Meta and TikTok for DTC brands
Paid social is a beton the second order
At median 2026 economics, a Meta customer does not pay for themselves on the first purchase. Nobody's does. The brands that scale profitably are not the ones with better targeting, they are the ones who get a second order. This is the acquisition channel inside our ecommerce practice.
★★★★★ 4.9 across 312 client reviews
Start with the arithmetic
Three published numbersfor the same customer
Ask what a Meta customer costs in 2026 and you will get three answers, all of them from credible sources, differing by a factor of six.
None of these sources is lying. They are counting different things. The platform number counts every purchase it can claim inside its attribution window, including from people who already buy from you. The operator number divides total marketing cost by net-new customers. The channel number goes further and asks how many of those purchases would have happened anyway.
That last question is the expensive one. Across 253 media-mix models covering $383M of spend, platform-reported return on ad spend ran two to five times above true incremental impact. So the number in the interface is not a slightly optimistic version of the truth, it is a different measurement entirely.
Now put any of them next to what an order actually contributes.
On a $95 order at median DTC margins, after cost of goods, shipping, fees and a one-in-five return rate, there is roughly $29 of contribution. Against a $39 platform CPA, the first order is already underwater. Against a real acquisition cost, it is not close.
This is not an argument against paid social. It is the reason paid social exists as a discipline. You are not buying an order, you are buying a customer, and whether that trade works depends entirely on what happens after the first purchase.
Which is where the second number that matters comes in. The average DTC brand gets a second order from about one in four customers over twelve months, and one 2026 study of 156,110 customers put the aggregate at 18.8 percent. Half of the second orders that do happen land within 30 days. Three quarters land within 90. And 77 percent of them are a reorder of the same product rather than a cross-sell.
So the window is short, the behaviour is predictable, and most brands spend it silent. That is the actual constraint on how far a Meta account can scale, and no amount of campaign restructuring fixes it. Email and SMS do.
Free, no email required
How many ordersuntil you get your money back
Take last month's paid social spend and the purchase count the platform reported. This converts it into a real acquisition cost, then works out how many orders a customer has to place before the trade turns positive, and what repeat rate you would need to get there.
Last 30 days
Meta plus TikTok plus any other paid social, combined.
The third field is the one people guess at, and it is the one that decides everything. It folds together two things: repeat buyers the platform counted as conversions, and purchases that would have happened without the ad. A post-purchase survey and a geo holdout are how you replace the guess with a number. Published DTC repeat rates run 25 to 30% over twelve months, higher for consumables and under 18% for durables. Contribution per order comes from the map on the hub.
What the trade actually is
Updates as you type. Nothing is sent anywhere.
This is a twelve-month view on contribution, not a lifetime value model, and it deliberately ignores discount rate and cohort decay. It is meant to answer one question honestly: at your current numbers, does the second order arrive before the money runs out. For the ceiling on what any single order can be worth, use the max CPL calculator.
How we build the account
Three layers,and only one of them tests
Meta's automated campaigns now take about 62 percent of ecommerce conversion spend on the platform, and they beat manual structures on cost per acquisition by roughly 22 percent, more when they are fed first-party purchase and email signals. That changes what an account is for. It is no longer a targeting exercise. It is a creative pipeline with a budget attached.
-
Layer 01
The scaling campaign
One consolidated automated shopping campaign carrying the majority of spend. Broad, no audience surgery, no daily budget fiddling. It is fed rather than managed.
- Server-side purchase events with real values
- Customer list and email engagement signals connected
- Existing-customer budget cap set deliberately, not left at default
- Budget changes in steps of 20 to 30%, then left alone for 72 hours
Typically 65-75% of spend
-
Layer 02
The testing campaign
The only place anything is deliberately experimental. New angles enter here, prove they can hold cost per acquisition at a fraction of scaling budget, and graduate. Losers get switched off rather than optimised.
- One ad set, one new concept per ad, no audience variables
- Enough budget for a decision inside 5 to 7 days
- Judged on cost per acquisition and hold rate, never on engagement
- Graduates get rebuilt into the scaling campaign, not moved
Typically 15-25% of spend
-
Layer 03
The retention exclusion
The layer most accounts do not have. Recent purchasers and engaged subscribers are handled by email and SMS, not by paying Meta to reach them again. What remains in paid retargeting is scoped tightly and measured against a holdout.
- Purchasers in the last 30 days excluded from prospecting
- Cart and checkout abandoners handed to flows first
- Whatever paid retargeting survives runs on a small, fixed budget
- A holdout group kept back so incrementality stays measurable
Typically 5-15% of spend
The percentages are starting positions, not doctrine. What does not move is the principle: one campaign scales, one campaign learns, and nothing else exists. Accounts with fourteen live campaigns are not sophisticated, they are splitting conversion data fourteen ways and starving the algorithm that was supposed to do the work.
