Every founder I work with has more data than they did five years ago, and less confidence about what to do with it. The dashboards are full. The weekly report runs three tabs deep. And the one question that actually matters — where does the next dollar go? — is no easier to answer than it was back when the whole plan lived on instinct and a single spreadsheet.
More information was supposed to fix that. It didn’t, because the thing that’s actually scarce changed, and most companies are still organized around the old scarcity.
Data used to be the hard part. Now it’s the cheap part.
Ten years ago, having the numbers was the advantage. You paid for the analytics stack, the attribution model, the research panel, a customer database, and someone who could run all of them. The data cost real money, so owning it set you apart.
Now it’s free, or close to it. Analytics are cheap, dashboards ship with every tool you buy, and AI will hand you a clean paragraph on what moved last week plus a tidy recommendation to “improve conversion” or “lean into retention.” Thirty seconds, no expertise required.
Which means the data can’t be your edge anymore. When a competitor has the same numbers and the same auto-generated advice, neither of you gains a thing from owning them. What’s left is the part nobody automated: deciding what the numbers mean and what to do about them.
A dashboard reports the past. Traffic up, conversion down, acquisition cost creeping again. Then it goes quiet, right at the moment you have to make a call. Closing that gap is the actual work, and it asks for three things no chart can give you.
Data doesn’t tell you what to do. It tells you what to ask.
What customer insight actually requires
You have to find the number that’s actually the problem. A falling conversion rate isn’t an answer; it’s a question with a dozen suspects. Maybe the offer went stale. Maybe a standing discount trained your best buyers to wait for the next one. Maybe a paid channel started spending on cheaper, lower-intent traffic that was never going to convert, in which case you don’t have a conversion problem at all. You have a traffic problem in disguise. The metric won’t choose among the suspects. Your job is to ask what would have to be true for each one to be the answer, then go find out. One chart, four different decisions, and picking the right one is judgment, not reporting.
You also have to develop real customer insight — knowing what your customers actually want, not what the funnel logged. Behavioral data shows you where people clicked. It rarely explains why, and the why is where the money sits. Take a pattern I’ve seen play out more than once. A bath & body brand, somewhere around $25M in revenue, has a repeat purchase rate of 16% against a category benchmark of 20–30%. The dashboard is unambiguous: you have a retention problem. So the team does everything a good team would. They build out email and SMS, launch a loyalty program, pilot a subscription, invest in post-purchase content. Six months and a few hundred thousand dollars later, repeat purchase has crawled from 16% to 18%, and everyone is quietly demoralized.
Then someone finally talks to customers, and the picture flips. Roughly half of first purchases turn out to be gifts, bought by people who were never going to repurchase no matter how good the email flow. Isolate the customers who bought for themselves and their repeat rate is closer to 22%, perfectly healthy. The 16% was never a retention problem. It was an acquisition-mix problem hiding inside a retention metric. The real fix lived upstream — change the packaging that made the product read as a gift, and shift acquisition toward self-buyers — and no amount of lifecycle work was ever going to find it, because the deciding fact (gift versus self-purchase) wasn’t anywhere in the stack.
The cost of getting that wrong isn’t only the few hundred thousand spent on the wrong infrastructure. It’s six months executing against the wrong hypothesis while the actual problem got worse. It’s also a harder conversation with your investors when you ask for a longer runway because your tactical retention plans couldn’t deliver. Misdiagnosis carries a financial cost, an opportunity cost, and a political one, and none of them are recoverable.
The deciding variable is usually outside the dashboard
The bath & body story turns on a detail worth sitting with: the fact that broke the case open — gift versus self-purchase — was nowhere in the analytics stack. That’s the rule, not the exception. The variable that decides what you should do is often the one your tools never captured: why a customer actually bought, what they expected versus what they got, why the ones who left went, and where they went instead. Some of it you can pull from a well-built post-purchase survey. Most of it only comes from genuine customer insight — talking to customers, the work most brands stopped doing once the dashboards got good. When the answer isn’t in the tool, more tooling won’t surface it.
And you have to know what your business can actually pull off. A brilliant strategy your team can’t execute, your operations can’t support, or your margins can’t survive is a wish with a deck attached. Good interpretation stays inside the lines of what’s real: true about your customers, and true about your people, your product, and your P&L.
This is the reason “we’re data-driven” stopped carrying any weight. Everyone is data-driven now. The companies that win read the same numbers everyone else has and make three or four correct, expensive decisions while their competitors are still reformatting the dashboard.
