Marketers today collect more data across more touchpoints and more channels than ever before. That means marketers are finally gaining actionable insights and making strategic, data-driven decisions, right?
Even with all of the data in the world at your disposal, if your analysis misses the big picture, your data insights will lead you down the wrong path. Worse, you’ll be incorrectly convinced that you’re on the right track.
In reality, marketing data is incomplete and unreliable most of the time. Have you ever compared advertiser platform click-out figures with arrival data on your website? Have you ever looked into who the individuals that “like” your social media advertising are? Do you know what hides behind “direct traffic” in your analytics reports?
Find the answers to those three questions and you will start to become a data skeptic like me.
Why I Became a Data Skeptic And You Should Too!
I first learned to not believe in everything the data tells you many years ago when I was a top affiliate in charge of driving high quality traffic from search engines directly into eCommerce sites.
The source of my success was a mastery of user journey data through supplementary and inventive tracking. Going further than most affiliates, I set up custom tracking that would show referrer data like the exact search engine query and environment variables of the user.
As a result, I could tell that when one particular user made a €18,000 purchase they had only entered the site less than two minutes ago. That purchase represented my largest affiliate earning ever.
At the time, I was all too happy to receive the last-click attribution credit for the sale. But I knew then as I do today that it was unlikely that I had acquired a new user and convinced them to make such a large purchase in two minutes. In all likelihood, this user had been convinced by a physical catalogue and arrived on the eCommerce site ready to buy.
The moral of the story: No matter how exhaustive your data tracking, if you don’t check your blind spots, the insights you extract will not be meaningful!
Two Opposing Trends: More Data & More Data Degradation
Today, there are two opposing trends at work in the digital marketing.
On the one hand, we are tracking more touch points across more channels than ever before, resulting in greater amounts of data to sift through. As David Amerland (author of The Sniper Mind) said in Innovell’s latest search strategies report, “The problem with data is that there is too much of it.”
On the other hand, there has been progressive degradation in data quality for a number of reasons. First, there are new unreliable and calculated data-incursions. For instance, because of retargeters, last-click attribution is now much less reliable. Second, data degrades because data carries value and advertising platforms or partners wish to keep a certain level of opacity. Case in point, Google shutting down our referrers. Finally, data can also degrade due to incompleteness and break-offs in the user journey: data corruption, device-changes, data-deletion, data-blocking or other limitations. Whereas tracking technology improves, all of these types of occurrences have an ever-increasing impact. The end-result is that we have incomplete and/or unreliable data.
Suboptimization: How Unreliable Data Emerges
There are massive shortcomings to optimizing in closed data systems (also known as “silos” or “boxes”). Within a box of finite size, once you reach a certain number of optimization iterations, you run into an effect called suboptimization.
Suboptimization happens when you are optimizing the marketing system within a box but the results outside the box are actually negative. To illustrate what that looks like, here are three examples.
Example 1: Suboptimization in Retargeting
In retargeting, you work on a piece of actionable data. This can be the knowledge that a product was viewed or was put in the cart by a user. Within the closed data system “box” where this campaign runs, there is a device receiving a cookie allowing the marketer to take action on that data. The retargeting system will optimize to this (actionable) event assuming that the user is interested in this product and has been disrupted in the journey towards fulfilling the purchase.
Events occurring outside of the box in which the retargeting campaign is run cannot inform the campaign. If the user has actually bought the product via another device , with another browser , or purchased offline this information which disqualifies the original information, never enters the box.
When there is no signal within the box indicating that the user has already purchased, the system will increase advertising pressure assuming it was previously not strong enough to make the user convert. This undesirable (and brand-killing) effect can also be achieved by not excluding recent converters from a retargeting campaign. In this case it could be incompetence on the part of the the digital marketer, but in most cases, it will be clear case of suboptimization.
Example 2: Suboptimization in Mobile Campaigns
Suboptimization can occur for mobile campaigns where users don’t acquire on a smartphone.
Let’s take the example of my company, Innovell, that commercializes a 100 page report on paid search trends in a printable PDF. The document is illustrated with the aid of engaging graphics making the file a bit heavy. Furthermore, many make the purchase using a company credit card which requires the entry of a VAT number in order to get 20% off.
In this scenario, the user is likely to discover the report via a smartphone because that is where most people find Innovell’s promotions on Twitter, Facebook and LinkedIn. But they are also likely to make the purchase on desktop.
Running campaigns to promote the report on mobile would give the data system the feedback that mobile never converts so any automated optimization to conversion data would reduce all mobile spend. In reality, mobile is probably essential. So we need to manually up-prioritize mobile to compensate for data not entering the box.
Example 3: Suboptimization when Optimizing for Events Rather than for Value Insights
Our digital optimization loops are a huge improvement compared to acting without data. We have moved from mental constructs to data-driven optimization. In any instance, doing that move is an improvement to your marketing system, but when you push and automate your optimization, you inevitably hit suboptimization at some point. That is when you need to expand your data dimensions to include the value of each transaction rather than simply the transaction event. Add more insight into your actionable data.
Search Strategies Outside the Suboptimization Box
The old granularity and optimization approach stays within the suboptimization box. It consists of adding more keywords, refreshing ads, more frequent bid adjustments and adding negative keywords. Today, this approach can no longer be called a strategy in its own right. Rather, it’s only the starting point for a truly successful search marketing campaign.
The new generation of strategies seek to break out of the suboptimization box. A comprehensive survey of campaigns that won search awards in the US, UK, and EU reveals that some strategies merge the search and shopping boxes into one and by putting structured data first, enables a quantum leap in performance for those adopting it. Other strategies integrate offline data into the marketing system and enables click-and-mortar businesses to take their overall performance to another level.
The source of the new crop of emergent strategies in paid search? It’s the data skeptics who understand the challenge of suboptimization. Ironically, in a data-obsessed world, it’s the data skeptics who are coming out on top.
Feature Image: Unsplash / Tobias Fischer