Most explanations of Facebook ads describe the buttons. This one describes the machine underneath, because the reason accounts succeed or fail almost never has anything to do with which checkbox you ticked.
How do Facebook ads actually work?
Every time someone opens their feed, Meta runs an instant auction for the ad slots in front of them. You don’t win it by bidding the most. Meta ranks each competing advertiser roughly on bid × how likely that specific person is to take your action × ad quality.
Which means an advertiser bidding less than you can beat you, repeatedly, because the system expects more people to respond to their ad. Your creative isn’t decoration sitting on top of the media buy — it’s an input into what the media costs.
That single mechanic explains most of what confuses people about this platform. Why costs vary between advertisers selling identical things. Why a new ad can suddenly halve your CPL. Why “we increased the budget and it got worse.”
What decides who sees your ad?
Less of what you choose than you would expect, and more than you would like of what the system infers.
Targeting has quietly become the smaller lever. Meta’s delivery system is now generally better at finding responsive people than manual interest-stacking is. Narrow audiences often raise costs, because you’ve shrunk the pool the algorithm can search. Broad targeting with strong creative and clean conversion tracking usually beats a carefully hand-built audience of interests.
What still matters: the conversion event you optimise for, whether your tracking actually reports it back accurately, and the creative. Get the signal wrong and the system optimises confidently toward the wrong outcome — plenty of accounts are efficiently buying the cheapest possible version of a worthless action.
What is the learning phase, and why does it keep costing you money?
When an ad set is new or edited, Meta’s delivery system doesn’t yet know who responds. Performance during this window is unstable and usually expensive. Meta’s documented threshold for exiting it is roughly 50 conversions per ad set per week.
Here is the trap: editing an ad set restarts the learning phase. The instinct when results look bad on day three is to change something — budget, audience, creative. Each change resets the clock, so the account never leaves the most expensive part of its life cycle.
Two practical consequences:
- Don’t structure the account into so many ad sets that none of them can reach 50 conversions a week. Fewer, better-fed ad sets beat many starved ones.
- Give changes time to resolve before making more. Most accounts we inherit are not under-targeted; they’re over-tinkered.
What do Facebook ads cost in 2026?
Benchmarks from Digital Applied’s Q1 2026 analysis:
- Average CPC: $1.72 — up 11% year over year
- Average CPM: $11.54
- Average CTR: 1.49%
Cost per lead averages roughly $27.66 in the US, and the industry spread is wide (AdManage):
| Industry | Cost per lead |
|---|---|
| Restaurants & food | $3.16 |
| Real estate | $16.61 |
| Legal services | $18.17 |
| Home improvement | $41.26 |
| Dentists | $76.71 |
Notice that Facebook’s average CPC of $1.72 is roughly a third of Google’s $5.42 — but that’s cheaper attention, not cheaper customers. Google clicks come from people actively searching. Facebook clicks come from people who were looking at something else. The right comparison is cost per customer, not cost per click.
How do you tell if they’re working?
Not by clicks, which the old version of this article recommended and which we’d now call a mistake. Clicks tell you the creative caught someone’s eye. They say nothing about whether the business made money.
Track the actual outcome — a booked call, a qualified form, a purchase — and hold it against what that outcome is worth to you. A “good” cost per lead is any number below what a customer is worth, not whatever the industry average happens to be.
Two examples from accounts we run, to show the range. A DTC footwear brand hit 4.47x return on ad spend across twelve months on Meta — $111,374 in tracked revenue from 1,110 purchases. A post-frame building company generated 2,259 leads at $27.06 over the last twelve months, holding cost per lead between $23 and $31 while scaling spend 3.5x — which is the harder achievement, because efficiency usually decays as budget grows.
Different businesses, different metrics, same discipline: measure the thing that pays you.
Facebook or Google?
They solve different problems, and the honest answer is usually both.
Google captures existing demand. Someone types “emergency plumber” and you appear. Intent is already there; you’re competing for a decision that’s being made right now.
Facebook creates demand. Nobody opens the app looking for your product. You’re interrupting, which is harder, but it reaches people who’d never search for you — and it’s much better for visual products, offers, and retargeting the people your website already lost.
If your service is urgent and people search for it, start with Google. If your product needs showing rather than finding, start with Facebook. If you’re spending seriously, run both and let retargeting connect them.
Want the leads without running the account yourself? See how we handle it in our Meta Ads service, or schedule a call.