Grade every ad twice: once for Meta's pick, once for yours

We have run over 18,920 Meta ads, and we graded every one of them. 1,175 became winners, which is 6.2%. A winner, in our grading, is an ad that spent at least 10 times its own account's median ad and at least $500. In plain words: a winner is an ad Meta chose to spend real money on.

That definition is honest about what it measures. It measures Meta's pick. It does not measure yours.

When we checked those winners against the thing an owner actually cares about, the cost of a lead, the result split almost down the middle. 47 in 100 winners brought leads cheaper than their own account's average. 53 did not.

Winners that brought leads cheaper than their own account's average

Source: Creative Benchmarks, benchmark.sneeze.it

So more than half the ads Meta poured money into were costing that account more per lead than its own average. That is not a scandal. It is how the delivery system works: Meta optimizes toward the result you asked for across the whole auction, and it will happily keep spending on an ad that gets results at a price you would not have chosen. The fix is not to fight Meta. The fix is to grade every ad twice.

The two grades

Grade one: Meta's pick. How much of the account's spend did this ad get? This is the spend share. An ad that took a big slice of the budget is one Meta likes. Spend share matters because money concentrates fast. Across our data, 49% of all money went to the winners, and 42% of a typical account's spend went to its top 6% of ads. Whatever Meta likes, it funds.

Grade two: your pick. What did each lead from this ad cost, compared with the account's own average for the same period? This is the only comparison that matters for you. Not a number from another business, not a number from a blog post. Your account, your average, the same dates.

Put the two grades side by side and every ad lands in one of four boxes:

High spend, cheaper than average. Meta's pick and your pick agree. Protect it and learn from it.
High spend, pricier than average. Meta's pick, not yours. This is where most of the quiet waste lives, because it is spending the most.
Low spend, cheaper than average. Your pick, not Meta's. Possibly a good ad that never got a fair run, or possibly luck on a few leads.
Low spend, pricier than average. Neither of you wants it. Let it go.

The fix is not to fight Meta. The fix is to grade every ad twice.

David Steel

A worked example

Say you spend $1,500 a month and get 50 leads. Your account average is $1,500 divided by 50, which is $30 a lead.

Now say one ad spent $600 of that and brought 15 leads. That ad took 40% of the budget ($600 of $1,500), so it is clearly Meta's pick. Its cost per lead is $600 divided by 15, which is $40. That is $10 more than your average.

Another ad spent $90 and brought 4 leads. Its cost per lead is $22.50, well under your average. But it only took 6% of the budget, so Meta barely tried it.

If you only looked at spend, you would call the first ad your best ad. If you only looked at cost per lead, you would call the second ad your best ad. Both readings are incomplete. The first ad is proven at scale but expensive. The second ad is cheap but unproven: 4 leads is a small number, and one more or one fewer lead would swing its cost a lot. The playbook below tells you what to do with each.

An illustrated two-by-two grid on a sheet of paper on a desk, with small ad cards placed in each square and one square shaded magenta — Two grades, four boxes. The box that costs you the most is high spend, pricier than average.

The playbook

Pull one month of ads with spend and leads
Use a date range long enough that most ads have had a chance to run, such as the last 30 days. You need three columns for every ad: amount spent, leads, and cost per lead.
Work out your account average
Add up total spend and total leads for the same dates, then divide spend by leads. Do not average the ads' cost-per-lead figures together; small ads would count as much as big ones.
Give every ad its first grade
Divide each ad's spend by total spend to get its share. Mark the ads that took a large share as Meta's picks.
Give every ad its second grade
Mark each ad as cheaper or pricier than your account average. Ignore any ad with only a handful of leads for now; note it as unproven rather than good or bad.
Act on the high-spend, pricier box first
These ads cost you the most. Try lowering their budget, moving them into a separate ad set with a cap, or replacing them with a fresh version of whatever made your cheaper ads work. Watch the account average for a week after, not just the ad.
Give your low-spend, cheap ads a fair run
Relaunch them in their own ad set, or duplicate them as new ads, so Meta has a fresh chance to deliver them. If they hold their cost as spend grows, you have found an ad that both of you can agree on.
Write down what the agreeing ads have in common
The high-spend, cheaper ads are your best teachers. Note the look, the opening and the offer, and brief your next round of new ads from them.

Two traps to avoid

Do not kill Meta's pick on reflex. An ad that is pricier than average can still be carrying the account. If you pause it and nothing else picks up the volume, your total leads drop and your average may not improve. Cut its budget gradually or test a replacement beside it, and judge the account, not the single ad.

Do not crown your pick too early. A cheap ad on a few leads is a hypothesis. Most ads never get the chance to prove much: in our data, 2 of 3 ads spent under $100 in their whole life. That is exactly why the low-spend, cheap box deserves a second run rather than a trophy. For more on judging young ads without killing winners early, read most ads barely run.

If you want the bigger picture on how spend piles up behind a few ads, and when to refresh them, where the money goes covers concentration and fatigue. And for how we built the grading in the first place, see what 18,920 Meta ads say about cost per lead.