How to build this into the way you run the company
Interpretation sounds like a gift some people have and others don’t. It’s mostly a set of habits. You don’t need a bigger data team or a smarter tool. You need to change the questions you ask and how you run the room, and most of it you can start at your next meeting.
Put the decision before the number
Refuse to discuss any metric in the abstract. Before a number goes up on the screen, the question on the table is simple: what would this change? If it can move in either direction and your behavior stays the same regardless, it doesn’t belong in the meeting. It’s there to make everyone feel informed.
That one rule clears out most of the clutter. A startling share of what teams report every week has no decision attached to it whatsoever.
Ask sharper questions
When a number moves, the reflex is “why did that happen?” Reporting question, reporting answer. The questions that actually produce insight sound more like:
- If this were the only number we looked at this week, what would we do differently?
- What decision does this change, and who owns it?
- What would have to be true for this to be a real problem and not just noise?
- Did we convert worse, or did we just buy worse traffic?
- What are customers telling us that this chart can’t?
- What customer insight do we believe in our gut that we can’t see on this dashboard?
- If our first explanation is wrong, what’s the next most likely one?
That last question earns its place. The teams that read data well are the ones in the habit of naming the second explanation before they marry the first.
Run the weekly meeting to produce decisions
Most weekly metrics meetings are a reading. Someone narrates the dashboard, everyone nods, nothing changes. Swap the recital for something that forces a call:
- One page, not three tabs. Three to five metrics, full stop. Put everything on the page and you’ve prioritized nothing.
- Four lines per metric: what changed, what we think it means (a real hypothesis, not a shrug), what we’re doing about it, and how we’ll know if we got it wrong.
- One piece of customer evidence, and not a flattering one. The point isn’t to pass around a glowing review so everyone feels good and nothing changes. Make the team bring the tension you’d rather not look at: a complaint, a shipment that got botched, a rude post, the long-time subscriber who just cancelled, the email with the hardest feedback in it. Rotate who brings it. Your job isn’t to feel warm and fuzzy; it’s to make the customer feel that way, which means facing the ugly thing and fixing it.
- One closing question: what’s the single thing we’re doing differently after this conversation? If the honest answer is nothing, you just learned something about the meeting.
Thirty minutes, and it replaces an hour of reading numbers out loud.
Know which customer metrics you watch and which you act on
Some numbers are vital signs. You keep an eye on them so you’d notice if something broke, but you don’t grab the wheel every time they twitch. Others should move money and attention directly. Trouble starts when a founder treats both kinds as equally urgent and chases noise as if it were signal. Sort them. The three to five you act on go on the one-pager. The rest live in a dashboard you glance at, not a meeting you run.
Talk to customers on a schedule, not in an emergency
Most teams only call customers once something is already on fire, so every conversation happens late and under pressure. A better approach is to build a simple, unglamorous cadence: a few customer conversations each month, a regular read of support and sales themes, and a short interview when someone cancels. You don’t need to stand up a research department. You just need an ongoing relationship with your customers, so the next time the data raises a question, you already have a feel for where to find the human answer.
Pressure-test the recommendation before you fund it
Before you commit to whatever the data seems to point at, ask yourself all the whys and hows. Can the team execute this well? Can operations carry it? Does it survive contact with your margins? And finally, does it still make sense given the customer conversations we’ve had recently?
Won’t AI just do this for you?
Fair thing to wonder, and the answer is no, for a specific reason. AI is genuinely good at the cheap half of this: gathering, summarizing, generating the obvious next suggestion. It’s weakest at the half that counts, which is weighing a murky signal against what you know about your particular customer and what your particular business can deliver, then owning the decision when someone asks why you made it. As generating data and generic advice becomes free, the judgment that turns either one into the right move gets more valuable, not less. The dashboards are converging. Interpretation and customer insight are where you get to pull ahead of your competition.
What to do with this
If you’ve got a stack of reports and still can’t say with confidence where the next dollar goes, the fix isn’t more data. It’s turning the data you already have into a short, ranked list of decisions, built on what your customers actually want and what your company can actually do. Most of how to get there is above, and you can run your next metrics meeting differently tomorrow.
If your dashboard is full but the next move is still unclear:
That’s the kind of thing I help founders work through as a Fractional CMO. As a starting point, I’ll get into the data you already have, talk with your team and a few of your customers, and come back with a short, ranked list of the moves I’d make and a couple I’d stop. It won’t be a canned, automated audit; it’ll be a real conversation that leaves you with a point of view you can act on. It’s also a low-stakes way for us to find out what working together feels like. If that’s useful, let’s talk →.
