3-Step Approach to Using Big Data for Personalization

Evergage Blog

Ideas and Strategies for Real-Time Personalization
3-Step Approach to Using Big Data for Personalization

August 2, 2016 by

Last week I spoke on a panel at the MITX eCommerce Summit entitled “Turning Mountains of Data into Streams of Revenue.” Along with other experts in the e-commerce space, I addressed the big data and attribution challenges involved in improving conversion rates.

One of the key challenges we discussed was how to uncover relationships within data and how to act on those relationships. In this post I’ll share highlights of that discussion.

1. Find a problem you want to solve and begin exploring the data

Rather than approach a mountain of data blindly, you should begin with a specific question. This means that you must be able to access to your data in a way that you can ask questions of it, such as within your web analytics tool, data warehouse, or a solution like Evergage.

For an example, let’s say you’re interested in the key differences between shoppers who convert and those who don’t. You would begin by digging into relevant numbers to uncover key differences between those groups, such as time on site, use of search, traffic source, initial page visited, etc.

As you learn more, ask follow-up questions to dive deeper. For example:

  • Why does this traffic source seem to convert better than that one?
  • For visitors who arrive on site directly on a product detail page, where do they go next? How does that impact conversion?
  • How does category affinity or brand affinity affect these results?

After you have explored the data on your shoppers and customers, pick one finding to explore further. For instance, it may seem that return visitors who have not yet purchased tend to land on the homepage and then engage in search repeatedly.

2. Formulate and test a hypothesis

Next, formulate a hypothesis. For example, perhaps these visitors are trying to find the products they looked at during their previous visits.

Determine an action that would be a way for your site to help these visitors find the products they looked at last time. For this situation, you could test placing a row of recently viewed products on your homepage in an easy-to-find place.

3. Measure the results

Finally, you need to measure the results of the campaign on relevant metrics. Of course, you need to ask if the visitors engaged with and/or clicked through to the recently viewed product display. But you also must ask, did conversions improve? What about average order value or revenue per visitor? You can also look at other behavioral metrics, such as time spent on page and number of searches per session.

The best approach is to test the campaign against a control so that you can be sure that the campaign was an improvement over the previous experience. Without A/B testing, you can’t be completely sure if your hypothesis about the behavior you observed in the data was correct.

Final Thoughts

It is possible to turn a mountain of data into a stream of revenue — one campaign at a time.

Rather than make tweaks to your site based on what you think will have an impact, explore all the data you have available to form realistic hypotheses about why your customers engage with your site in the ways that they do. Then be creative about how to make it easier for them to accomplish their goals. Keep iterating and trying new things based on what you see in the data, always putting the customer at the center of your strategy.

Share This