Data-driven marketing: how to actually use your numbers.

The problem often looks like sophistication: a beautiful dashboard with sparklines and a dark-mode toggle. Many of those dashboards get glanced at, nodded over, and then ignored when the real decisions get made on gut feel in a Tuesday meeting.

A survey of executives reported in Harvard Business Review tracked exactly this question, and the share of firms that call themselves data-driven has actually been going down, not up, slipping from 37.1 percent in 2017 to 31.0 percent a couple of years later. More tools, more data, less confidence.

The fix is a smaller, sharper set of questions, asked on a schedule, answered by numbers you trust.

The first thing to do is sort your numbers into two piles. One pile is vanity metrics: numbers that go up, feel good, and change nothing about what you do next. The other pile is decision metrics: numbers that, when they move, force a different action out of you. Most teams report almost entirely from the first pile and wonder why nothing improves.

• Vanity. Impressions, total followers, page views, email opens, "reach." They rise with spend and say little about whether any of it paid back.
• Decision. Cost to acquire a customer, what that customer is worth over time, the rate at which visitors convert, the return on each channel, and how many customers stick around. These hurt to look at, which is exactly why they are useful.

A simple test: for any metric on your report, finish this sentence out loud. "If this number drops by 20 percent, I will _____." If you cannot fill in the blank with a real action, the metric is decoration. Keep it off the page you actually look at.

You do not need forty-one charts. You need five numbers you understand cold and can recite without looking. Each one, and why it earns its seat:

• Customer acquisition cost (CAC). Total sales and marketing spend divided by the number of new customers it produced. This is the price of growth. If you do not know it, you are bidding blind.
• Lifetime value (LTV). What an average customer pays you across the whole relationship, minus what it costs to serve them. The number that tells you how much you can afford to spend in the first place.
• Conversion rate. The percentage of people who take the action you care about, whether that is a booked call, a purchase, or a form fill. Small moves here ripple through every other number you have.
• Channel ROI. For each source of traffic and leads, what you put in versus what came back. This is the number that tells you where to pour more fuel and where to stop lighting money on fire.
• Retention. The share of customers who are still with you a month, a quarter, or a year later. Acquisition gets the applause, but retention quietly decides whether the whole machine is a business or a treadmill.

Notice the relationship between the first two. The old rule of thumb is that LTV should comfortably clear CAC, and many operators want it several times higher to leave room for the costs the spreadsheet forgot. You do not need a perfect model. You need to know whether a new customer is worth meaningfully more than it cost to win them. If you cannot answer that, nothing else on your dashboard matters yet.

A dashboard feels like truth because it is visual, real-time, and tidy. But a chart is a set of choices, and you usually did not make those choices, so you cannot see them. As HBR put it in a piece on how dashboards mislead, single-screen snapshots of your KPIs can be visually elegant and intuitive, which is precisely the danger.

The usual culprits are mundane. A date range that quietly defaults to the wrong window. A metric averaged across two groups that behave nothing alike, hiding the one that is on fire. A "conversions" line that counts a newsletter signup the same as a fifteen-thousand-dollar deal. None of these look like errors. They look like Tuesday. And they steer real budget.

The defense is to ask, of any number you are about to act on, three small questions. What exactly is being counted? Over what window? Compared to what? If you cannot answer all three in one breath, do not move money on it yet.

Attribution is the art of deciding which marketing touch deserves credit for a sale. Someone sees an Instagram ad, forgets it, googles you a week later, clicks a brand search ad, reads a blog post, and finally books a call from an email three days after that. Which one gets the credit? The answer is all of them and none of them cleanly, which is why attribution is both essential and a little bit of a polite fiction.

Google's own documentation is direct about this. As the GA4 guide to attribution explains, attribution is simply the act of assigning credit for important actions to the various ads, clicks, and factors along the path, and by default all that credit tends to land on the last thing someone clicked. Last-click is easy to compute and often wrong, because it hands the trophy to the bottom of the funnel and starves everything that did the real persuading earlier.

So use attribution, but hold it loosely. Treat it as a directional argument, not a courtroom verdict. The teams that get burned are the ones who cut a channel because a single attribution model said it underperformed, only to watch their whole pipeline sag a month later because that channel was doing the quiet work no model could see.

This is also where most people meet Google Analytics 4 and feel the floor move. GA4 swapped the old session-and-pageview world for an event-based model, which is genuinely more flexible and genuinely harder to read on day one. The reports moved, bounce rate hid, and "engagement" showed up under a name nobody recognized. GA4 is a real upgrade for understanding behavior across devices, and it will happily let you stare at it for an hour and walk away with nothing.

The practical move is to resist the urge to master every report. Configure a small number of key events that map to money, mark the ones that matter as your conversions, and ignore the rest of the interface until you have a specific question. Most people drown in GA4 because they open it with no question and let the dashboard set the agenda. A telescope is wonderful, but it does not tell you which star to look at. Walk in with a question, get the answer, close the tab.

Lambrecht and Tucker catalogued the four mistakes most managers make with analytics, and the throughline is that people reach for data without first reaching for a clear question. When the question is fuzzy, the data will cheerfully confirm whatever you already believed. This is how the pattern tends to show up, and the small habit that beats each one.

• Starting with the data instead of the question. Open the report only after you have written down the decision it is meant to inform. No question, no chart.
• Confusing correlation with cause. Two lines that move together are not proof one caused the other. When the stakes are real, run a small test before you commit the budget.
• Drowning the signal in noise. Reporting everything is the same as reporting nothing. Cut your dashboard down until it hurts a little.
• Trusting the number over the obvious. When a metric and your eyes disagree, do not just believe the metric. Go check how it was built. Often the number is broken, not your judgment.

None of this requires a data science team. It requires the discipline to ask a question first and to treat your own dashboard with a healthy, friendly suspicion.

The teams that actually run on data do not have better tools than you. They have a smaller, duller, more consistent habit. The reporting that changes companies fits on one page and gets looked at on the same day every week.

• One page, five numbers. CAC, LTV, conversion rate, channel ROI, retention. Each one shown against last period and a simple target. That is the whole report.
• One question per number. Next to each metric, write the decision it informs. "If CAC rises, we pause the worst-performing channel." Make the action obvious before the number moves, not after.
• One standing meeting. Fifteen minutes, same time weekly. Not to admire the numbers, but to decide one thing differently because of them. A meeting that ends with no decision is just a slideshow.
• One owner. A single person responsible for the numbers being right. Shared ownership of data means nobody trusts it, which means everybody quietly goes back to guessing.

That is it. No new platform, no migration, no consultant. A page, a habit, and the willingness to actually change course when the page tells you to.

Data will not make your decisions for you, and any tool that promises it will is selling certainty it cannot deliver. The numbers narrow the range of reasonable choices and catch you when you are about to do something dumb. The judgment is still yours, and it should be.

So start embarrassingly small. Pick your five numbers. Get them onto one page you trust. Ask one honest question of them every week and let the answer actually move a dollar. You will learn more in a month of that than in a year of installing tools you never open.

If you get there and the foundation feels shaky, the tracking is a mess, the channels are muddy, or you just want a second set of eyes on what the numbers are really saying, that is the kind of unglamorous work we do at Mining Wells, mostly around growth, ads, and conversion. No promises, no all-knowing dashboard. Just the boring, durable habit of turning your numbers into decisions you can stand behind.