Facebook Ads Statistics That Actually Change What You Do (Benchmarks, Creative Rules, and Setup)

Most Facebook ads statistics are useless to you

Not because they are wrong. Because nobody tells you what to do with them, and nobody tells you when the number was collected.

A stat is only worth reading if it changes one of three things: a decision you make, a number you compare yourself to, or a setting you leave alone. So that is how I sort them.

• Actionable — change your campaign because of this.
• Plain stats — compare your own numbers to these, carefully.
• Automated — set it up right, then let the algorithm work.

Below is the full list, with what I would actually do with each one, and where I would not trust the number at all.

1. Actionable stats — use these when building or fixing a campaign

Captions on video ads

Tests conducted by Facebook found that adding captions to a video ad boosted view time by an average of 12%.

Twelve percent more watch time for a task that takes a few minutes. There is no tradeoff here. Caption everything.

This pairs with the next one, which is the reason captions matter so much.

85% of Facebook videos are consumed without audio

Blue Corona found that 85% of Facebook videos are consumed without audio.

Read that as: your voiceover is decoration. If the offer, the proof, and the call to action only exist in the audio track, 85% of your viewers never got the message.

Practical rule I use — write the ad so it works on mute first, then add sound as a bonus. Key text on screen. Product visible in the first two seconds. Captions burned in, not relying on the platform toggle.

This one I trust even though it is old, because the reason behind it has not changed. People scroll in public. Sound-off is a behavior, not a platform setting.

Vertical beats square and landscape more often than not

In 70% of tests, vertical ads have an incremental boost in brand lift and a 3 to 9 point increase in ad recall.

70% is not 100%, so this is a default, not a law. Start vertical. If you have budget to test, test landscape against it — but do not make landscape the starting point when the feed and the Reels surface are both vertical.

Image sizes Facebook actually recommends

Ad type	Recommended image size
Single image ad	1200 x 628 pixels
Carousel ad	1080 x 1080 pixels

Boring, but this is where a lot of small accounts leak performance. Uploading one asset size for every placement means something gets cropped, and the crop usually eats the headline text you spent an hour writing.

Build the two sizes above at minimum. Check the placement preview before you publish. Every time. And check the spec page yourself before a launch — this is one of the few numbers on this page that Meta updates, and every roundup on the internet is quoting a snapshot of it.

Five power words

Five power words to get a higher response from your ads: You, Free, Because, Instantly, New.

I do not treat these as magic. I treat them as a checklist for a first draft that reads too corporate. "You" forces second person. "Because" forces a reason, and a reason is what makes an offer believable. "Instantly" and "New" force specificity about timing. "Free" only works if the free thing is real.

If your headline has none of the five, it is usually because you wrote about your company instead of the reader.

2. Plain stats — benchmarks, and why I trust them the least

Read this section differently from the other two. I am going to quote the numbers everyone quotes, then tell you honestly what is wrong with them.

Here is the problem. The published Facebook benchmark figures circulating right now come from a small handful of large studies. Those studies get cited, then the citation gets cited, then a roundup cites the roundup. By the third hop nobody carries the collection date forward. Meanwhile the auction, the placements, the attribution window defaults, and the signal loss underneath all moved. A cost-per-click average from a pre-iOS-14 dataset is not describing the account you are logged into today.

So before you compare yourself to any number below — including mine — do this:

• Find the original study, not the roundup. If the roundup will not name the source, ignore the number.
• Find the collection window. If there is no date on the page and no date in the source, treat the figure as a shape, not a target.
• Check whether the sample is your industry, your country, and your objective. All-industry averages blend lead gen with ecommerce with app installs. Those are three different businesses.

With that said, here are the numbers, and the part of each one I think survives.

Click-through rate

• One Wordstream study found that across all industries, the average click-through rate on Facebook ads is 0.90%.
• Legal has the highest average click-through rate out of any industry at 1.61%.
• Employment & Job Training has the lowest click-through rate at 0.47%.

The spread matters more than the average, and the spread is the part that ages well. Legal runs more than three times the CTR of employment. Absolute costs drift with the auction. Relative differences between categories drift much slower, because they come from how much the audience wants the thing, not from what advertisers are bidding this quarter.

So use the ratio, not the raw figure. If you are in a low-CTR category and you benchmark yourself against an all-industry 0.90%, you will kill campaigns that were fine.

Conversion rate

The same study found that the average conversion rate on Facebook ads is 9.21%.

Worth sitting with, mostly for the shape of it. A sub-1% click rate and a 9% conversion rate on those clicks means the click was the hard part, not the landing page. That diagnostic still holds even if the exact figure has drifted a long way.

So use it as a decision rule, not a score. If your landing page converts under your own historical average and your CTR is fine, fix the page. If your CTR dropped and your page still converts, fix the creative. Do not fix both at once or you will not know what worked.

Cost

Fit Small Business found that the average cost-per-click on Facebook is $1.86 and the average cost-per-thousand-views is $11.20.

I would not plan a budget off that CPC. Not because it was wrong when published — because cost figures are the fastest-decaying stat on this entire page, and this one has been re-quoted for years without a date attached.

What I would do is run the arithmetic pattern. At $1.86 per click and 9.21% conversion, a lead costs roughly twenty dollars in raw click cost. Now swap in your own CPC and your own form conversion rate and you get a number that is actually yours. If your product cannot support that lead cost — before creative, before management, before the leads that never answer the phone — the math problem is not the ad account.

