Most D2C ad accounts are not underperforming. They are under-measured. Meta cannot optimise towards a purchase it never hears about, so it optimises towards the cheapest thing it can still see.
The first month of paid advertising goes the same way for most D2C brands. A page post gets boosted because the button is right there. Somebody sets up a campaign with an audience built from interests that sounded correct. Money leaves the account. Quite a lot of it. At the end of the month there are some orders, no clear idea which ads produced them, and an argument about whether any of it worked.
Almost none of that is a creative problem. It is a setup problem, and the setup is the part you only have to get right once.
What follows is the order to do it in: the tracking to fix before you spend anything, an account structure that will not fight you, the first campaigns worth running, and how to judge them without strangling the thing while it learns.
Fix the plumbing before you spend anything
Meta renames its tools often enough that any screenshot ages badly, so this is by function rather than by menu path. Before the first campaign goes live you want all of the following in place:
A Business Manager account that your business owns, with the ad account, page and assets inside it. Running ads from a personal profile or a freelancer’s account is how brands lose access to their own history.
Your domain verified against that business, which is what lets you control how your own pages are used and prioritised in measurement.
The pixel, or dataset, installed on every page of your ecommerce store and firing the standard events: view content, add to cart, initiate checkout and purchase.
The Conversions API sending those same events from your server, matched to the browser events so they deduplicate rather than double count.
A product catalogue that is complete and syncing, because the strongest retargeting format on the platform depends on it.
App events wired up if you have an app, so installs and in-app purchases report back rather than disappearing.
Payment, billing and a backup card sorted, so nothing pauses mid-flight.
Doing this after you start spending means the first weeks of data are partly wasted, and that data is exactly what the system uses to find buyers later.
Tracking is the whole game now
Browser-side tracking alone under-reports, and has done since the privacy changes that let people opt out of app tracking. Ad blockers, blocked cookies and people who buy on a different device all quietly remove conversions from the record.
That matters more than most brands realise, because the platform does not simply report fewer sales. It optimises towards what it can still measure. Starve it of purchase events and it will happily find you the cheapest clicks and the cheapest add-to-carts it can, which is not the same job.
The fix is boring and effective. Send every important event twice, once from the browser and once from your server through the Conversions API, and give each one a shared event identifier so the two copies collapse into a single conversion. Then check in your events tool that purchases are arriving from both sources, that the match quality is reasonable, and that you are not counting each sale twice.
If you also sell through an app, get the app events reporting too. A buyer who saw an ad on a phone, installed the app and bought there is invisible otherwise, and the ads that produced those orders will look like failures and get switched off.
Keep the account boringly simple
The instinct is to split everything: audiences by interest, ad sets by age, campaigns by product. That instinct comes from an older version of this platform and it now works against you.
Each ad set needs enough conversions to learn. Splitting a modest budget across fifteen of them means none of them get there, so all of them stay in the expensive guessing phase and your results look random.
For most D2C brands the structure that works is close to this: one prospecting campaign with broad targeting, one retargeting campaign driven by the catalogue, and budget consolidated at campaign level rather than micromanaged per ad set. Let the system find the people. It is better at that than an interest list you assembled from intuition.
The first three campaigns worth running
Prospecting, optimised for purchases, targeted broadly. Resist the urge to narrow it. Your creative is doing the targeting now, and the system reads who responds far faster than you can guess who might.
Catalogue retargeting for people who viewed a product or abandoned a cart. This is usually the best-performing thing in a new account, and also the one people over-credit, because those buyers were already most of the way there.
One conversation-led campaign, if your customers prefer messaging to forms. Click-to-WhatsApp ads put the enquiry in a thread you can follow up, which suits considered purchases and anything that needs a size, a quote or a question answered before somebody commits.
If you sell through an app as well as a website, add an app install or app event campaign once the first two are stable. Do not start there. An install is worth very little until the app is the thing your repeat buyers actually use.
Creative is the targeting now
Since the targeting levers got blunter, the creative does most of the work of deciding who sees you. That changes what you should spend your time on.
Make more of it than feels reasonable. A handful of concepts a month, not one polished film a quarter.
Earn the first two seconds. Most of the audience decides in that window, and nothing after it matters if they have already scrolled.
Shoot vertical and square, because that is where the placements are.
