How to get actionable, C-suite-relevant customer insights from big data. During the research for “The 12 Powers of a Marketing Leader,” we spoke with many leading CMOs about how they draw real learnings from big data. Because there’s so much hype and confusion about big-data insights, here are some practical suggestions from your peers.
“Big data” often means that there’s lots of it, but also that most of it is generated as a byproduct of other things—for example, the company’s routine operations or consumers’ social media conversations—rather than specifically for insight purposes. It’s typically spread around the company and elsewhere in different formats, and, quite likely, on incompatible legacy systems. It includes errors and missing values and isn’t easy to analyze or distil. Basically, it’s a mess. Relax. This is normal.
As a leader, you may get swamped with proposals for approaches and tools, most of which are almost impossible for non-specialists to evaluate. Again, chill. This, too, is normal.
Big-data customer insights are primarily a leadership challenge, not a technical challenge. Your job, as a senior leader, isn’t necessarily to become a data expert but to own the business thinking. See the big picture, ask the right questions, and only then bring in the experts. Here are the seven rules:
1. Step back. Ask yourself: What are the business issues we’re cracking to serve customers better? There are endless things you could analyze—buzz, churn, distribution, feedback, comments on the competition, revenue, price trends, customer preferences, profitability, share of wallet, and so on. Don’t try to include everything in your insight plan. Confirm which big issue you want to tackle for the business. Then, together with your team, set up hypotheses about which data might help you most.
2. Create an information map of your company. What customer-relevant information is flowing into and around the company today? Where? In what form? Few marketing departments have such a map, but it’s a powerful way to find really good customer insights.
3. Pull out some data manually, and play with it. However complex your systems are, you should be able to extract customer insights manually from a small sample of the relevant data. Do an initial, small-scale pilot analysis. By getting your hands dirty with this data, you’ll develop an intuitive feel for what’s there, how good it is, and the types of pattern and insight that might emerge from a full-scale analysis.
4. Get several views of potential full-scale data-insight projects. Invite, say, three firms to show you how they’d help generate insights that address your key business issues. Tell them what these issues are, and brief them on what data you have and what you’ve found in your pilot analyses. Ask them what approach they’d use to generate and validate these types of insight routinely and at scale in the future. Ask them about costs, timelines, and other companies where they’ve used their recommended approach. Ask if you can talk to these other clients. Hearing several different views will give you a basis for comparison and greatly expand your understanding of what is and isn’t possible and sensible.
5. Continue with manual insights analysis while you implement an IT solution. A temptation with big data—often encouraged by service suppliers—is to set up a big bells-and-whistles project that will supposedly get you amazing results … one day. This approach is typically expensive, inflexible, and slow, with a high failure rate. It may take a long time before you get any insights. Worse, you’ll likely end up with the wrong design if people add stuff on the way, as they think of more features that would be “nice to have.” Instead, demand that every big-data project should produce immediate insights at each step of the implementation—even if done manually along the way—so you can continuously learn and steer the project well.
6. Hire some good data analysts of your own. This will enable you to use data analytics in parallel with more traditional sources of customer insights. Everyone else is looking for good data analysts, so be prepared to pay a bit over the odds—this is a strategic investment. Every contemporary marketing team must have strong analytics capabilities.
7. Don’t run too many parallel big data projects. To get real big-data customer insights that every C-suite executive wants to see, avoid running more than one or two big-data projects at the same time ...
... and don’t compromise on the seven rules.
(I version of the article has appeared on cmo.com)