I recently watched a reel by Rory Sutherland, Vice Chairman of Ogilvy, where he criticises the over-reliance on data in business. Companies use numbers as a defensive shield, everyone draws the same conclusions, and competitors end up copying each other without realising it. Rather than using data as an engine, they use it as camouflage.
Even though I'm very much a data person, I get what he means. What I didn't see in the reel was a complete solution.
Numbers tell you something. Data tells you something. But how you interpret it, or who you let interpret it, matters just as much. People who aren't comfortable with data have to trust what a specialist tells them. And if that specialist knows enough, they can always spin the numbers in a way that works in their favour. I wrote before about how to check if your agency is showing you the right numbers.
But the answer isn't to abandon data. Because ignoring data is at least as dangerous. Just because you sometimes have to look at something ugly doesn't mean you're better off being blind.
Flying blind
For a long time I had to make do with the tools that were available. These days I solve that by connecting data sources myself and building dashboards, but in the past I was stuck with what GA4 could tell me, or what a backend could spit out. I squeezed everything I could out of pivot tables and complex formulas, but that usually ate up enormous amounts of time just to get a snapshot of the current situation, which meant you were still missing things.
A few years ago, a client made a change to their shopping cart. It was a no-brainer. The CTA sat below the fold on mobile. That needed fixing, right?
The change went through and the KPI they were tracking showed improvement: conversion rate went up slightly. But something else changed. Average order value started dropping. By about 5 to 10 percent.
The end result? Revenue went down.
Average order value kept declining week after week and nobody connected it to the cart change. AOV fluctuations happened before. Maybe the new converters were just customers who typically put less in their basket, a different product mix or something?
It took not weeks but years before we finally pinpointed the real cause. When I built a graph showing average order value over time, using actual order data combined with traffic data rather than just what GA4 could tell us, there was a clear crash at the exact moment of the optimisation. And it clearly wasn't temporary, because the effect was still there two years later.
One KPI went up. Another, more important one went down. Nobody looked at both at the same time.
That's not the problem Sutherland is describing. But it is what happens when you don't look at data at all. You keep walking around with extra weight on your back, weight you could have easily thrown off if you had actually looked at what was going on.
Looking at things the wrong way
But Sutherland is also right. Sometimes data is the problem. Not because data looks backward, but because someone looks at it and draws the wrong conclusions from a bad interpretation.
I ran into this pattern at another client, but I'll use a metaphor to make the point clear.
Large retailers generate more revenue when there are more Saturdays in a month. A month typically has four or five Saturdays. The chance of getting five is higher for months with 31 days.
Take November. In 2024 and 2025 it had five Saturdays. This year, only four. What you typically do is compare against last year. So naturally you see a steep decline for that month.
The reaction to a wrong interpretation makes things even worse: we have a gap in orders and revenue. We need to launch new campaigns, run promotions, maybe even cut prices.
But if you just look at it differently, by week number for instance, you'll see there's barely anything wrong for that period.
The result? Higher costs, lower average order value, for a problem that didn't exist in the first place.
The problem does look real though, and if you don't have the means to look at it from another angle, you go straight into panic mode. You turn a non-existent problem into an actual one.
Because if you lower your prices based on that wrong interpretation to compensate for the decline, you create the revenue loss you were trying to prevent. The drop that wasn't there now is. Caused by your own reaction to a calendar artefact.
It sounds like a strange story, but I've come across this more than once. This specific case is about a calendar effect, but I've seen management spot a problem that isn't there through other lenses too. When you're stuck with one fixed dashboard that doesn't give you enough data, you still have to make decisions based on it.
Bad news dressed up as good news
An example that's come up in previous articles. A client saw that the share of returning customers was improving. Repeat share was going up, so retention clearly wasn't the issue this year.
But repeat share had gone up because the absolute number of repeaters was only declining slightly, while the absolute number of new customers was dropping off a cliff.
Repeat rate? Nobody was looking at that. Repeat share told them how satisfied customers were with the brand and how well their retention marketing was working.
On top of that, they assumed that everyone coming in through direct traffic or organic was a returning customer. Because you must already know us if that's how you arrive. That was an assumption, not a fact. And that assumption coloured everything that followed.
What I was able to show them afterwards, and at that point I was still doing it with pivot tables in Excel, was that repeat rate had been declining for years. Combine a shrinking pool of new customers with those same new customers repeating less often, and you get a cocktail that tastes very bitter.
You can read more about this specific example of looking at the wrong KPI with the wrong attribution and the wrong interpretation here: why your most loyal customers are quietly leaving.
It can go differently
That same historical data, read correctly, can help you look forward.
Something I noticed at one of my clients: the data showed that conversions during the winter months had the highest average value. It made total sense for that business field: people who buy early tend to spend more. But there was something else underneath. Those early conversions filled the pipeline for the entire peak season. They spread evenly across April through August. Investing later only catches the last-minute demand concentrated in the next month or two.
Historically, barely any marketing budget was allocated to that winter period, because it was off-season. When we tripled that budget, the return per euro invested stayed well above that of the peak months. Every euro spent in that period outperformed the months that traditionally received the most budget.
The pattern was sitting in the historical data. But it wasn't a self-fulfilling cycle. We didn't see that we always generated more revenue during that period. We saw that we could potentially generate more revenue during that period at a lower cost. And the outcome matched the prediction.
The difference with the previous examples? Someone who understood the data and asked the right questions.
Nuance
I'm sometimes asked to provide data that helps push a decision in a particular direction. I try to always be transparent, even when it means arguing against my own interests. Because if there's one thing data can do, it's bring nuance. Nuance without emotion attached to it or a political decision that needs justifying.
A conversion rate that drops can be good news. A rising repeat share can be bad news. Growing revenue can hide shrinking margins. A declining month can be a missing Saturday. A shorter session duration might mean the customer journey on your website just got better.
No single number tells the full story. Context, context, context.
Anyone who isolates one metric and decides based on that misses everything around it. Anyone who doesn't look at data at all misses everything. But anyone who looks at data and reads it wrong makes the worst decisions of all. Because they carry the confidence of someone who thinks they know what's going on.
Repeating the same mistake every day
Sutherland is right about the diagnosis. Companies misuse data to cover themselves. They look at numbers to appear rational. They make the same decisions as their competitors because they're looking at the same spreadsheets.
But what Sutherland seemed to be suggesting, that companies should rely less on data and more on intuition and creativity, works better when you're Rory Sutherland than when you're a small business trying to grow. For most companies, the answer isn't less data. The answer is someone who knows what to look at, who understands the context, and who has the nerve to say that the 20% decline is a missing Saturday. I wrote before about why marketing reports stay so simple. The same applies here: without the right context, every report is a half-truth.
Data looks at history, that's true. But why do you study history in school? Why is history even worth learning about?
To learn from past mistakes. To understand patterns from the past.
A donkey doesn't trip over the same stone twice. But a business without the right data insight trips over it every day.