Caption everything. A large share of views happen with the sound off.
Vary the format rather than the filter: someone talking to camera, the product being used, a customer’s own footage, a plain before and after, a straight offer.
Let the losers die quickly and keep the winners running far longer than feels comfortable.
A note on user-generated footage: it works because it looks like the rest of the feed, which means the moment it gets polished into an advertisement it stops working. That is the whole trick.
Budget, the learning phase, and leaving it alone
A new ad set enters a learning phase while the system works out who to show it to. Meta’s own guidance puts the exit from that phase at roughly fifty optimisation events a week, which is a useful planning number: it tells you the smallest sensible budget is whatever buys about fifty purchases a week at your current cost per purchase.
If that is out of reach, optimise for an event that happens more often, such as add to cart, and move up to purchases once volume allows. That is a deliberate compromise rather than a permanent setting.
The most expensive habit in a new account is daily fiddling. Every meaningful edit restarts learning. Give a test a week and enough conversions to mean something, judge it then, and make changes in batches rather than every morning.
What to measure, and what to ignore
The platform reports the numbers that flatter the platform. Read them accordingly.
Watch blended performance: total revenue against total ad spend for the period. It is unglamorous and hard to argue with.
Know your contribution margin per order. Return on ad spend means nothing until you know what is left after product cost, shipping, payment fees and returns.
Compare acquisition cost against what a customer is worth over a year, not against a single order, particularly if people reorder.
Treat platform-reported return on ad spend as directional. Attribution windows overlap and every channel claims the same sale.
Ignore click-through rate, cost per click and engagement as success measures. They are diagnostics for why something is or is not working, not the thing you are buying.
One more that gets missed in markets where cash on delivery is common: measure delivered revenue rather than orders placed. A campaign that produces cheap orders which never get accepted is not a cheap campaign.
Where the first month usually goes
Spending before the Conversions API is live, then wondering why results decay.
Fifteen ad sets sharing a budget that could support two.
Turning off an ad after a day because it had no sales yet.
Judging everything on the platform’s own return figure and never checking it against the bank account.
Retargeting a tiny audience with most of the budget, then celebrating a return the brand would mostly have earned anyway.
One creative concept, run until the audience is sick of it.
Frequently asked questions
How do I set up Meta ads for a D2C brand?
Get the account, domain, pixel, Conversions API and product catalogue in place first. Then run one broad prospecting campaign optimised for purchases and one catalogue retargeting campaign, keep the structure simple, and give each test a week before judging it.
How much should a D2C brand spend on Meta ads to start?
Enough to buy roughly fifty purchases a week at your current cost per purchase, since that is the volume Meta’s guidance associates with leaving the learning phase. Below that, optimise for a more frequent event such as add to cart until sales volume catches up.
Do I still need the Conversions API if the pixel is installed?
Yes. Browser tracking alone misses conversions from blocked cookies, opted-out devices and cross-device purchases. Sending the same events from your server and deduplicating them gives the system a fuller picture to optimise against.
Should I use broad targeting or detailed interests?
Broad, in most cases. Interest targeting fragments budgets and keeps ad sets stuck in learning. Creative is the effective targeting mechanism now, so put the effort there and let the system find the audience.
Are click-to-WhatsApp ads worth running?
They are, where buyers prefer a conversation to a checkout. WhatsApp ecommerce suits considered purchases, anything needing sizing or a quote, and markets where messaging is the default. The catch is that somebody has to answer quickly, so only run them if that is covered.
What is the difference between Meta ads and boosting a post?
Boosting optimises for engagement on one post from a simplified interface. Proper campaigns in Ads Manager let you optimise for purchases, use the catalogue, control placements and structure budgets. For online advertising that is meant to sell rather than to be seen, use the full tool.
How long before Meta ads start working?
Plan on a few weeks rather than a few days. The first week is mostly the system learning, the second gives you a readable signal, and a sensible judgement on creative and structure comes after that. Changes made daily reset the clock each time.
Where to go from here
For the wider picture of how paid and organic digital marketing work together, we wrote about social media and ads for ecommerce. Ads also only pay back if the brand behind them is worth returning to, which is why branding is important. If you would rather this was run for you, our marketing plans include managed Meta campaigns and the creative that goes in them, and they are on the pricing page.
