These Major Accuracy Problems Are Affecting Your Marketing Reporting
Marketers thrive on data; without it, we’re glorified guessers. It’s easy to say that you think a certain strategy or campaign would be effective, but if the numbers aren’t there to back you up, your belief is meaningless. That’s why we’ve seen an explosion in the popularity of different marketing analytics platforms and dashboards designed to track and analyze different marketing metrics. We now have the technology to take more sophisticated measurements in more diverse areas, and we can mine that data less expensively as well.
2017 Digital Marketing Predictions From the Experts
There’s a problem, however. Even state-of-the-art data measurement platforms can give you wrong information, or present it in a way that makes you interpret it incorrectly. We rely on data to be an objective backbone for our conclusions and future plans, but if its accuracy is in any way compromised, how can we trust our conclusions?
The Accuracy Problem
These are some of the most common and impactful accuracy problems in marketing data analytics:
- Traffic filtering. Most online marketing platforms help you track the traffic that goes to your site, but how can you tell—for sure—where this traffic is coming from? By default, you’re probably tracking all your own internal traffic, meaning that when your web developers visit your site 100 times a day, those visits are counted alongside your organic search visitors. You may also be tracking rogue hits from bots and web crawlers, which shouldn’t be included in an analysis of human visitors. To compensate for this, most platforms (like Google Analytics) allow you to filter your traffic sources, so you can rule out visitors from certain IP addresses, or ones that come from non-human sources.
More robust platforms like Simply Measured’s Social Attribution product can tell you exactly which product pages and content pieces are getting shared across social and the web, where your visits are coming from, and can even enlighten you on how your content is being shared privately via Slack, text messages, and email (i.e., Dark Social).
- Confirmation bias. Confirmation bias isn’t a problem inherent in your data so much as a problem with how you interpret your data. These days, marketers have access to a plethora of different metrics, but that excessive quantity won’t necessarily lead you to better conclusions. Confirmation bias is a cognitive bias that guides people to overly favor metrics and data points that agree with their current assumptions; in short, confirmation bias can make you favor only the metrics that fall in line with your expectations, causing you to ignore data that might contradict it. Be aware of this and seek to disprove yourself.
- Compensation for hidden variables. Most tracking platforms keep things as simple as possible, which means tracking metrics literally and without exceptions; they don’t take hidden variables into account that could affect how you interpret the data. For example, direct visitors could have clicked a bookmark to your site by accident. Your page duration metrics could be bogged down by visitors who bounced. There are hundreds, if not thousands, of variables that can influence user behavior, so it’s important not to take any metric too literally. It’s better to look at each metric for exactly what it is—usually a response to a specific technical condition, rather than direct insights into subjective user behavior.
- Technological limitations. In some cases, there are limitations to how accurately our technology can currently track user behavior. For example, Google currently tracks visitors and “unique” visitors separately; this is a good way to gauge how many people are returning to your site, versus how many people are finding it for the first time. They track this by installing cookies when a user visits a site, which is somewhat effective; however, it doesn’t work especially well when a user disables cookies, deletes cookies, or visits a site from multiple devices.
Google and other companies are working on fixes to improve the accuracy in areas like these, but there’s an upper limit to how precisely we can monitor and track web traffic and user behavior.
- Status quo bias. Another bias that interferes with your perception of data is the status quo bias, which causes people to be resistant to change. With this bias, you might see a noticeable drop in traffic and rationalize it as a coincidence—after all, your strategy has worked well so far, right? Status quo bias can also make you resistant to trying new strategies or tactics, especially if your current set is working reasonably well, but that’s more a problem with interpretation and action than it is with the data itself.
How Precise Does Our Data Need to Be?
For the most part, these biases and accuracy problems have only a marginal effect. You might see 2,000 visitors instead of 1,800, but if you’re consistently seeing an increase in visitors month over month, that 10 percent difference isn’t going to lead you to a false conclusion.
That being said, it’s important to avoid seeing your data too literally, or assuming that it’s 100 percent accurate in all cases. Most of the time, it’s better to take a high-level snapshot of your performance, drawing broad conclusions, than to nitpick over each individual bit of data you’ve tracked. In time, our tools will grow even more advanced and more precise, but until then, remain as distant and unbiased as possible, and do what you can to improve the accuracy of your data.