The KPIs a small business should track

You open the Meta dashboard and there are dozens of numbers. Reach, frequency, impressions, one kind of click, another kind of click, cost per result. You close the tab and still don't know whether the month was a good one.

You're not short on metrics. You have too many, and none of them comes labeled "this one matters."

I'll tell you which four to look at, in what order, and what's normal for each. One warning first: there is no public, measured study of these numbers for Chile. The ranges that circulate on local agency blogs don't say where they come from. The ones I use here are from the United States, they're measured, and they come with their fine print.

A KPI answers something you need to decide

KPI stands for "key performance indicator." In plain terms: a number that answers something you have to settle before you decide what to do. If a number answers nothing, it's a data point. Maybe an interesting one, but not a KPI.

A small business needs to answer four things, in this order, because each depends on the one before:

Is the right audience reaching me?
Does that audience do anything?
What does a new customer cost me?
Does it leave me money?

If your answer to any of them is "I don't know," the numbers below it don't mean anything yet. That's your task.

1. Is the right audience reaching me?

The number: what you pay each time someone lands on your site from an ad.

What's normal. On Google search, the 2026 average is $5.42 per click. On Meta, in campaigns that aim to get the person to leave their details, $1.80. In campaigns that only aim for visits, $0.60.

It varies a lot by industry. On Google, a click for a lawyer costs $9.87. For a restaurant, $2.05.

The trap. Pushing this number down isn't the goal. A cheap click from someone who never buys is wasted money, just in smaller installments. The people who publish these benchmarks say so themselves: if the click gets cheaper and fewer of the people arriving do anything, the final cost may not improve. This number only reads correctly next to the next one.

2. Does that audience do anything?

The number: out of every 100 people who arrive, how many leave their details or buy.

What's normal. On Google, about 8 out of 100 (8.18). On Meta, in lead campaigns, almost the same: 8.5 out of 100.

This number says more about your site than about your ad. The ad already did its job: the person arrived. What happens next depends on whether they understand what you sell, whether they trust you, and whether the form or the button works.

There's a data point that backs this up. In 2026, cost per lead on Google fell for the first time in five years, even though clicks kept getting more expensive. It fell because more of the people who arrived ended up doing something, not because the auction got cheaper.

If yours is well below 8, look at your page before you touch the ad.

3. What does a new customer cost me?

The number: what you spend on ads, divided by the customers you get.

What's normal. First, a clarification almost nobody makes: benchmark tables don't measure customers, they measure leads — people who left their details. They haven't paid anything yet.

With that clear, the average cost per lead is $66.69 on Google and $27.39 on Meta. The extremes are wide: $131.63 on Google for lawyers; on Meta, $12.30 for employment and $61.56 for dentists.

To get to cost per customer, divide that number by the share of leads that end up buying. That share is yours. No table has it.

The trap. Don't compare $66.69 with $27.39 and conclude Meta is 2.4 times cheaper. Each platform counts a lead differently, and the campaigns that produce one work differently: on Meta, the click in a lead campaign costs almost three times what it does in a traffic campaign, $1.80 versus $0.60. It's the same problem as testing Google and Meta halfway: two numbers that look comparable and aren't.

4. Does it leave me money?

The number: how much you sell for every dollar you put into ads, against what's left of that sale after your costs.

This is the only one of the four without a table that's useful to you, and that's on purpose. For example: a customer that costs you $70 to win is a good deal if it leaves you $200 and a bad one if it leaves you $40. Your margin decides that, not your industry.

The break-even math is one division: 1 divided by your margin. At a 25% margin you need to sell 4 for every dollar invested just to get to zero. I broke it down in full, with two examples in money, in what is a good ROAS in your sector.

The ones you can ignore for now

Reach, impressions, followers, likes, frequency. They're not useless. None of them answers one of the four things above.

They help later: when one of the four comes out badly, these help you look for the cause. Until then, they're noise that makes you feel like you're measuring.

Before reading any of them: did the platform see the sale?

Numbers 2 and 3 are only real if the platform managed to see what happened. If your sale closes over WhatsApp, by phone, or at the counter, it didn't. What it shows you is the last thing it managed to record — a click, a message, a form — and it calculates everything else on top of that with total confidence.

Of 23 Chilean small-business sites we scanned in July, 12 couldn't tie a single dollar from Instagram to a customer who actually paid. And of 8 Chilean B2B startups we scanned afterward, none measured properly. Those are small samples, not a national study, but the pattern repeats.

If you're in that group, your KPIs aren't good or bad. They're fiction with decimals. The way to check is in how to know if your advertising works.

Compare the right way

There are two different comparisons, and they're for different things. Against your previous month, to know whether you improved. Against your industry, to know whether you're being charged more than everyone else for the same thing.

For the second one, the fine print on these numbers: they're from the United States, in dollars, from campaigns measured by a single provider (WordStream by LocaliQ), with data from April 2025 to March 2026. They say "thousands of campaigns" and don't specify how many or whether the figures are averages or medians. They work as an order of magnitude, not as an official yardstick.

And there are two typical ways to get the math wrong: comparing your mix of new and returning people with someone else's, and forgetting that whoever buys this month may have seen the ad last month. Both are covered in the ROAS article.

What to do this week

Three steps.

One. Write down the four answers today, with numbers, on paper. If one of them is "I don't know," that's your task for the week. Don't move on to the next.

Two. Ignore the rest of the dashboard for 30 days. Yes, all of it.

Three. Once you have all four, compare them against the reference above, with the fine print in hand. If you're not sure how much to spend for those numbers to mean anything, start with how much to spend on ads, for real.

Frequently asked questions

What is a digital marketing KPI?
A key indicator: a number that answers something you need to decide. If a number doesn't change any decision, it's a data point, not a KPI.

How many KPIs should a small business track?
Four are enough to start: what each click costs, out of every 100 who arrive how many do something, what a new customer costs, and whether that customer leaves you money. More numbers don't give you more clarity, they give you more noise.

What are the most important marketing KPIs for a small business?
Cost per new customer and return on your margin, because those are the ones that connect your spending to your cash. The other two, cost per click and share who do something, help you find where the problem is when those two come out badly.

How do I know if my KPIs are good?
With two comparisons: against your previous month, to know whether you improved, and against your industry, to know whether you're expensive. For the second, use benchmarks as an order of magnitude and with the fine print in view: most are from the United States and in dollars.

What do I do if I sell over WhatsApp and can't track sales?
Start by writing down by hand, for one week, how many conversations come from ads and how many end in a sale. It's crude, but it's yours. Without it, the dashboard numbers are calculated on a click or a message, not on a sale.

If you haven't decided where to start yet, look at Google Ads or Meta Ads. And if you want to see which of these four you can measure on your site today, and how they compare with your sector's, scan your site for free. No install, no agency.