Once you’ve gotten your data clean and trustworthy, it’s time to see what you can learn from it. As we’ll talk about later, part of building a data-driven culture is removing any sense of judgment or doom from data, and just using it as the neutral information it is.
One paradigm that can foster an objective approach to data is to look at is as a scientist would—or a detective. For good scientists and detectives, every data point is a clue to be followed and investigated, not feared or dreaded. This framework shows how to take that approach in three steps.
Step 1: Look and Listen
Break your desired outcomes into a series of single-focus questions that can be answered through data. Using your own insight and that of your team, form a hypothesis to test the aspect specified in the question. “Begin with the end in mind” applies here, as long as the end you’re seeking is an accurate answer to your question rather than a particular answer.
Step 2: Analyze and Learn
Develop a plan to test each hypothesis individually, as independent from other factors or changes as possible in the real world. Before testing, make sure you’re tracking the relevant metrics consistently and accurately. Once you’ve gathered the resulting data, analyze the results.
Step 3: Act and Iterate
Use the data you gathered during Step 2 to inform your next question(s) for your data. View your team’s progress through the lens of an Agile framework, where the ultimate goal is improving the process itself and it’s understood that the testing environment (the market) is constantly changing.
Following this framework is simple, but not necessarily easy. The most crucial thing to keep in mind when looking at data is this: approach it like a scientist. Bring your questions and hypotheses to your data; don’t look at the data first to see what questions it might answer. (Your team can do that later, and it might yield some interesting results, but for strategic matters, put your questions first.)
If the idea of being a scientist isn’t appealing, think of acting like a detective. Good detectives begin with a question (“Who has committed this crime?”) and use the data sets they have to answer it. By comparison, most depictions of bad detectives show them starting with the most easily accessible evidence and shaping it into a narrative of their own design. Bad detectives are big fans of vanity metrics.
Data in the Sales Funnel
Like interrogations in police shows, interrogating your data is all about directing the right question to the right resource at the right time. For instance, using the traditional sales funnel as a framework for informing general marketing decisions, you can test your hypotheses by directing these questions to these specified data sets:
AWARENESS – Where do our customers come from?
- Geographical data from analytics & paid search tools
- Online source data from behavior & path tracking
CONSIDERATION – How do our customers look for our solutions? Where do they go for information?
- Sales by source for buyer-oriented traffic
- Search query data for general traffic
- Traffic by landing page
- On-site search data
DECISION – Where do they need more information to help them purchase?
- On-site search data
- Cookie data for more global concerns & interests
LOYALTY & ADVOCACY – What is similar about our best customers?
- Demographic data
- Geographical & time-based data
- Site usage patterns
- Content consumption
Trust the Process
The scientific method has served those seeking verifiable information for nearly 500 years. Researchers the world over trust that process, because it begins with a question and seeks an answer, not the other way around. When you apply the scientific method to your own data, you can similarly trust its process, knowing that you’re proceeding in the right direction.
Many scientific experiments reach the conclusion of “Hypothesis not confirmed; more research needed,” and many detective cases aren’t fully resolved and get classified as “cold,” which essentially means “more research needed.” This is bound to happen to some of your data interrogations, and that’s to be expected. Use those experiences to inform the next question(s) you ask, which metrics you’re tracking, and even how those metrics are tracked.