The beauty — and business — of imperfect data
If there’s one universal truth in digital analytics, our data is messy, biased, and, at times, downright unreliable.
And yet, for decades, businesses have treated data like some sacred, infallible gospel.
Well, here’s the bad news: no matter how much you clean it, structure it, or feed it to your AI overlords, it will never be perfect for more than a brief moment, and then someone will mess it up.
…again.
But here’s a bit of good news — it actually doesn’t have to be perfect.
The real magic isn’t in chasing perfection; that’s a lost cause that will sidetrack you from both actions and insights.
The real magic comes from learning how to extract massive value from imperfect, biased, and “flawed” data.
Because if we know where the cracks are, we can still build something solid.
A lesson from the Pacific
Horrible headline, right?
But please indulge me for a minute..
Let’s rewind to the Battle of Midway. As history goes, one of the factors that secured the Americans the victory was technology because they used radar, giving them a critical edge.
The funny part is that the Japanese actually also had radar, but their leadership didn’t believe in it, so they had simply sidelined the technology and not invested in it.
The result?
Radar vs. no radar?
A decisive loss.
And a painfully simple lesson.
Having access to technology is not enough, you have to use it.
And just because your data isn’t flawless doesn’t mean it isn’t valuable.
It’s about how you use it.
Value does not come from volume but from understanding, use, and activation.
Those who dismiss imperfect data outright are, well, the ones left scrambling when the competition moves faster because they found the gold nuggets in the dirt.
The “garbage data” that can make you millions
Digital analytics has been around for more than three decades in one form or another, and the industry consensus is pretty clear most of the data we collect online is, on a global quality scale compared to BI, rubbish.
Tracking errors, missing cookies, broken attribution- if you’ve worked in this space for more than five minutes, you know the list goes on.
But does that mean it’s useless?
Not even close.
Consider that companies have spent billions optimizing online performance, tweaking landing pages, and A/B testing ad creatives, getting bottom line value out of the data by a systematic approach.
How we normally use it.
Far from perfect, but it still works and delivers results.
But that’s the easy stuff.
What about the blind spots? The indirect insights? The behavioral signals that imperfect data reveals?
All the times the customers do not go with “the obvious choice”.
Instead of dismissing the “mess,” the smartest companies are using it to understand:
What biases drive user behavior?
Which segments respond irrationally to certain messaging?
How do external factors (like market momentum) impact conversion trends?
It turns out that looking at the noise rather than just the clean, structured metrics tells you a lot more about what’s really happening.
The power of biased data
People love talking about the dangers of bias in data, and yes, if you’re making medical decisions or sentencing people in court, bias is a problem.
But in business?
Bias is often exactly what you need to understand.
If 70% of visitors to your website bounce, it tells you a lot about your first impression.
If a product page gets shared wildly but doesn’t convert, it reveals social influence vs. purchase intent.
If users consistently behave differently from what your models predict, it’s most likely not bad data, it’s just human nature.
What’s critical is knowing how to read biased data rather than fight it.
Because no matter what tool, dashboard, or machine-learning model you deploy, your customers will continue being irrational, emotional, and… let’s be real… often fairly unpredictable.
Tips for imperfect data
It’s the price of working with data from people. They do not think linearly, and they are not always rational, but they are certainly always open to a new fad, craze, or trend.
So 4 friendly ideas to help find the value in the bias.
Stop trying to “fix” the bias lean into it
Instead of obsessing over accuracy, ask: What does this bias reveal about behavior?Look for patterns, not precision
The exact numbers might be off, but directional trends tell you where things are heading.Embrace the uncertainty
If you only act on data that’s 100% “trustworthy,” you’re already too late.Combine multiple weak signals
One flawed data source? Unreliable. Five flawed sources pointing in the same direction? A goldmine.
It’s damn counter-intuitive and goes against everything we have learned in school about logic, math and problem solving, but the point is that we simply can’t always use that type of logic on people…
That would be too easy, right?
The radar you refuse to use
Your data is imperfect. It’s incomplete. It’s biased.
Because so is the real world.
And the companies that figure out how to harness those imperfections?
They get ahead, they win a little more every day.
Just like the radar in World War II, the value isn’t in perfection. It’s in how you use what you have.
So the question becomes, will you embrace your customer's imperfection and bias, or will you wait until your competitors do?
—
This article was inspired by a keynote I gave at CDOIQ Nordic 2025 on other ways we can use the analytics data we already have gathered by thinking outside the box.
Sign up for my newsletter on Linkedin (Please, Make me Think) to get info, when the video is released on Youtube”.
About me:
Director of Data Innovation at IIH Nordic, Board Member CDOIQ, Community Builder & International Keynote Speaker- I Connect People, Data and Business Outcomes.
With more than 20 years of experience, I’m on a mission to make people realize the importance of the business and commercial aspect of data.
I’m here on Medium to write and help people in Data, AI, Digital and Web analytics navigate better in their Journeys to make stuff happen.
