Three years of social media recruitment in figures

How many people do you reach with vacancies on Facebook and Instagram?

Most figures about social media recruiting come from surveys or from platforms that sell advertising. These come from our own ad account at Meta and from our own application forms, over exactly the same period: 10 August 2023 up to and including 31 August 2026. In those three years we ran 195 recruitment campaigns for 65 clients. Across the whole account, including our own campaigns for SocialFind, around €400,000 in campaign budget was spent. That's media budget only, without the work around it; campaigns that clients or other agencies run in their own account don't count.

Meta counted 28.7 million impressions of our ads in that period. An impression is one time an ad appears on a screen, so the same person can see it more than once. The estimated reach, the number of accounts that saw at least one ad, stands at 3.7 million. That's an estimate from Meta, not a count of people: someone with both a Facebook and an Instagram account can be counted twice, and the figure says nothing about where those people live or work.

For an employer, this mainly means reach is rarely the problem. Facebook and Instagram reach almost everyone you're looking for, including people who aren't actively job hunting and so never land on a vacancy site. The difference between a campaign that delivers applications and one that only racks up impressions lies in what happens after that impression.

Why we started going through our own campaigns

This research started while we were rebuilding our own website. Nikolay Broeks, who works on the site and the measurement at SocialFind, did not want to copy the usual claims about recruiting through social media without knowing whether they held true for us. The problem was that the answer was not in any single system. What an ad cost was in the ad account. What someone filled in was in the application form. What happened to that application afterwards was in the employer's applicant tracking system. Three systems that do not know each other.

He linked those three together, on an encrypted form of email address and phone number, and worked through three years of campaigns. The analyses themselves contain no names and no contact details. Every finding on this page was then recalculated by a second check from the same scripts, and the results that did not survive that check are not here.

What three years of campaigns taught us

Four things kept coming up, and they apply just as much to employers who advertise themselves as they do to us.

1. Applying has to fit how candidates actually behave. With us, 95 percent of submissions arrive on a phone, almost a third after six in the evening and well over a quarter at the weekend.
2. The advertising platform's number is not the same as the result. In 19 of the 38 campaigns we were able to compare, the cheapest lead put a different ad on top than the cheapest application.
3. Ads wear out, and more budget does not solve that. By week 3, an application costs one and a half times as much as in week 1 in half of the cases.
4. Cheap need not be the same as good. Across these 77 campaigns we found no demonstrable relationship between the cost per application and the chance that someone reaches the first conversation.

Below, each lesson sets out what exactly we measured, which campaigns it covers and what you can do with it.

How do people apply through our application forms?

On their phone. Of the submissions through our forms, 95% arrived on mobile, and 45.8% came straight from the Instagram app, meaning the browser built into that app. The median completion time is 32 s: half of applicants are done within half a minute. These figures come from our own application forms, and we advertise almost only on Facebook and Instagram. That audience is on a phone, which is why the share on mobile is this high with us; read it as behaviour within this channel, not as how everyone applies.

The timing is what surprises employers most. Of the submissions, 31.2% came in between 6 pm and midnight, and 26.2% came in at the weekend. Sunday, at 14.2%, is just as busy as a weekday. This too comes from our own application forms, and we advertise almost only on Facebook and Instagram. So it is about the times at which people fill in our form, not about how everyone applies.

What should you steer a recruitment campaign by: leads or applications?

Meta reports a number of leads for a campaign: the number of times Meta's tracking code registers a completed form. Anyone steering a campaign on the lowest cost per lead is assuming that number tracks real applications. Often, it doesn't. For each campaign, we compared the ads on cost per lead according to Meta and on cost per real application in our form. In half of the campaigns, 19 of the 38 with at least three ads and leads according to Meta, the cheapest lead pointed to a different winning ad than the cheapest application did.

And cheap isn't the same as good. Between a campaign's cost per application and the chance that an applicant reaches the first conversation, we found no relationship across 77 campaigns.

How quickly does a vacancy ad on Facebook and Instagram wear out?

