longer-term brand equity effect responsible for the remaining “base” of 70-90 percent of sales. Without ongoing investment in the brand, the value of this base erodes over time and creates a stiff head wind for future sales.

To understand long-term effects, companies first need to create a baseline by estimating the potential decline in base sales if all marketing activities would be stopped. Reviewing marketing investments and brand performance of multinational companies across regions is a good place to start since that data often yields a useful set of measurements of the impact of “brand leakage” (ie. how much base sales is lost and at what rate). That estimate then needs to be tested and adjusted systematically based on the unique situation of the company using the experience and judgment of marketing and sales managers, as well as other internal data (e.g. customer surveys).

These estimates can then help determine the Net Present Value (NPV) of the long term effect of marketing in terms of future sales. This NPV provides marketers with a reasonable understanding of the long-term implications of marketing in addition to the short term impact measured by Big Data and help make necessary trade-offs when it comes to making spend decisions. Although this is not a perfect science, we think it’s better to be “roughly right” than “precisely wrong.”

One consumer food brand almost fell into this short-term trap. It launched a campaign using Facebook advertising, contests, sponsored blogs, photo-sharing incentives, and shared shopping list apps. The approach paid off, delivering sales results similar to those generated by more traditional marketing (which included heavy TV advertising and significant print), at a fraction of the cost.

Given the overwhelming success of this effort, the brand considered shifting significant spend from TV and print advertising to digital and social media channels. When the long-term effects were included in the calculations, however, the proportion of impact of digital dropped by half because most of digital activities (search, display, etc.) typically are short-term calls to action and contribute little to building the brand and consumer loyalty. Significant cuts to TV spend as suggested by traditional econometric modeling would have reduced the net present value of the brand’s profit.

Marketing analytics is far from a monolithic approach. It’s actually a collection of approaches and techniques that, when systematically applied across a specific set of issues, delivers useful insights for making marketing investments that pay off. The latest wave of data combined with the right models can illuminate a lot. But smart marketers will spend just as much time looking for their data’s blind spots.

This article originally appeared on the Harvard Business Review (HBR) Blog Network site