If you've read my last few posts, you may have noticed a pattern: I wrote about coffee shop lines, the matcha hype, Trader Joe's and delivery apps — and in every one of them, I ended up chasing numbers. How much a viral video boosts foot traffic at a coffee shop. How much Trader Joe's makes per square foot. How long DoorDash takes to deliver an order. That wasn't an accident. I'm someone who observes behavior, but I only rest once I can measure what I'm observing.
And that led me to a question that, honestly, runs through my entire career: is marketing an art or a science? For a long time, the field was known as the place for gut feeling, for the creative idea, for the beautiful campaign everyone talks about. There's a famous line, attributed to the 19th-century American merchant John Wanamaker, that sums up the dilemma: half the money he spent on advertising was wasted — the trouble was, he didn't know which half.
More than a hundred years later, we have more data than ever. The question now is a different one: knowing which number actually matters.
Data, metrics and KPIs are not the same thing
First, it's worth separating three words everyone uses interchangeably, even though they mean different things. Data is the raw record: every click, every purchase, every person who walked into the store. A metric is data organized in a way you can track: monthly visits, email open rate, average order value. And a KPI (Key Performance Indicator) is the metric a company has consciously chosen as the thermometer for a business goal.
The difference sounds like a detail, but it isn't. Every company has hundreds of metrics, but it should have only a few KPIs. If everything is a key indicator, nothing is. A good KPI answers one simple question: if this number improves, does the business improve with it? If the answer is "not necessarily," you're looking at a metric — maybe an interesting one, but not a decisive one.
Likes don't pay the rent
In my post about coffee shops, I mentioned that a viral trend can drive a huge spike in traffic to a place in a single week. That's real. But the question an analyst would ask right after is: what about the week after? How many of the people in that line came back without having seen any video? This is where the difference between vanity metrics and actionable metrics comes in — a concept made famous by entrepreneur Eric Ries in The Lean Startup.
A vanity metric looks great in a report, almost always goes up and makes everyone happy, but it doesn't help you make any decision. Followers, views, likes, the length of the line out the door. An actionable metric is one that, when it changes, tells you what to do next.
| Looks like success (vanity) | Shows whether it is success (actionable) |
|---|---|
| Video views | Conversion rate: how many watched and bought |
| Number of followers | Customer acquisition cost (CAC) |
| Line out the door at launch | Repeat purchase rate after 30 days |
| App downloads | Active users and month-over-month retention |
| Campaign reach | Revenue attributed to the campaign (ROAS) |
None of the metrics in the left column is useless — reach and followers matter, especially when a brand is just starting out. The problem is when they become the final score. A coffee shop with 100,000 followers and customers who never come back has a problem no post will fix.
What I learned by measuring in practice
I didn't learn this in a classroom, but on the street — literally on the beach. At one of the largest breweries in Brazil, in Natal, I was a sales supervisor and took on a project along Ponta Negra Beach, a region where our beers had a low market share. The initial diagnosis was intuitive: "our beer is more expensive than the competition." But intuition doesn't tell you how much more expensive, who it matters to, or what to do about it. I spoke directly with about 60 beach bar owners, we designed a proposal with a minimum monthly purchase and equipment support (umbrellas, tables, chairs, buckets), and we started tracking adoption and volume. The result was a 43% adoption rate among partners in the region and a 20% increase in beer volume sold. Without measuring, all I would have had was a feeling that things had "gotten better."
At the same time, there was a problem that looked operational but was actually about data. Draft beer has a short shelf life, just a few days, and kegs were being exchanged every 15 days. The result: product expiring at the point of sale and recurring losses. When you cross two simple pieces of information — shelf life and delivery frequency — the solution becomes almost obvious: we moved the exchange to weekly. That's also where I started using indicators like OTIF (On Time In Full, which measures whether a delivery arrived on time and complete) and NPS (Net Promoter Score, which measures how willing customers are to recommend the brand).
Later, at a delivery super app operating across Latin America, the logic was the same, just at digital scale. As a Key Account Manager, I tracked each partner's indicators — sales, satisfaction, delivery time. And one of the most revealing KPIs was one few outsiders would guess: how much of the partners' marketing budget was actually being used. When I took over the partnerships area, only a small share of the budget was going into campaigns, which meant idle money and unhappy partners. Within three months, we were using between 70% and 80% of the budget and beat the launch GMV target (Gross Merchandise Volume, the total value transacted on the platform) by about 35%.
