I agree with Jon Loomer's Master Brief. Here is the store version.

On July 2, Jon Loomer published the Master Brief, one document covering Loomer’s current approach to Meta ads, from targeting and structure to creative and tracking. The detail I liked most is how it was made. A tool Loomer built, with help from Claude Code, turns posts, videos and podcast transcripts into ad briefs, and the Master Brief merges 24 of them. Loomer reviewed every section and takes the blame if anything is wrong.

A document about trusting Meta’s algorithm, assembled by a tool and signed off by a human. I would like more of my own work to be organized that way.

I agree with it, and this is what it looks like when the ad account belongs to a store. Most of the brief’s examples come from lead generation, and a store has a catalog, margins, returns and a checkout. The ideas are Loomer’s. The store details are mine, and where I add a rule the brief does not have, I say so. The brief is a July snapshot, so check settings in your own account before acting on any of it.

Tell the algorithm the truth#

The brief starts from one idea: Meta’s delivery is literal. Ask for purchases and it looks for people likely to buy. Ask for cheap clicks and it finds cheap clicks, whether or not they ever order. So the purchase signal is the input that matters most, and in a store it comes down to a short checklist:

• The Purchase event carries the order value and currency and fires once per order. Decide whether VAT and shipping are in the value, then keep it that way.
• The browser event and the server event share a name and an event ID, the order ID being the obvious choice, so Meta counts one order once.
• Product IDs in your events match the IDs in your feed, or catalog ads cannot connect what someone viewed with what you sell.
• Reloading the confirmation page does not fire the purchase again. The brief counts staff opening confirmation pages among the ordinary reasons for results that look too good.

Keep the pixel and add the Conversions API, because the pixel alone is no longer reliable. Meta reports 17.8 percent lower cost per result on average for accounts that do, which is an average and not a promise. Shopify has an automatic setup, and elsewhere the brief points to the Conversions API Gateway. Server-side tracking is not a way around a visitor saying no, so what you send has to match your consent setup.

Then check against reality. Place a test order and confirm in Events Manager that exactly one Purchase arrived with the right value. Each week after that, compare purchases according to Meta with orders according to the store. They will not match, and the brief says to look for the boring explanation before suspecting Meta of inventing sales. Treat a purchase counted from a view as much weaker evidence than one counted from a click.

My addition: keep one number that no platform can attribute away, total revenue divided by total ad spend across all channels. It is blunt and a bit boring, which is why I trust it.

Stop describing your customer#

The brief’s boldest claim is that most targeting inputs are now suggestions. With Advantage+ Audience on, age, gender, detailed targeting, lookalikes and custom audiences guide Meta instead of fencing it in. What still binds is location, an age minimum and custom audience exclusions.

Stores are prone to this, because most have a persona document somewhere. It says the customer is a woman between 25 and 44 who cares about sustainability, so the ad set is restricted to that profile, results disappoint and nobody connects the two. Meta wants purchases too, so it was never going to spend much on people who do not buy. The restriction only removes options. And in retail the buyer is often not the user: gifts, parents, partners. The persona describes who uses the product. The Purchase event describes who bought it.

So target the countries you ship to, group the ones where advertising costs about the same, and split one out only when it brings enough volume of its own. Use the age minimum for age-restricted goods. That is the first of the two reasons the brief accepts for a restriction.

The second is a proven performance problem where you hold information Meta does not, and a store holds plenty. Meta sees the purchase. It does not see your margin, your return rate or whether the customer comes back. If your own orders show that one country, placement or age group brings purchases that get returned or leave no margin, the brief’s answer is a value rule: bid less there instead of switching it off. In Loomer’s lead-generation example, bidding 90 percent less on the oldest group cut its share of spend from 45 percent to under 2 percent while keeping it in the auction. If you cannot point to the numbers, skip the rule.

Fewer campaigns, with arithmetic#

The brief’s default is one campaign and one ad set, optimized for a conversion, with the budget in one place. Every extra ad set dilutes the money, adds auction overlap and makes it harder to collect the roughly 50 optimized events a week that Meta needs to leave the learning phase.

Put numbers on it. The brief suggests assuming, conservatively, that a sale costs about what the product sells for. Take a store with an average order of €60:

If your real cost per purchase is lower, which is the point of advertising, the numbers get kinder. The shape stays the same: only the last row gets even one ad set to the 50 a week the brief wants, and the budget that feeds one ad set is the budget five would starve. Five ad sets on €100 a day are not five experiments. They are five ad sets with a couple of purchases each and a lot of noise. My reading for a small budget is one campaign, one ad set, and judging by weeks instead of days.

A split earns its place in three cases the brief names: separate business goals, product lines with different messaging, and seasonal products, each only if it can be funded properly. That could be a sofa range and a candle range that share nothing but the logo. It is not one ad set per product because the catalog is long, one per audience, or one per creative test.

Creative is the targeting now#

With targeting loosened, the ad is what tells Meta who might want the product. The brief calls creative diversification the main lever: genuinely different visuals, formats, angles and text, so delivery can match the right message to the right person. Variety gives Andromeda, the retrieval step that narrows tens of millions of ads to a few thousand, something to choose from. Volume does not. Fifty near-identical ads are one idea wearing fifty hats.

For a store, look for differences in the angle (the problem it solves, a review, the comparison with the cheaper option, the offer), the objection (price, fit, delivery time, trust in a brand nobody has heard of), the buyer (for themselves, for a gift, buying again) and the format (vertical video for Reels and Stories, a 4:5 static elsewhere). Personas can share an ad set, and Meta matches them.

Say a store sells espresso machines. One ad is a vertical video about mornings with no time. One is a static with a review that mentions how quiet it is. One answers why it costs more than the cheap one. One is for the person buying a gift. Four ads, and each could win with someone different. Four photos of the machine on white with the headline changed are one ad.