Where the angles come from
Read the ad librarybefore writing anything
Every advertiser in your category is publicly listed in the Meta Ad Library, with every live creative and the date it started running. Most operators have never opened it. It is the cheapest research in the business and it produces a testing matrix rather than a mood board.
We are not looking at production quality. We are looking for which claims a competitor has kept paying to make. A creative that has been live for four months has survived a decision to keep funding it, and that is the closest thing to public performance data you will ever get.
| What we log | What it tells you |
|---|---|
| Run length | The single most useful field. Sort every competitor's creatives by start date and the survivors at the top are the angles that work in your category. Anything under two weeks is a test, not a result. |
| Angle, not format | Classify each survivor by the claim it makes: price, speed, ingredient, founder story, problem demo, social proof, guarantee. You end up with a short list of angles the category has already validated and a shorter list nobody is running. |
| Variant count | How many live creatives a competitor is carrying at once tells you the production cadence you are being asked to match. If they are running 40 and you are running 4, targeting is not your problem. |
| Destination | Where the ad points. A brand sending paid traffic to a purpose-built landing page rather than a product page has usually learned something expensive about their product page. |
| The offer | Threshold, bundle, first-order discount, subscription incentive. An offer that has run unchanged for months has beaten the alternatives, and it sets the number your own offer has to survive against. |
Out of that comes a matrix of angles against formats, which becomes the testing campaign's queue. Creative fatigue in fast-moving categories sets in within a week or two, so the queue is the deliverable, not any individual ad. A brand that cannot produce continuously cannot scale paid social, and that is a resourcing decision worth making with open eyes before the budget goes up.
You've got questions
We've got
straight answers
What ROAS should we be targeting?
Whatever your margin says, which is a different number for every brand and no number at all for a brand that has not worked out its contribution per order. The median Meta ROAS across a year of tracked ecommerce spend was 1.88x. On typical DTC margins, break-even sits above 3x. Both of those facts are true at once, which tells you how many accounts are running at a loss with a healthy-looking dashboard.
Work out your own break-even first. The contribution map does it in a minute, and the full working is here.
Meta says we made $80,000 and Shopify says $50,000. Who is right?
Both, under their own model. Meta claims purchases inside its attribution window including view-throughs, and it does not know about the ones it did not influence. Shopify counts orders and does not care what happened first.
The practical move is to stop reconciling and start measuring incrementality directly: a post-purchase survey asking how people heard about you, and periodically turning a channel off to see what actually happens to revenue. The attribution page covers how to set both up.
How much creative do we actually need?
More than you are making, almost certainly. Fatigue in competitive consumer categories can set in within five to seven days, which means the constraint on spend is usually production throughput rather than budget or audience size.
The honest way to size it is to count how many new concepts your testing campaign consumes per month at your current spend, then double the budget on paper and see whether that number is producible. If it is not, scaling will fail regardless of how the account is structured.
Should we be on TikTok as well?
Only if the category and the creative supply support it. Published TikTok acquisition costs run roughly $90 to $129, above Google and below the channel-level Meta estimates, but the creative does not transfer. Reposting your Meta library there produces the cost without the performance.
The test is whether you can sustain native creative for a second platform on top of the first. If the answer is no, more angles on Meta is a better use of the same production budget.
Our costs keep rising. Is that us?
Partly not. Meta CPMs rose roughly 20 percent year on year into 2026 and the increases have been compounding since the 2021 signal loss. Everyone in your auction is paying more for the same impression.
What is within your control is what each impression is worth to you. Conversion rate on the store, contribution per order, and repeat rate all raise the price you can afford to pay, which is the only durable way to win a rising auction.
Do you need to rebuild our account from scratch?
Usually not, and we would rather not. Rebuilding discards conversion history the account has spent months accumulating. What normally happens is consolidation: fourteen campaigns become two or three, the testing layer gets built properly because most accounts do not have one, and the event setup gets fixed underneath.
Where a rebuild is genuinely warranted is a broken or duplicated pixel setup, because in that case the history is not worth keeping. We will tell you which situation you are in during the audit rather than after.
The rest of the system
Acquisition is one of five
Paid social is the only channel here that reliably loses money on the first order, which is exactly why it cannot be bought on its own. It works when the other four are doing their jobs, which is how the whole ecommerce practice is set up to run.
Next step
Thirty minutes.
Bring your Ads Manager.
We will pull the last 90 days, work out what a customer really costs you against what one is really worth, and tell you whether the account or the retention program is the thing holding you back.
No obligation · Usually booked 1-2 days out