Build your own benchmark instead

This is the part the stat roundups cannot give you, because it is specific to your account. It takes about twenty minutes and it beats every number above.

1. Pull the last 90 days in Ads Manager. Break it out by campaign objective, not by everything at once.
2. Record CTR, CPC, CPM, cost per lead, and lead-to-appointment rate. Write down the median, not the average — one runaway winner distorts an average badly.
3. Do it again by month so you can see seasonality in your own category.
4. Note the floor and the ceiling. Your working range is the number you judge new campaigns against.
5. Redo it quarterly. Auction costs move, and your own numbers from nine months ago are already a historical document.

Benchmarks are for deciding whether to keep testing. They are not proof that your campaign is good or bad. Your own trailing 90 days beats any industry average published by anyone, including me.

3. Automated — set it up correctly, then get out of the way

This group is different. There is nothing to optimize by hand. The job is making sure the campaign is structured so the algorithm can do what it already does well.

Mobile is where the impressions are

• Facebook users are 1.5x more likely to watch video on their mobile devices than on desktop.
• In one experiment, mobile ad placements outperformed desktop by over 45%.

The takeaway is not "exclude desktop." The takeaway is that manual placement carve-outs usually make things worse. Leave placements open, make the creative mobile-first, and let delivery find the cheap impressions. Then check your landing page on a phone, because a mobile advantage disappears fast if the form is unusable with a thumb.

The call-to-action tradeoff

• One test found that changing the CTA from "Sign Up" to "Learn More" improved click-through rate by 22.5%.
• The same test found that "Sign Up" resulted in a 14.5% higher conversion rate.

This is the cleanest example of a real tradeoff in the whole list. "Learn More" buys you traffic. "Sign Up" buys you intent. Note that it is a single test, so treat the direction as the finding and the percentages as approximate.

How I think about which to pick:

• Cold audience, unfamiliar offer, need volume for the algorithm to learn — "Learn More." The extra clicks feed the learning phase faster.
• Warm audience, retargeting, offer already understood, sales team paying per lead — "Sign Up." You want the conversion lift and you do not want to pay for tire kickers.

There is no universally correct button. There is only which cost you would rather pay.

Ad copy length

After analyzing 752,626 Facebook ads, AdEspresso found that the median headline length is 5 words, and average ad text contains 14 words.

Five and fourteen. That is the shape of a typical ad on this platform. Big sample, so the central tendency is probably still roughly right, but remember it describes what advertisers wrote, not what performed. It does not mean long copy never works — it means long copy is the exception and you should know why you are writing it. If your first draft headline is eleven words, cut it in half and see if anything of value was lost. Usually not.

The part no stat will tell you: the setup underneath

You can nail every number above and still get nothing out of a campaign, because the plumbing between the ad and your CRM is broken. I spend more time on this than on creative. These are failures I have had to fix on live builds.

Lead data gets mangled in transit

Facebook lead-ad payloads relayed through Zapier into Google Sheets commonly split the lead name, so only the first name maps through. You end up with a sheet full of half-names and a sales team that cannot address anyone properly. The fix is normalizing first and last into a single full-name field at the webhook layer, before it hits the sheet. Small thing. Costs you nothing. Nobody checks it until a client asks why every row says "Maria."

Permissions break lead flow with no warning

On one build, Meta app permissions blocked direct lead ingestion from a Facebook lead form entirely. Rather than pause spend and wait on app review, we deployed a Zapier webhook fallback to forward leads into the app. Lead flow stayed alive. That is the point — always have a second path for leads, because platform review timelines are not yours to control.

Same category of problem when running more than one page. Supporting multiple Facebook pages under a single lead-ingestion app required a dedicated page-access-token mapping table in the database instead of one stored token. And a schema cache error had to be cleared before the new table was even recognized by the app. If you run ads for several locations or brands under one system, plan for that from the start — retrofitting it while spend is live is not fun.

Launch blockers are rarely the ad itself

One webinar registration funnel needed four separate fixes before it was launch-ready:

1. The registration redirect target
2. The thank-you page contact email
3. A generated privacy policy page — required for Facebook ad approval
4. Ad copy plus targeting

Only one of those four is the thing people think of as "the ad." The privacy policy one gets accounts rejected constantly. Handle it before you build creative, not after your ad gets disapproved on a Friday.

A short order of operations

1. Get the tracking and lead routing working end to end. Submit a test lead yourself and confirm it lands in the CRM with a full name, correct email, and correct source.
2. Confirm compliance basics — privacy policy page live, pages and permissions connected, fallback path for lead delivery.
3. Build creative mobile-first, vertical, captioned, at the current recommended sizes.
4. Write to five-word headlines and roughly fourteen-word body copy. Check for the power words.
5. Leave placements open. Pick your CTA based on whether you need volume or intent.
6. Pull your own 90-day medians and judge new campaigns against those, not against an industry average from a study with no date on it.

None of this is clever. That is the point. Most accounts I look at are not losing to a better strategist — they are losing to a cropped image, a broken redirect, and a benchmark they were never supposed to be measured against.

If you want a second set of eyes, I will run a test lead through your setup, pull your own 90-day medians, and tell you which of the three — tracking, creative, or offer math — is costing you money. About half an hour, and you keep the numbers either way. Book a time here if that is useful. If the answer is "your account is fine," I will say that too.