Big data has been one of the key buzzwords in the analytics world over the past couple of years. Though the biggest difference between big data and regular data is the sheer volume of it (variety and velocity come into play as well), many marketers have been stumped by it. How does a marketer in the trenches make the most of the big data brouhaha?
Believe it or not, the answer may lie in something you may vaguely remember from your days in high school science class—the scientific method. As defined by Merriam-Webster, the scientific method is comprised of "principles and procedures for the systematic pursuit of knowledge involving the recognition and formulation of a problem, the collection of data through observation and experiment, and the formulation and testing of hypotheses."
Using the scientific method to tackle big data is really about testing your hypotheses. Let's walk through each of the steps with our marketing hats on and see how the scientific process might work.
1. Formulate your question. Are our customers more likely to complete a purchase on our Web site if they are provided with targeted content across multiple media channels—social ads, email promotions, and customized offers on our Web landing page?
2. State your hypothesis. My hypothesis is that customers have a preferred engagement channel, and that customers who see messages across multiple platforms will be more likely to click through (and purchase) than customers exposed to targeted content in only one channel.
3. Consider what the data will tell you. The data will help me determine which channels are the most effective. I think the data will also enable us to segment our customers by their preferred mode of engagement.
4. Run your test or experiment. If you already have historical campaign data, work with your analytics team and have them analyze the data to test your hypotheses, as well as see if there are any other nuggets of insight. If you haven't tackled your question more formally, take your next campaign and map it to your hypothesis above. Make sure your Web properties, emails, and social media ads are tagged properly to capture the results you will need—this is where you need to work with your developers and analytics team to make sure this is done correctly.
5. Analyze the results. Dive into the data—ideally with the help of your in-house analyst or data scientist. Tell him what your hypothesis was and what you are thinking will come out of the data, then ask him to play around with it. Is your hypothesis confirmed? Was another insight found?
6. Repeat to see if the analysis and results are replicable. In true scientific fashion, you would rerun your experiment to make sure the conclusions you drew were valid—and you would also let others take a look at your data and try to replicate the results. In the real world, this is a bit difficult, as "others" could very well be competitors, or your data could contain personally identifiable information and thus is not something that you could easily share with those outside of your company. But you can certainly work with your team internally to run your test scenario again. Perhaps you can introduce a slightly different campaign and see if the "campaign" or offer itself is a decision lever or not a relevant variable.
Like a true scientist, leverage all of your corporate resources to help you with the scientific method. The following are especially important:
Be a partner with your resident data scientist. Good data scientists or analytics experts are hard to find. There are a lot of people who can crunch the numbers but not many who know what to do with those numbers. Your business expertise is the winning ticket here.
Work together. Let your analytics experts know what your business plans and strategies are; if you are looking for information that will help you more effectively cross-sell a product, tell them, so when they start playing with the data, they will have some direction on their focus. Over time, as your partnership grows and the analytics guru's insights help you achieve greater business success, you will learn to do this automatically, and the data person will learn to ask for business goals before every analysis.
Let automation and technology do some of the work for you. Work with your analytics and IT teams to build in logical next steps for your customers. There are countless examples of companies doing this well. I'll give two examples.
When a customer searches for a hotel in New York City, the Expedia Web site doesn't just list the hundreds of hotel options haphazardly. Instead, Expedia uses an algorithm to prioritize the options presented based on what is most likely to appeal to that specific customer. For example, if the customer checks off that he's traveling with children, the options presented first would be family-friendly hotels.
Expedia did get some flak for this customization last year as it was discovered that Mac users were presented with "pricier" hotel options first. Customers still have the option to adjust the search filters according to their personal preferences if what Expedia presents isn't quite what they are looking for. The goal is to simplify the purchasing decision for the customer. Faced with hundreds of choices, it's hard for a consumer to decide. But faced with attractive choices at the top of the search results, the consumer can save time and quickly make a booking, a win-win for both the customer and Expedia.
Amazon.com uses not only your historical purchases but those of others to help narrow down choices—you see the familiar "customers who looked at this" also checked out products X, Y, and Z. Or if you are frequently buying something (say, diapers for new parents), Amazon suggests a "subscribe and save" option for you to simplify your life and prevent any potty accidents.
These are not perfect algorithms, as customers may have been buying on behalf of someone (a baby gift for a friend rather than for themselves), but over time the systems can learn to differentiate and discount these "anomalies" in your customers' behavior.
So instead of letting big data stump you, put on your scientist's cap and focus on the questions you are most interested in. The scientific method, a good partnership with your analytics team, and logical automation in your processes and IT systems can all help you deliver great experiences for your customers.
Olga Spaic is the manager of analytics at Metia. She creates test and measurement plans for marketing initiatives, develops dashboards with KPIs and supporting metrics, and provides comprehensive analyses and optimization recommendations. She honed her marketing and analytics skills at companies including Microsoft and AT&T.