The raw material is already in the store. Reviews, support tickets, return reasons and gift messages are customers describing what worried them, in their own words. Sorting them into themes is a job for a spreadsheet, a script or, with enough volume, a model like the one in my article on Jev. If you run catalog ads, the feed is your creative.

Build in phases, and read results for themes instead of a champion. The brief’s creative testing tool runs two to five ads inside the live ad set with an even split, and it warns that an underpowered test teaches nothing. Back to the arithmetic: five ads on €50 a day buy about six purchases a week at €60 each, roughly one per ad. That test cannot say which ad sells. Fewer ads, a bigger budget or a longer wait are the honest options.

Meta already does your remarketing#

The brief’s most surprising number is about remarketing. With no custom audiences in the setup, Meta tends to put roughly 20 to 25 percent of a sales campaign’s budget on people who already know you, sometimes 10 percent and sometimes 40. The Audience Segments breakdown shows it by splitting results into engaged audience, existing customers and new audience. It needs the Sales objective.

It tells the truth only if you define the segments properly, and a store has the data to do that:

• Engaged audience, defined broadly: everyone who visited the website in the last 180 days, plus your whole email list.
• Existing customers, defined precisely: everyone with a Purchase event, up to 730 days back, plus an email audience of every paying customer.

What goes wrong in a store is completeness. Customers who ordered by phone, in a physical shop or on a marketplace never trigger the pixel, so if you have their emails and your privacy terms allow using them for ads, the list is how they get counted.

The conclusion is that a separate remarketing campaign mostly reaches people twice, and a retargeting ad set with a high ROAS deserves a look at how much of it is view-through. Excluding past customers from acquisition needs a proven problem like anything else. For a mattress or a dishwasher, showing acquisition ads to last month’s buyer is a plausible waste, and the breakdown will show how much. For things people reorder, existing customers may be the point.

My addition: abandoned carts and browse sessions are a first-party job. An email or SMS flow to people who agreed to hear from you reaches them without an auction.

When it goes wrong, walk the funnel#

When results sag, the brief says to look at your own ads and setup before blaming targeting, structure or a broken algorithm, and to ask three questions in order. Each has a specific answer in a store.

1. Why are people not clicking? The creative and the offer. Is there an angle people stop for, and does the offer, including delivery time and returns, give a reason to act?
2. If they click, why do they not buy? Often the store, not the ad. Look at where sessions fall away: product page, cart, checkout. A shipping cost that first appears at checkout, a best seller out of stock, a missing payment method or a slow mobile page will each look like a Meta problem.
3. If they buy, why does Meta not know? Back to the first section: events, deduplication, consent choices and delays.

Do this before touching a campaign setting. Most of the fixes live in the store, the tracking or the creative.

Two more habits are worth stealing. Evaluate in aggregate: one ad taking half the budget is normal, and forcing money elsewhere usually does more harm than good. And read seven-day windows, not before the campaign is two weeks old. My addition for stores: mind the calendar. Payday, a sale, a free-shipping week, a stock-out and the weather all move results without Meta changing anything, and a campaign judged across a promotion measures the promotion. Most bad months begin as a bad week followed by a nervous human.

Ask what problem it solves#

If the brief has one sentence to keep, it is the question Loomer wants asked before any change: “What problem does this solve?” The audit is easy to copy. For every ad set, restriction, removed placement and disabled enhancement in your most important campaign, ask what problem it was meant to solve, whether it did, and whether a better answer existed.

For a store I would walk this list:

• Every extra campaign and ad set: a separate goal, a distinct product line with its own budget, or a seasonal range? If not, merge it.
• Every country split, age or gender restriction and removed placement: a legal requirement, a proven number, or a persona and a hunch?
• Every excluded audience: is there a proven problem, and is the audience complete?
• Every cost cap, ROAS goal or spending limit: which margin figure set it?
• Every disabled enhancement: if the worry is that the product looks different, look at what it actually produced.

Good reasons in a store are a legal requirement, a country you cannot ship to, a product you cannot supply and a number from your own orders. A feeling, an assumption about who buys and attachment to an ad are not.

Write the answers in a change log: what changed, which problem it solves, how you will know and when you will look. In two months it is the difference between knowing why the account looks the way it does and inheriting a puzzle from yourself. The brief is also not a license to rip out a structure that works. An ad set that solves a real problem stays.

A first month#

If I were setting this up for a store, I would go in this order.

1. Fix the signal first. A test order, one Purchase with value and currency, browser and server events matched, product IDs matching the feed, consent respected.
2. Define the two segments and look at the Audience Segments breakdown to see what Meta already does for you.
3. Collapse what has no reason to exist, with a written problem line for whatever stays separate. If the account works, go in steps: one change, then wait.
4. Launch a first set of genuinely different ads in both formats, sized so each can plausibly earn purchases.
5. Leave it alone for two weeks. Read seven-day windows, walk the funnel when something looks wrong, keep the change log, and build the second batch from the themes, not from the single ad on top.

Where this leaves me#

I agree with Loomer because the advice follows from how the system is built. A literal optimizer rewards the signal and the creative you give it, and punishes the controls you add out of worry. Loomer also says the honest answer is often that it depends, so treat all of this as defaults to test against your own numbers.

What I like best is that it moves the work to where a store has real advantages: its data, its catalog, its customers’ words and its checkout. None of those is a setting in Ads Manager. Shouting at the referee does not change the decision, and fiddling with Ads Manager at midnight does not change the auction.

Source: Jon Loomer, The Master Brief: My Complete Approach to Meta Ads, a free snapshot from July 2, 2026. The member version is updated monthly, so some details here may have moved on.