The lesson I take from both places is that the right KPI is rarely the most obvious one. It's the one closest to the business's real problem.
The KPIs that really count
There's no universal list, but there is a logic: each stage of the customer journey asks a different question, and each question calls for a different indicator.
| Stage | The business question | The KPI that answers it |
|---|---|---|
| Attraction | Am I reaching the right people? | Cost per lead, qualified click-through rate |
| Conversion | Do the people who arrive buy? | Conversion rate, CAC |
| Revenue | Does the investment pay back? | ROAS, average order value |
| Retention | Do the people who buy come back? | Repeat purchase rate, churn |
| Relationship | Do the people who come back recommend us? | NPS, LTV |
If I had to pick a single ratio to track, it would be LTV to CAC: how much a customer is worth over time compared to how much it cost to bring them in. That's what separates a business that grows from a business that just buys growth. And that's why retention matters so much. Classic research by Frederick Reichheld of Bain & Company — the same person who created NPS — showed that increasing customer retention by 5% can raise profits by 25% to 95%, depending on the industry.
Trader Joe's is a good example of a company that focuses on the last two rows of that table, retention and relationship, rather than on attraction and promotion. It has no points program, no delivery and almost no aggressive promotions — in other words, it gives up the classic tools for bringing in new customers or pushing short-term sales. Its logic is different: about 80% of what it sells is private label, so customers can't compare prices with competitors, and they come back because that product only exists there. As one of the company's executives put it in an interview, nobody needs to join a special club to be loyal; loyalty comes from the value delivered on every visit.
And the results show up precisely in the right-hand indicators. In the 2026 American customer satisfaction survey, with about 31,000 respondents, Trader Joe's ranked first among grocery chains, overtaking Publix for the first time. Its sales per square foot are estimated at nearly three times the industry average. And much of its publicity comes for free, from fan accounts: customers who come back and recommend it do the job that, at another company, a paid campaign would do.
The curious thing is that, even with so much data available, the market still doesn't use everything it has. A Gartner survey of 377 marketing analytics professionals found that analytics influences only 53% of marketing decisions. And the obstacles they pointed to weren't a lack of data: they were inconsistent data, data that was hard to access, and the biases of the people making the decisions.
When the number becomes the target
There's a side of KPIs few people talk about. In economics, there's something called Goodhart's Law, which says roughly this: when a measure becomes a target, it ceases to be a good measure. The reason is human. If a team is held accountable for a number, it will find the shortest path to move that number — even if that path doesn't improve the business.
In marketing, the examples are easy to imagine. Reward only lead volume, and you'll get plenty of bad leads. Reward only cost per click, and the campaign will become cheap and irrelevant. Reward only short-term conversion, and someone will discount until the margin is gone. At the delivery app, if the only indicator had been "use the budget," it would have been easy to spend everything with no return; that's why it only made sense alongside GMV and partner satisfaction.
The way out isn't to measure less, it's to measure in pairs: an outcome indicator next to a quality indicator. And to accept that not everything fits on a dashboard. Remember my post on delivery apps, where food safety had no statistical weight in the decisions of most people surveyed, but for me it's a dealbreaker? Data shows average behavior. Anyone who works with data needs the humility to know that the average never tells the whole story.
Marketing is analysis too
If Wanamaker were alive today, he could probably figure out which half of his budget was being wasted. The question is whether he'd know what to do with the answer. Because the challenge in marketing today is no longer collecting data — it's asking the right question, choosing the right indicator and having the courage to change strategy when the number contradicts your intuition.
The American market has already figured this out. The Bureau of Labor Statistics projects that employment of market research analysts will grow 7% between 2024 and 2034, well above the average for all occupations, driven precisely by the growing use of data and market research across every industry.
For me, this is where everything I've done so far comes together: the beach bar in Natal, the delivery app's partner dashboard, the coffee shop line in Manhattan. In all of them, the question was the same — what is really happening, and how do I prove it? This intersection of marketing and data analysis is exactly what I want to dig deeper into from here on.
And you: what number do you track at work — and does it really show whether things are going well?