Judge each ad against your target cost, then cut it or scale it

1. Why the ad itself does the work

The view here is that Facebook and Instagram have grown skilled at locating buyers within almost any audience you hand them. In 2020 the game was hunting for the perfect audience: interest types, maybe even detailed lookalike setups. The money is not there anymore, in this view. It comes down to the ad: your creative is what tells the platform whom to go and find.

That turns paid ads into a volume game. Produce a lot of different ads, test every one, and back with money only the few that prove themselves. The starting point is usually a single piece of content or a single format: a long video is tested for free as organic content, the formats and angles that land there become paid ads, and those are scaled. The volume comes from a system, not from hiring more people to pile onto it.

On interest targeting the team's answer is short: it never uses it. Broad targeting is its default. Audiences of people who have engaged with the business, by viewing its social profiles or its website, still perform well for it, and so do lookalikes made from a list such as a thousand past buyers. The reasoning: outside the creative, what decides the result is the data you send back to the platform.

Flag: targeting, or the ad itself? This section lets the ad tell the platform whom to find, with broad targeting as the default. SOP 36, section 1, holds that a paid-advertising problem is a targeting problem, and SOP 213, section 5, starts a new channel on interest groups; both carry the same flag. This page does not settle which reading is right.

Testing on organic content before paying is SOP 38 — Turn the best organic post into an ad.

2. Keep testing apart from scaling

Run test campaigns apart from scaling campaigns. While testing, set the budget for each test group yourself, so the platform is not free to spend everything on a single ad before any lesson is in. Once an ad proves itself, move it into a campaign built to scale. How to set up either inside an ad platform is not covered on this page.

A second view. How the account is set up counts for a lot less than you think, in the view of someone who ran ads at an agency. What strong accounts share is simplicity at the top level, with the ads underneath split into very specific messages.

3. Test one angle or one format per group

An angle is a distinct way of addressing the customer. A format is a different way of saying the same thing to them. Give each its own test group: in one example, the founder answering questions from a stage sits in one group and customer testimonials sit in another. A new angle opens a new pocket of buyers: the same person, reached through a different door.

How many ads per group. Ideally one or two. The team's own groups hold no more than four, and a fifth is starting to push it. Inside a group, the platform decides from early indicators, such as the cost of showing the ad and the click-through rate, where the money goes, whether it is right or wrong. Put in too many and the money is split so finely that no single ad gets enough to show what it can do; the more control you keep over where it goes, the better.

Where a group does hold more than two, they are close variations of one ad: the same ad with two or three different opening moments or cuts, there so a winner shows sooner. The team can do that because it has so many variations of creative to test. Taking money out of a scaling campaign to test, it would still start with maybe one, two at most, per group, and run several groups side by side, unless there is a large pile of creative to push through at once.

4. Point the platform at the furthest step you can track

Most advertisers start by optimizing for an opt-in, or maybe for a purchase if they sell goods online, which is fine. For its own account, the team pushed further down the funnel and optimizes for booked meetings. That tells the platform to find people more likely to book, not just to opt in. In its view this is probably one of its biggest levers, and it made a large difference to the quality of its leads. Optimizing a campaign for return rather than for leads is SOP 213, section 8.

5. Set the test budget, and the share kept for new ads

The test budget. Be prepared to spend as much as a customer is worth to you to learn whether something works: that is, to break even on what it costs to acquire a customer. Put plainly: if you will not at least break even to see whether something works, advertising is probably not for you yet. This applies above all to an advertiser who has not yet run anything on the platform.

Flag: a second test budget. A second reading lets each new ad spend up to twice what a new customer brings in over their first month while it brings leads, and cuts it once it has spent that figure if it brings none; it is set out on SOP 56, section 5. This page does not settle which reading is right.

The share kept for new ads. Most recently the split here is 70 and 30, with 70 percent of the budget behind what already performs well, and 30 percent on fresh versions of the winners or on concepts not tried before. The split changes over time. It is not always an even spread across the creative. The same split applies when a scaling campaign is already running: set some of its money aside and spread it across new test groups. Money for a test might also come from something that is not performing: a weak ad is a good excuse to kill it and move the budget.

Where the split starts. If you are starting from nothing, everything is testing. How fast the scaling side grows depends on how many winners testing produces; the scaling campaigns are grouped by angle and format, the same way the test groups are.

A learning budget that might be 5 or 10 percent, for ideas you expect to fail, is SOP 42, section 6. The 30 percent here also funds variations of ads that already win, so the figures measure partly different things. Splitting hook effort 70, 20 and 10 is SOP 122.

6. Read the early signs in the first days

Before there are enough sales to give a clean read, look at simple signals: the click-through rate, and the cost of showing the ad. For that team, a click-through rate of 1.5 percent or more is a good first sign, and the cost of showing the ad may be low; that figure is for its own account, and it is not a benchmark. Judge both against your own past averages. Sometimes those who have been running an account have a feel for what is getting traction, and the signals differ a little from one business to the next.

Early numbers that stay dreadful for a couple of days are a pretty good reason to cut the ad and send its budget elsewhere.

7. Manage to the cost of a qualified lead

The number to manage to is the cost per qualified lead: a lead your sales team would think worth calling. Set a target for it. The rules below measure against different targets.

The cut rule. The team calls this its only cutting rule. Once an ad is starting to get traction, if each qualified lead from it runs to about three times your target, the ad is finished: switch it off.

