--- title: "A four-week creative test plan for Meta ads, built from 18,920 ads" description: "How many ads to launch, which kinds to test first, and how to judge them at the end of the month without fooling yourself." section: Playbooks author: David Steel published: 2026-10-10T20:34:56.856Z url: https://one.sneeze.it/blog/a-four-week-creative-test-plan tags: ["benchmark-playbooks", "meta-ads", "creative-testing", "planning"] --- # A four-week creative test plan for Meta ads, built from 18,920 ads _How many ads to launch, which kinds to test first, and how to judge them at the end of the month without fooling yourself._ We have run over 18,920 Meta ads, and the earlier playbooks in this series each pulled out one finding: how often to launch, how little most ads spend, where the money goes, how to grade an ad, which looks and openings bring cheaper leads. This one puts them together into a plan you can run for the next four weeks. The plan is simple on purpose. Four new ads a week. Each ad tests one idea. Nothing gets switched off early. At the end of the month, every ad gets two grades, and the next month's ads come from what earned them. ## Why four ads a week Across 97 accounts with at least 50 leads, we looked at how many new ads each launched per week and how many winners it found per month. A winner is an ad that spent at least 10 times its own account's median ad and at least $500. In plain terms, it is an ad Meta chose to put real money behind. :::chart column "Winners found per month, by new ads launched per week" Under 1 | 0.37 1 to 2 | 0.43 2 to 4 | 0.68 4 or more | 1.00 source: Creative Benchmarks, benchmark.sneeze.it (median across 97 accounts with 50+ leads) ::: Accounts launching four or more new ads a week found about one winner a month. Accounts launching under one a week found 0.37. The typical account launched 1.6 a week, which puts it in the 0.43 group. Two honest notes. First, this shows that the two things go together, not that launching more ads causes more winners. Accounts that launch more may also differ in budget or in other ways. Second, look at the top number: about one winner a month. Four new ads a week does not mean four winners. It means you give Meta enough new options that it finds one worth backing. [How many new ads a week](/blog/how-many-new-ads-a-week) covers the cadence finding on its own. :::pullquote Four new ads a week does not mean four winners. It means enough new options that Meta finds one worth backing. ::: ## What to test first A new ad should test one idea, so that when it wins or loses you know why. We labeled 2,126 ads three ways: how the ad looks, how it opens, and how it was produced. The average cost per lead across all of them was $26.62. Some labels sat well below that: - **Look:** before and after at $18.73 (40 ads), product demo at $20.58 (69 ads), problem, then fix at $21.73 (115 ads). - **Opening:** contrarian take at $14.30 (41 ads), bold claim at $20.16 (99 ads), curiosity at $23.43 (57 ads). - **Production:** phone-shot UGC at $22.21 (64 ads). Every one of those groups is small, between 40 and 115 ads. That is why they belong in a test plan and not in a rulebook. They are the places where our data says a test is most worth running, not a promise that they will win in your account. At the other end, headline only ads averaged $34.35 per lead, 83% more than before and after, and announcement openings averaged $73.48 (another small group, 40 ads). If most of what you run today is one of those, testing away from it is the obvious first move. :::figure wide ![A planning desk with four columns of index cards, one column per week, each card a small sketch of a different kind of ad](/blog/media/d577960a7a1574d285a3ea98.jpg) One idea per ad, four ads per week. The cards change; the rule does not. ::: ## The four weeks :::steps 1. Before week one: write down your baseline Pull your account's cost per lead for the last 30 days and the spend of your middle ad (the median). Ten times that median, or $500 if that is higher, is your winner line for this test. 2. Week one: test the look Launch four new ads on the same offer. Make one a product demo, one a problem, then fix, one a before and after if you can make one honestly, and one in the style you already run as a control. 3. Week two: test the opening Take the look that seems most promising so far, but do not judge it yet. Make four versions that differ only in the first line or first few seconds: a bold claim, a curiosity line, a contrarian take, and your usual opening. 4. Week three: test the production Reshoot your strongest idea on a phone, by a real staff member, in your real location. Launch it next to a more polished version of the same idea, plus two fresh ideas from weeks one and two. 5. Week four: fill the gaps Launch four more. Use them to try what you have not yet tried, or a second version of anything that is pulling ahead. 6. Leave everything running Do not turn ads off in their first week because Meta spent little on them. In our data, 2 of 3 ads spent under $100 in their whole life; low early spend is normal, not a verdict. 7. End of week four: grade every ad twice For each of the 16 ads, ask two questions. Did Meta spend past your winner line? Did its cost per lead beat your baseline? Write down both answers. 8. Build next month from the grades Make new versions of the ads that passed both tests. Give ads that brought cheap leads but little spend a small budget of their own before calling them losers. ::: A worked example, purely hypothetical. Say you spend $1,500 a month and launch 16 ads. Spread evenly, that would be about $94 an ad, but it will not be spread evenly. Meta will push most of the money to a few ads, which is the point. In our data, 49% of all money went to the winners, and a typical account put 42% of its spend into its top 6% of ads. Expect the same shape in your test. Another hypothetical, for the winner line. Say your middle ad has spent $30. Ten times that is $300, which is under the $500 floor, so in your account any ad past $500 counts as a winner. If your middle ad has spent $80, ten times is $800, and that becomes the line. ## How to judge at the end The two grades exist because they often disagree. Of the winners in our data, 47 in 100 brought leads cheaper than their own account's average. The other 53 did not. So Meta's favorite ad is often not your cheapest source of leads, and your cheapest source of leads is often not Meta's favorite. Use the grades this way. An ad that passed both is the one to copy. An ad that Meta backed but that brings expensive leads is worth keeping only while your overall cost per lead holds; it is the first candidate to replace. An ad with cheap leads and little spend is the one most likely to be thrown away by mistake. Give it room. [Grade every ad twice](/blog/grade-every-ad-twice) goes into each case. A month is also short. Some ads will not have spent enough to judge. Say so in your notes and let them run into month two rather than forcing a call. :::callout tip "How to measure this in Ads Manager" At the ad level, set the date range to the life of the test and turn on columns for amount spent, your lead result and cost per lead. Sort by amount spent to see which ads Meta backed and where your winner line falls. Then compare each ad's cost per lead to your account's cost per lead over the same dates, not to last month's. Keep a simple sheet with one row per ad: the idea it tested, its spend, its cost per lead, and both grades. ::: :::takeaways "The short version" - Accounts launching four or more new ads a week found about 1.00 winner a month; under one a week found 0.37. The typical account launched 1.6. - One idea per ad: look in week one, opening in week two, production in week three. - Start with the lower cost labels, but treat them as tests. Every group behind them is small. - Leave new ads running. Two of three ads in our data spent under $100 in their life. - Grade every ad on spend and on cost per lead. Only 47 in 100 winners beat their account's average. ::: :::cta button="Get your free benchmark report" href="https://benchmark.sneeze.it" See where your account sits on launch pace and ad mix before you plan the month. ::: For the cost per lead picture behind every label in this plan, read [what 18,920 Meta ads say about cost per lead](/blog/what-18920-meta-ads-say-about-cost-per-lead).