Reliable Data Is Within Reach: 6 Steps to Data Confidence

April 6, 2017

Data may not be the most exciting part of marketing, but it is the foundation for gaining the marketing insights you need to continuously optimize your site.

An analytics strategy is only as good as its data: you need to be sure the data being collected is "clean" before gathering insights from it. If the data being analyzed is defective to begin with, no matter how good the analysts are, the insights derived and given to decision makers will likely be misleading or outright wrong.

6 Steps to Data Dependability

To prevent these kinds of situations, we outlined the following specific steps to ensure you are working with dependable and reliable analytics.

1. Implement a Tag Manager

If you are not using a tag manager, now is the time to start. It reduces your dependence on IT, allows you to quickly measure the impact of new features and fix bugs you uncover in data collection methods.

In other words, you’ll no longer have to live with corrupt data as you wait for IT to prioritize and deploy your code. You will be in charge of your own data gathering.

Tag managers can also be used to build reliable connections to all your analytics tools. Adobe’s Dynamic Tag Manager and Google Tag Manager are two popular, easy to use tag managers. And best of all? They are absolutely free.

2. Test your analytics implementation across all browsers and devices

Testing should be a critical part of your implementation strategy. Since data collection is JavaScript-based, its behavior can vary across different browsers and devices.

Take a tip from your IT teams and test analytics implementations across platforms to ensure data is being collected reliably. Write test cases for analytics, just like a QA tester does for code.

3. Implement data governance processes

Unless you are the analyst who just finished testing all the code, how do you know what the data means? Far too many analytics implementations do not come with documentation, which leaves marketers and other analysts with very little to rely on when it comes to interpreting reports. Clear documentation can make all the difference between deriving reliable insights and making embarrassing claims.

First and foremost, keep solution design documentation current and make it available to all analytics users. Then share starter dashboards and reports that demonstrate best practices for custom reports, offer analytics training, and establish communication channels to ensure reporting consistency across all departments.

4. Data pre-processing

Reporting tools give you the ability to pre-process and filter click stream data before it makes it into the reporting suite. While mechanisms like these are powerful, they need to be used carefully.

Fiddling with incoming data before it is analyzed can lead to confusion and is often difficult to debug. To minimize complications, configure your data process to touch the data as little as possible during collection.

5. Reporting

Is testing the data the first thing you do when reports show anomalies? Are you sure that the drop in conversions was caused by actual events or is it a tracking failure? Do you have sufficient diagnostics to prove the data is reliable? Since analytics platforms are so customizable, build diagnostics into your analytics process from the beginning.

6. Segmentation

Today, analytics reporting software can produce much more than simple reports. Now we can create advanced segmentation that significantly expands the capabilities of reporting. We can also nest statements and conditions to derive the specific results.

But how do we know when a segmented report is reliable? Start with simple segments that can be tested on related reports. For example, a segment for US traffic on a country report should only return US traffic. Knowing that the filter conditions are returning the expected results are a critical first step to advanced segmentation.