The wait rule. If every early sign looks good but no sale has come in yet, the team usually lets the ad run to about three times its target cost of acquiring a customer before it decides to cut, because its sales cycle is a little longer.

Rule	Measured against	Applies when	In an invented example
Cut	Target cost per qualified lead	The ad is getting traction	Target $100 a qualified lead; cut at about $300
Wait	Target cost to acquire a customer	Early signs are good, no sale yet	Target $1,000 a customer; let it run to about $3,000

The worked figures on this page are invented.

A second view: switch off what closes worse. More reasons to turn an ad off come from the agency side. An ad with a very high click-through rate and a cheap cost per lead whose leads do not qualify should be shut off and never run again; reframe its message instead. And an ad that still brings leads or calls, but whose leads buy less often than others', should go too: keeping it, in that view, tells the platform that an average account is fine. Do not be afraid to shut off most of your ads. A higher cost per lead at times is acceptable as long as you are fully sure that the message a person sees, from the ad through the page, is as accurate as it can be for your business. Higher-cost leads that are worth more are SOP 213, section 6.

8. Read the numbers weekly with two questions

Have the report sent to you each week. Then put two questions to them: does the cost look normal? Does the volume? If either one is off, that tells you exactly where to dig, and you do it before you even open the ad account. Putting the lead count first in a weekly marketing meeting is SOP 220, section 2.

9. Give the winners more money

When an ad wins, there are two moves: give it more budget where it sits, or copy it into the scaling campaign, which keeps its likes and comments. Sometimes the team simply lets it scale in place.

Raise spend in modest steps. Leads do not always grow in step with budget: putting 50 percent more behind an ad need not bring 50 percent more leads. The size of a safe daily step is an ad-platform setting and is not covered on this page.

10. Read results by the type of customer

Here is the most important part, and the one people miss, in this view. Do not stop at which ads brought in qualified leads. Look at which types of customer actually turn into sales, and optimize toward the lifetime value of those customers. The team sorts results by industry; for you, it might be one ideal customer profile. Where one type closes well for little spend, there is more money there: make more ads for that type.

Work with your sales team to see where the good leads came from, and what is said on those calls, then do more of it. Within each type, look at which ads perform, without digging through the ad account again. Finding your best customers and the places that bring them is SOP 219, sections 2 and 4.

A second view: go back to the runners-up. This is one of the biggest tips from the agency side. When five ads launch and one stands out, the second and third are often never looked at again. Go back through past spend to the ads that came second and third, and rework them with what you know now. Reworked runners-up usually stand a better chance against the variations of your present ads. Losing ads, in that view, hold some of the most useful data there is.

11. Cut fast, and keep feeding the system

Switch off losing ads quickly and move their money to winners. Simple as that is, a lot of people are too slow at it. Since cutting is fast, new ads have to keep coming in so that winners keep being found. The whole approach holds up only if creative is made in large amounts, enough to keep it supplied.

How many ads to make for a given spend is SOP 210, section 1. Making new versions of a proven ad is SOP 113 — Remix and remake a proven ad, whose section 8 left the measure of a winner and the test budget open; this page gives one team's answer to both.

Flag: more ads, or a better page? This section treats a steady supply of new creative as the lever. A second view, from an ads practitioner, holds that trying 20 versions of the page tells you more about your ads than a thousand new creatives would. The same flag sits on SOP 11, section 1. This page does not settle which reading is right.

12. What this page does not decide for you

• How many new ads to make each week. The figures that appear do not line up. One benchmark is 10 new ads a week for each $10,000 of monthly spend; another sets at least 10 to 15 creatives against $10,000 a month, with no period stated; a third is 20 pieces to rotate at $1,500 a month. This page gives no figure for how many to make; SOP 210, section 1, gives benchmarks.
• How long a test runs. Beyond a couple of days of very poor early signs, and the wait to about three times what a customer is meant to cost, this page gives no figure for how long a test runs.
• What counts as a normal cost or volume. The weekly read compares against normal; what normal is for you is not established on this page.
• How to split money between testing and scaling once winners pile up. It depends on how many winners testing produces. Beyond the 70 and 30: not established on this page.

13. The checklist

Step	What to do
1	Keep test campaigns, where you set each group's budget, apart from campaigns built to scale
2	Put each angle or format in a test group of its own; ideally it holds one ad or two, and no more than four
3	Point the platform at the furthest step you can track, such as a call that gets booked
4	To learn whether an ad works, be prepared to spend as much as a customer is worth to you
5	Keep about 70 percent behind what performs and 30 percent on new variations and concepts
6	In the first days, read the click-through rate and the cost of showing the ad against your own averages; cut if they stay dreadful for a couple of days
7	Set a target cost per qualified lead; cut an ad with traction at about three times it
8	If the first signs look good and no sale is in yet, wait until it has spent about three times what a customer should cost
9	Weekly, get the report delivered and check that cost and volume both look normal
10	Give winners more budget in modest steps, or copy them into the scaling campaign
11	Sort results by type of customer, make more ads for the type that buys best, and ask sales where good leads came from
12	Keep new creative coming, so that cutting fast does not leave you short of ads to test

14. What this page does not cover

Ad-platform settings are not covered on this page: campaign budget types, separate ad accounts for each funnel, the size of a daily budget increase, tracking and event setup, and reporting software. Choosing the first platform is SOP 35, and making the creative itself is SOP 212 — Make better ad creative with hooks, permutations and assembly.