Viral video devotees might be familiar with “Teens React To,” a YouTube series in which today’s smartphone-toting kids meet yesterday’s obsolete tech. 

A recent episode, for example, involves young people bewildered by Windows 95 PCs. Accustomed to touchscreens and ubiquitous WiFi, most of the kids can’t even turn on the computer, let alone deal with a dial-up modem.  

What Analytics Can Learn from Search 

If you’re not actually a teen, the videos will make you feel a little more ancient than you actually are. But they’ll also make you think of other rapid tech evolutions — and about the difference between young tech that tantalizes us with its potential and mature tech that actually works.

I kept imagining an episode in which teens accustomed to today’s highly-personalized search engines have to confront the generic, impersonal versions we used a few decades ago. Not an obviously hilarious topic, I know — but stay with me. 

The more I thought about the evolution of search engines, the more I realized they’re a great metaphor for how today’s marketing analytics technology can evolve to realize its potential. 

Personalization Is the Future

Back when today’s teens were still infants, Google’s algorithms valued interconnectedness — meaning, broadly speaking, the more links that pointed to your page, the more juice that page had in search results. Google basically treated these links as proxies for a page’s authority or legitimacy, almost as though clusters of hyperlinks represented some crowdsourced consensus about the best search results. 

Compare that to the modern Google. Today, teens would find it absurd that anyone would care what the average or typical search return is, as though all the people who enter a given query actually want the same answer. Instead, Google has conditioned us to expect personalized results. 

We often don’t care which pages have the most links pointing to them; we care about the pages that — based on our past search and internet browsing histories — actually address our concerns.

Indeed, we see this shift everywhere as search engines cease to be landing pages that one visits, such as Google.com, and increasingly become features within other services. After all, what is the Facebook news feed if not a kind of predictive and curated search engine? 

Personalize Your Analytics Methodology

The point is this: Given the same question, different methodologies often produce different answers. Even though search’s fundamental purpose — to help you find the most relevant information — hasn’t really changed, our search methodologies have changed dramatically and so, too, have many of the results we receive. 

Let’s bring this back to today’s data-driven marketing. When we think of old Google versus modern Google, or modern Google versus modern Bing versus modern Facebook, we understand implicitly that the substance and quality of search returns varies across platforms. 

As a “Teens React To” video would demonstrate, different methodologies yield different results, some of which are drastically different from what we expect or want. 

Analytics: Think Search Engines for Datasets 

When marketers attempt to select analytics solutions, they face basically the same situation: Marketers are awash with consumer data — social media, forum posts, online shopping carts, you name it. 

Analytics solutions act like a search engine for these datasets. Just as Google tries to find the elusive signals that match your query inputs, analytics engines try to distinguish buzz and noise from the data points that actually affect your business. 

Here’s the trouble: Marketers know they need to use data to understand their customers better but, so far, many solutions haven’t delivered the expected results. The reasons for these failings are numerous but here’s a big one: Marketers aren’t paying enough attention to methodology.

Know Your Intent 

A variety of analytics products claim to measure consumer intent, but vendors’ methods for defining and deriving “intent” vary wildly. Sometimes the solution uses click-through rates as a proxy for intent. Other times, the solution might be based on natural language processing. At my employer, Quantifind, for instance, we look for indicators of intent in the consumer language patterns that correlate with a client’s financial movements. 

Marketers don’t need to be data scientists who can discuss R-squared values as easily as email and social media campaigns. But we must understand that if seven different methodologies claim to measure “intent” or “lift” or “awareness,” it’s very likely that some of them are defining terms and calculating results differently from others. 

Understanding Your Preferences 

Even when Google and Bing purport to do the same thing, we know from experience that we prefer one over the other and we understand that methodological differences are the reason for these preferences. We need to apply that same understanding when we assess analytics products. 

Every methodology rests on assumptions about what certain measurements imply, whether x can be a proxy for y, what kinds of relationships are valid, and so on. Data science techniques can validate whether an assumption is well-supported by data but assumptions still underpin the entire enterprise. 

Use the Past to Predict the Future

Most techniques are useful for measuring something, but to know whether they’re measuring what you need to know, you need to identify your solution’s assumptions. Does it make sense for a product to assume social sentiment is a proxy for consumer intent? Does the solution assume the loudest online voices are also the most important? 

Asking these questions of your analytics solution will go a long way toward shifting your mindset from the past to the future. Meanwhile, enjoy that brief, shining moment when you know something your teenager doesn’t. 

Title image "" (CC BY 2.0) by  steve p2008