Why Marketers Should Borrow From Google’s Playbook

Understanding the connectedness of data makes it inherently more valuable

The data sets marketers have in their possession are more than just a collection of attributes and behaviors.
Getty Images

Remember back in the early days of the web, when search engines relied on indexing algorithms to create a taxonomy of the web? They’d ingest data and then sort them into buckets as best they could. As a user, finding useful and relevant information took considerable amounts of skill and patience.

Then Google changed everything, and in so doing, set itself up to dominate the industry. Google realized that the real power of assessing the relevance of a webpage lay in its connectedness to other pages, more so than the content itself. This realization led to its PageRank algorithm, which is still used today.

Webpages with the highest levels of connectedness—i.e. with the most number of backlinks or other webpages that point to it—are deemed more relevant by PageRank, and move up in the search results.

Google proved definitively that understanding the connectedness of data makes it inherently more valuable. I think it’s high time marketers followed their lead.

From a machine-learning point of view, customer data is like website data

Today’s marketers have an unprecedented amount of customer data at their disposal. But it’s messy, difficult to use, and very confusing. In other words, it’s just like the early web that search engines struggled to make sense of.

Like early search engines did back in the day, marketers slice and dice massive amounts of data in relational databases based on rudimentary queries, hoping to unlock patterns and insights to leverage. They’ll ask things like: “Who purchased our product last month?” or “Which customers abandoned their cart this year?”

One can build a campaign based on those query results, but just don’t expect anything other than an anemic performance, as these queries have virtually no strategic value for planning future marketing activities. Let’s face it, this approach is little more than last-click targeting. But here’s the thing: We know the data, and its connectedness, is getting more complex and nuanced, it’s just that marketers have yet to take advantage of it.

Marketers tell me all the time they want to make better use of their data and are building data science teams internally in order to do just that. That’s great, but to succeed they really need to follow Google’s lead. Google realized early on that real insights come from applying machine learning to understand the underlying connectedness of data, and in my opinion, what’s true for Google is true for marketers. The data sets marketers have in their possession are more than just a collection of attributes and behaviors.

I guarantee you that buried within them are patterns that connect these customers, reveal their common affinities and identify actionable trends. What’s even more exciting is the way these connections can become veritable roadmaps to create campaigns and content geared towards specific segments, so not just who purchased my product last year, but what news sites she reads, what she’s passionate about.

Google leveraged machine learning and graph technology to power its search engine, an approach that I think will be beneficial to marketers in terms of segmenting audiences for prospecting purposes.

Machine learning helps eliminate bias

Graph technology uses unsupervised machine learning to identify connections, and to group data points into clusters. Humans don’t have to tell the machine what to look for, the machine’s algorithms crunch through massive datasets, applying weights to observed connections and identifies common patterns. By definition, machine learning helps eliminate the bias marketers bring to their audience segment definitions, and lets the data speak for itself.

This is the exact approach marketers should take when creating strategies. Too many build personas based on their perceptions of who their customers are rather than who they may actually be.

Recommended articles