In short, the answer is no unless your current analytics tool does a poor job at helping you clean your data. There are a few different schools of thought when it comes to the best way of fixing data quality issues, but everyone agrees on one thing. Bad data needs to be addressed quickly because it messes up everything: analytics, machine learning/AI components, and, ultimately, your company’s ability to make good, swift decisions. So how does data go bad and what should you do about it?
We’ll discuss all the root causes of bad data in detail in future articles, but the simplified formula for disaster is:
How do we address bad data then?
Given the usual culprits above, the logical first step is to set up robust guidelines for data quality. This is a must no matter what and should include, among other things, data collection guidelines, periodic data auditing, and who owns data quality. Today, bad data cannot be fully solved by machines. So it’s mostly a people problem that can be mitigated with tools. Until machines can fix bad data proactively, people need to safeguard the quality of data. It’s annoying at first (ugh, one extra thing to worry about, right?) but not addressing it upfront will allow data issues to fester and cost you more and more resources down the road.
Bad data is primarily a people problem, not a tool problem.
Next, check if you can leverage your existing analytics tool(s) to fix your data. In some cases, this means fixing the data warehouse that’s pumping data to your tool. Usually, bad data comes in a few different forms (e.g. badly labeled data, duplicate data) and good product-led analytics platforms empower you to fix it. If that’s not the case, getting a new tool that aligns better with your team resources, supports your new data standards, and makes your team’s job easier might be the next step. But not before doing a bit of ROI math: compare the cost of carrying on as is (subscription for current tool + resources for fixing and maintaining data) vs. the cost of the new analytics tool + cost of migrating to the new tool + data maintenance savings.
How do you know if you’re using good enough analytics tool(s)?
Let’s recall the different types of bad data and see the minimum requirements for a good analytics platform:
- Badly labeled data. This usually manifests in the form of misnamed events and attributes/properties. A good analytics platform will let you enforce data quality controls. (Avo, Amplitude with Ampli). It will flag data issues proactively, and, if badly named data does get through, it will let you adjust data labels accordingly without necessarily changing the data source.
- Inaccurate data. This translates to things like missing data, duplicate data, or just plain wrong data that doesn’t reflect reality. A good platform enables you to troubleshoot the root cause and then fix your data by backfilling missing events, removing or hiding duplicate data, or correcting your existing data.
I say these are minimum requirements because all the potential data issues I mentioned are fairly common. Chances are you encountered at least one or two of these so far. So if you’re using a combination of tools for your analytics needs or an all-in-one solution, they need to be able to solve your data challenges effectively.
To recap, our main takeaway is this: don’t spend time, energy, and money on a new tool if what you have 1) can be fixed and 2) aligns with your business goals.