An ad works best in the first few weeks. For each ad we compared the cost per application per week against week one. The median ad costs one and a half times as much per application in week 3 as in week 1, and almost twice as much from week 9 onwards. The number of clicks per impression drops as well. That last point is what we measure; the most obvious explanation is that the same people have already seen the ad, but we did not test that separately.

More budget does not fix that. Across 80 substantial budget increases, of at least one and a half times, the cost per application changed by a median of 1 percent, so next to nothing. A higher budget did not make an application cheaper in these campaigns.

Does a cheap application also produce a better candidate?

In our own figures it is not demonstrable, and that was the finding that surprised us most. Across 77 campaigns we found no demonstrable relationship between a campaign's cost per application and the chance that an applicant reaches the first phone conversation. That is not proof that no such relationship exists; it means we could not find one in these campaigns. The practical implication is clear enough: you cannot assume that a campaign in which applications come in cheaply also produces candidates who get further.

That does not mean price is irrelevant. It means that price on its own says nothing about quality, and that you therefore need a second measure alongside it. For us that is what happens to the candidate in the applicant tracking system.

What we do differently from simply running a campaign

Everything on this page comes out of one thing: we do not let the measurement stop at the advertising platform. For every application we try to tie four steps together.

1. The ad. The link in our application form carries the number of the ad someone clicked on. Since August 2024 that link succeeds for 90 to 98 percent of submissions per month; across the full history since July 2022 it is 74 percent of 25,662 submissions, because that information was not yet in the link before then.
2. The form. We count an application in the form itself, not through a measurement code on the page. That way it also counts in forms that do not carry such a code.
3. The applicant tracking system. We find 83 percent of submissions back as a placement in the employer's applicant tracking system, linked on an encrypted form of email address or phone number.
4. The first conversation. Only for the applications we found back in step 3, and only if they are far enough in the past (at least 90 days), can we see in that system whether someone reached the first phone conversation. That fourth step therefore does not cover every submission but the portion that made it through the three steps before it, and only where the employer keeps the status up to date in their system. It is the measure by which we compare ads and forms with each other, and every finding on this page states how many campaigns or forms it covers.

What that yields is not a prettier report but a different decision. An ad the platform calls expensive stays on if it produces candidates who get further. A form that attracts many submissions gets rebuilt if those submissions do not reach the conversation.

What it does not yield: a figure in which all the steps divide neatly into each other. The counts in steps 1 to 4 cover different groups and different periods, so we do not put a conversion rate between them. Anyone who does is flattering themselves.

What do these figures mean for an employer who wants to hire?

In short: there's plenty of reach, the difference lies in the steps after that. You can check four things as early as tomorrow. Open your own application form on a phone and fill it in with a timer running; if it takes longer than a minute, it's too long. Check whether your ads keep running in the evening and at weekends. Agree who calls the weekend inflow on Monday morning. And judge your campaigns on applications and conversations, not clicks. How we approach that within a campaign is covered under job marketing and social recruitment.

How do we measure and check these figures?

Every figure on this page has a fixed definition, a source and a period, and was checked before it went up here. Budget and impressions come from our own ad account, retrieved through Meta's official integration, and are mechanically reconciled each month against Meta's account total; those two match exactly. Reach is Meta's own estimate across the whole period; that figure cannot be added up and we take it as Meta gives it. The number of submissions through our forms was counted twice independently. We only publish the advertising figures for complete months, because the current month still changes at Meta. Note the two periods: budget, impressions and submissions run from 10 August 2023 to 31 August 2026, while the analyses of what happens after the application rest on a snapshot of 10 September 2026 and cover all 25,662 submissions in the same forms, including those from before August 2023. Until September 2026 this page also carried a count of how often the form page loaded after a click, measured with Meta's measurement code; a re-check showed that figure was not reliable enough and we therefore no longer publish it.

What we deliberately don't publish: cost per application or per hire in euros, sector averages, and client names. An average across three years, dozens of clients, and very different roles produces a number that's useless to any single employer. All definitions are in our measurement methodology.