How to explain bad data

Photo by Robin Glauser on Unsplash

How do you instill confidence in your data presentation when the underlying data isn’t perfect? In today’s post, let’s talk about what bad data is, why we have to use it, and some do’s and don’ts for presenting “bad” data to leadership.

What is bad data?

Not to be confused with “bad results”, “bad data” is data ought to be called “imperfect data”.

True bad data – data of such poor quality that it can’t be used – shouldn’t be used.  

Rely on your professional judgment and organizational standards to understand what acceptable thresholds for data quality are.

In some settings, near 100% accuracy and completeness in the dataset is a requirement. In others, a small sample or “whatever is available” may suit your needs. Understand what boundaries apply to your data before getting too far into the analysis.

Why are you using bad imperfect data?

Before getting into how to explain your imperfect data to leadership, it’s good to understand why you’re even using it in the first place.

Reason to use bad imperfect data 1: Data is never 100% perfect

Even with high tech, highly accurate sensors and data sources, data is never 100% perfect. Sensors may be out in the elements, installed incorrectly, rely on spotty wifi, or need maintenance.

Reason to use bad imperfect data 2: You’ve done everything you can to “clean it up”

You’ve taken every step to recover as much data as you can out of your imperfect source. What is leftover still meets the quality requirements for this application.

Reason to use bad imperfect data 3: You don’t have access to “good data”, but imperfect data can still be used.

Imperfect data is still usable depending on its quality. A common misconception in business is that if there is imperfection in the data, the whole dataset should be thrown out. Of course, this isn’t a given. Data isn’t a bucket of apples – one bad row doesn’t spoil the lot.

You might have old equipment, an outdated collection process, or historical data that was captured using pen and paper. Showcasing what you can do with the data you have is a good way to build a case for obtaining better quality data in the future. 

How to confidently communicate insights derived from bad imperfect data to leadership

You can skip commenting about the data quality. After all, it’s up to you how you address data quality issues in a presentation. When there are new metrics, audience members, or you’re new to the program, you’re going to get more questions so it’s better to be prepared.

Honesty is always the best policy when communicating with leadership. The key is to find the right balance between sharing too much or too little about the imperfections underlying your data.

The goal is to be confident about the insights you’re presenting and transparent about the limitations of the data without sounding like you’re covering something up. Here are some do’s and don’ts to start you on the right path.

DON’T: Call it bad data in your presentation.

Don’t undercut your own credibility by using this phrase. Calling it “bad data” suggests that you shouldn’t be using or trusting it. “Bad” can mean a wide range of quality issues that all  have a negative connotation – and it allows the audience to begin making their own interpretations.

Remember, the data is imperfect. Data is never perfect. It’s ok to acknowledge and normalize that.

DO: Be prepared to describe the level of “badness” of the data

Now that you’ve been transparent about the data quality, expect more specific questions about the quality.

Pre-empt these questions by including the details in your presentation script. Or, have the information at hand in your notes or a hidden slide. Consider the following questions.

• Are you suffering from a one-time data loss, or an ongoing outage?

• What percentage of the expected data did you lose?

• What is your confidence level in the data you do have? Be specific.

Transparency and specificity are your friends when describing your data. Giving vague answers to specific questions makes it sound like you don’t know the answer. If you actually don’t know an answer, don’t hesitate to say so. It’s infinitely harder to walk back a vague or incorrect statement.

DON’T take questions about data quality personally.

You are not your data quality.

It can be uncomfortable to have the quality of your data under a spotlight. It might feel personal.

When you get questions about the quality of the data, your audience is doing their due diligence. They want to trust the information they are using to make decisions.

You’ll find you naturally get more questions about the data quality when:

• The analytics practice, data governance, or you are new to the company;

• The data set or intake process is new;

• The audience is new to the role.

If you manage these situations smoothly, over time you build trust and likely get fewer questions.

Be upfront, confident, and relaxed when answering these questions. You never want to be on the defensive – that WILL affect the perception of your credibility.

DO: Propose improvements to get better quality data for future

Be proactive. Don’t just share why the data is imperfect. To effectively present to senior leaders, prepare by anticipating their questions.

When the audience hears you say, “The data is imperfect”, they will follow up with, “What can be done to improve it?”.  

Think about what could be done to mitigate or eliminate the data limitations in the future. Even better, begin taking steps to improve the data quality if that falls within the scope of your role.

A fully formed plan isn’t necessary for your presentation. And it might not be appropriate. What matters is that you’ve given it some thought.  

Identifying what you’d need to improve the data quality shows that you’re proactive and understand the business needs.

Final thoughts

No matter the dataset, there will be issues.

Data quality issues don’t make the data unusable. They require you to be excellent at explaining complex concepts to your audience.

As the analyst, you can make recommendations on making the data better for next time. I always think about improving future reporting cycles. Even if it takes time, proactively working toward better data will result in better insights. And, it can save you time in data cleanup and explaining that data cleanup.

The more you practice, the more confident you’ll feel communicating the limitations of your data and your ideas to make it better.

If you enjoyed this article, watch this space! I'm going to be posting more of my thoughts on developing soft skillsets for well-rounded analysts, along with practical tools that can help you in your day-to-day life.