How Overlooking QA Can Sabotage Data Analytics

How Overlooking QA Can Sabotage Data Analytics


Too often quality QA goes MIA.

Data analysis is a fundamental part of the world we live in today, responsible for driving any number of important business decisions, so ensuring data quality is of paramount importance. Yet, too often, a crucial step gets overlooked in the process.

Knowing what to measure, on websites and applications, and how to measure it, is an extremely important factor when gathering information about users, predicting behaviors and measuring performance, among other relevant variables for a business. Yet too often, businesses don’t invest enough in ensuring that data is of the utmost quality.

In most analytics projects we overlook the importance of adequate testing, leading to the data collected not meeting expectations, which can cause a number of different problems: extra hours to find and fix problems, constant bugs, untrackable errors, data mistrust, data leakage and non-retroactive fixes, to name a few.

One of the most important things that differentiates great analysts from adequate ones is attention to carrying out proper QA throughout the process of collecting, inputting, and analyzing data. Unfortunately, most analysts aren’t familiar enough with QA best practices.

Analytics QA consists of testing and assessing whether inputs and outputs work seamlessly and generate quality data during the collection, ingestion, and storing of analytics solutions, in alignment with business goals and requirements.

There are two main pillars that drive Analytics QA:
•Data QA: Collecting the right data, such as: having the complete e-commerce information of each transaction enabled for reporting.

•Implementation QA: Verifying that the implementation is working as expected, such as the tag management tool correctly loading, tracker scripts working, custom JavaScript without errors, and tags being triggered.

Our Analytics QA process is based on our CIS Analytics Framework, which assesses the entire data analytics pipeline and helps us focus and audit the different stages in which data collection travels. To accomplish this, a solid Analytics QA will consist of testing the following things: workspaces and preparation, the implementation itself, tags, triggers, and variables, the payload, data processing, and the stored data. Of course, this is not a static method, and the testing cycle should happen at different stages – such as testing and production, when testing new tags, and when apps are sent for approval to the App or Play Store.

Perfecting data analytics QA practices means never getting tired of testing. While the level of effort required to deploy QA is higher during a testing phase, data quality analysis efforts actually carry more weight during production, since that’s the final set of data they’ll be applying at the end of the process.

Needless to say, depending on what you’re testing, and your approach, methods and tools can vary. If data can’t be trusted or it isn’t reliable enough to inform decision making, it doesn’t matter if you use a free analytics tool or pay for the most expensive tools on the market: either will be equally useless. That’s why choosing the right analytics partner is crucial to ensuring a best-in-class implementation where the client can trust that their data is reliable enough to drive important business decisions.

Taken from: