How do you know if your Facebook and Instagram ads are working?

You can judge them without being an expert. Here is what to look at, and what to ask.

Short answer: follow the numbers down to paying customers, and allow for how long buyers take to decide. Then ask about anything nobody can explain.

If you're still deciding whether to pay for help, start with whether hiring an agency is worth it. This article picks up after that decision. Say someone runs your ads, or soon will. How do you tell whether the work is paying off?

I run Facebook and Instagram ads for businesses, so I'm one of the people you'd be judging. Read what follows with that in mind. The checks work the same whoever runs your ads, me included.

Do likes and followers show your ads are working?

In the early days of Facebook and Instagram, owners watched likes and followers. I call those vanity metrics. Likes and followers are easy to count, and they can feel like progress.

They count attention. Attention has a job, since nobody books a business they've never heard of. But a like doesn't fill a room or an appointment.

So start with a different question: what were the ads meant to change? More bookings, more new customers, or old customers coming back? Write that down before you open a report. Every number after it gets judged against that answer. Setting that goal belongs to whoever is responsible for your marketing, even if someone else runs the ads.

How do the numbers in an ads report fit together?

Feeds, a website or a chat each report a different first number. All lead down one chain to enquiries, bookings, payment.

The numbers cascade down. You need a certain reach to get enough touch points, and enough touch points to get a lead. In my client work, this chain is what owners mix up most.

Here is the chain for a business that sells appointments:

  1. Reach: how many people saw the ad.
  2. Touch points: views, clicks, taps and chats started.
  3. Enquiries: people who asked about booking.
  4. Suitable enquiries: people you can serve.
  5. Bookings: appointment slots that were filled.
  6. Payment: customers who came and paid.

Each step is a different event. A click isn't an enquiry, and an enquiry isn't a booking. The numbers shrink at every step, and that's normal. What matters is where the chain thins, and why.

Watch how a report labels its numbers. Meta's ads report has a "results" column. It counts whatever the campaign was set up to get, and that may be a click or a chat started.

Placements add a layer. Your ads can appear in feeds, Stories and Reels. The report can break the numbers down by placement, so you can see where results came from. What counts as a result is set by the campaign, though. One set up for video may report views. One that opens a WhatsApp or Messenger chat usually reports conversations started. Those don't add up into one total. Follow each one down the same chain, to bookings and payment.

Two habits keep the chain honest. Read every rate next to its count. Half of ten enquiries booking is five people. Half of a hundred is fifty. And a gap isn't a zero. If nobody recorded where bookings came from, the answer is unknown, not none.

What can you judge yourself, and what needs an expert?

Three questions: was the work done and what happened, which you can check, and what the ads caused, which you judge.

Three questions sit inside "is it working?" They have different answers, and different people can answer them.

Was the agreed work done? You can check this. Ask to see the ads as they ran, and the spend as the platform reports it. Did they run on the agreed budget, in the agreed places? What changed since the last review?

What happened in the business? You can check this too, often better than anyone. Enquiries, bookings, payments and returning customers sit in your own records.

What difference did the ads make? This is the hard one, because some of those customers would have come anyway. Working out how many takes a method, and a specialist should explain it to you.

You don't need to follow the method step by step. Ask what it assumes, what it leaves out and how sure it is. Then judge whether the answer fits what you see in your own records. The same goes for technical set-up, such as what the tracking counts. Ask for a plain explanation.

The craft belongs to the person running the ads. Judging what it produced belongs to you.

I see running ads as an art mixed with science. I pay other people for design, video and photography. I judge their work the way I judge ads: room to experiment, but tracked and measured.

What is the simplest check you can run yourself?

The simplest check I'd suggest takes two questions, asked over a set period. Did sales go up? And can the rise be traced to the ads that ran?

The first answer is in your own books. The second takes a little record keeping. Ask new customers how they found you. Note which bookings started as a chat from an ad. Keep returning customers apart from new ones, because they answer different questions. A full diary of familiar faces can hide a lack of new ones.

A trace is useful evidence, but it isn't proof. Someone who tapped your ad may have heard about you from a friend first.

Researchers know this is hard. Randall Lewis and Justin Rao studied 25 large ad experiments, each with a group held back from the ads. They ran with big retailers and brokerages in one country. The return was hard to pin down, because each customer's spending jumped around so much. In at least half of them, the likely range of returns was more than 100 percentage points wide.

Their finding, the authors noted, was unlikely to apply to firms that get nearly all their customers from advertising. A business like that may see the link more clearly. For most advertisers, though, the authors expected the uncertainty to last.

Can an ads report prove the ads caused the sales?

Not on its own, because a report shows who saw an ad and later bought. It can't show who would have bought anyway.

Researchers tested this on Facebook itself. Brett Gordon, Robert Moakler and Florian Zettelmeyer analysed 663 advertising experiments on the platform. In each, some people were kept from seeing the ads, so the effect could be measured without bias. The team then tried two shortcut methods, which skip the held-back group and use thousands of details about each user.

Neither shortcut matched the experiments reliably. Their median estimates came out at about three times the experiments' median, or more. These were large campaigns in one country, from 2019 and 2020. The size of the gap won't transfer to your account. The lesson likely does, because the authors had richer data than most advertisers can use.

Chart: in Facebook experiments the median effect was 29, 18 and 5 percent. Two shortcuts estimated far more at every step.
The median effect of the ads in 663 Facebook experiments, against the median estimate from two methods that skip the held-back group. Source: Gordon, Moakler and Zettelmeyer, 2022 working paper (arXiv 2201.07055v2).
The numbers behind the chart
StepExperimentsShortcut 1Shortcut 2
Page views29%83%173%
Add to cart18%58%176%
Purchases, sign-ups5%24%64%

Median lift, in percent. Shortcut 1 is double machine learning, shortcut 2 is stratified propensity score matching. A revised version appeared in Marketing Science (2023).

The authors also suggest that a biased estimate won't always change the decision. So ask whether you'd decide differently if the true effect were far smaller than the estimate.

The strongest evidence of cause comes from a controlled test, where a group is held back from the ads. For a small business, even a test can leave a wide range. Ask whoever runs your ads whether one is worth running at your size, and what it would cost. When a test isn't practical, ask for a careful estimate with its limits spelled out. A good answer states its assumptions and points to something in your own records.

Ask for that estimate as a range. Anne Marthe van der Bles and colleagues ran five experiments with 5,780 people on how uncertainty is shown. Figures given as a range came across as less certain. In most of the experiments, they kept people's trust. A plain statement that a figure was uncertain, with no range, lowered it. A range tells you how unsure an estimate is. Those were public statistics, not ad reports, so treat this as a pointer.

How long should you wait before judging?

How long buyers take: an impulse buy the same day, a high-ticket purchase after research, an aircraft or jet over years.

In my experience, a few weeks in is usually too early to judge. How long depends on what you sell.

Different niches have different buying cycles. An impulse buy can happen the same day. A high-ticket purchase takes research, as buyers compare options and consult other people first. At the far end, selling an aircraft or a private jet can take years.

So ask how long your customers take to decide, and judge on that clock. Then match the dates to the question. Cash this month may come from ads that ran last month. Enquiries from this week may not have booked yet. Before the ads start, or at the next review, agree what you should see by when. Chats and enquiries come first, bookings later.

When should you change course?

I change course when the cost per lead runs higher than what's expected for that market.

Expectations vary by place and business. Our guide to Facebook and Instagram ad costs shows one way to set that figure. Agree it up front, or at the next review, so you have a line to judge against. A lead still sits high in the chain, though, so check what those leads became before acting.

Before you change anything, check what else moved. A price change, a full diary, slow replies on WhatsApp or a holiday can all shift results. Each shows up in the ads numbers without being the ads' fault. "The leads are poor" is a starting point, not a diagnosis. The cause could be the audience, the offer or the follow-up. The ads may still be the problem, and the point is to know which.

Ads also need room to experiment. I give campaigns that room, but we track and measure what happens.

Changing course isn't the same as stopping. If you're weighing a pause, read what happens when you stop advertising first.

Should good results get questions too?

Yes. My mentor says it's the questions he can't answer that worry him. Here is one of them: why did our cost per lead get cheaper?

Cheaper leads might come from a better ad. They might also come from a holiday, a competitor going quiet or a change in how results are counted. If you don't know why it improved, you can't count on keeping it.

Sriraj Aiyer and colleagues ran a study with 692 people who rated decisions in one medical scenario. The same decision was rated better when it turned out well. That held even in a small group who said outcomes shouldn't count. This was a made-up medical case, not an ad account. The pull is still worth knowing.

So judge the reasoning as well as the result. Was the decision sound, given what was known then? A good result can follow a weak decision, and the reverse. A sound decision doesn't excuse poor results that keep coming, though.

Result against reasoning: a good result from weak reasoning still needs the question of why it worked.

How do you review your ads without being an expert?

Take one real report, or the plan for one, and fill this in. Leave a gap wherever you don't know. Each gap is a question for whoever runs your ads.

QuestionWhat the report or your records showStill unknown
What were the ads meant to change, and by when?  
What was done since the last review: spend, placements, changes?  
How far down the chain did the numbers go: reach, chats or clicks, enquiries, bookings, payment?  
How many were new customers, and how many returning?  
What else changed: price, capacity, follow-up, season?  
What is the strongest conclusion, and one other explanation?  
What has nobody explained yet, good or bad?  
What happens next, who owns it, and when is the next review?  

Nothing here adds up to a score. The answers might say keep going, change something, wait or stop.

If whoever runs your ads can answer these plainly, you can hand the work over and still judge it. You don't need their skills. You need their answers.

Questions owners ask

How do you tell if a Facebook ad is performing well?

Follow its numbers past the click, to enquiries, bookings and payment. An ad with fewer clicks can still bring more paying customers. Judge it against the goal you set before it ran.

How long does it take for a Facebook ad to start working?

Ads can bring clicks within days, but sales follow the buyer's own decision time. A few weeks in is usually too early to judge sales. Agree what early signs to expect, and by when.

Why isn't my marketing working?

Check three things in turn. Was the agreed work done? What changed in the business, such as price, capacity or follow-up? Did the ads reach people who could buy? The cause can sit in the ads, outside them, or in a buying cycle longer than the wait.

What should a marketing report include?

For Facebook and Instagram ads, start with what was spent and what was done. Then the numbers down the chain, to bookings and payment. Then new against returning customers, what else changed and anything unexplained. A plain summary should say what the numbers prove, and what they don't.

How do you measure the success of a marketing campaign?

Against the goal set before it started. Count the business result it was meant to move, over a period that fits your buying cycle. Then ask how much of that change the campaign caused.

How do you measure marketing ROI?

Take the extra profit the ads brought, subtract what they cost, then divide by what they cost. Count profit after the cost of serving those customers. If an ads report shows a return figure, it usually divides tracked revenue by ad spend. That leaves out your other costs, and customers who would have come anyway.

What is marketing effectiveness?

It's how much your marketing moved the result it was meant to move, beyond what would have happened anyway. Reach and clicks show activity, but effectiveness shows up in the business.

Moses Azorbo, Founder of 5 Digital Marketing

About the Author

Moses Azorbo has run Facebook and Instagram ads for businesses hands-on since 2012. He founded 5 Digital Marketing, which publishes this article. The idea of the chain, the timings and the simplest check come from his client work. The review table is a prompt for questions, not a score. He'd rather a client judged him on what they earned than on what they spent. More at mosesazorbo.com.

Services How It Works Insights Book a Growth Diagnostic