Meta Ads Management
Meta counts a sale when someone saw an ad and bought within a week, whether or not the ad had anything to do with it. Retargeting your own site visitors reports beautifully for exactly that reason. We work out what the spend is genuinely adding, and build the creative that adds more of it.
From £1,000/month. Ad spend paid direct to Meta, never marked up.
Three-month minimum, then rolling.
None of this is Meta behaving badly. It is a platform reporting every sale it can reasonably claim, using rules that were set to make the platform look useful. Knowing the rules is what stops you scaling into a number that is partly borrowed.
A useful test costs nothing: add up the revenue every platform claims for last month and compare it with what your bank received. The gap is the size of the problem.
For years the skill in Meta was audience construction: interest stacks, lookalikes, exclusion layers. Broad targeting and automated campaign types have taken most of that away. The algorithm now decides who sees an ad, and it decides largely on the basis of who responds to the creative.
Which means the creative is no longer the thing you make after the strategy. It is the strategy. Change the hook and you change the audience, because a different hook pulls a different person into the auction.
Practically, that shifts the work from account management to production cadence. A steady flow of genuinely different concepts, tested properly, retired when they fatigue. Not fifteen colour variations of the same idea, which teaches the algorithm nothing and burns budget finding that out.
01
New concepts on a schedule rather than when performance drops. Different angles, formats and messages, not variations on one idea. Winners scaled, fatigued creative retired before frequency does the damage rather than after.
02
New concepts tested in their own structure, so learning does not disturb the campaigns carrying your revenue. Mixing the two is how accounts end up volatile and nobody can say which change caused what.
03
Reported together, retargeting flatters everything and hides how acquisition is really performing. Split apart, you can see what it costs to reach someone who had never heard of you, which is the only number that grows a brand.
04
Conversions API properly implemented, event match quality raised, deduplication against the browser pixel checked. Poor signal quality means the algorithm optimises on a fraction of your real conversions and you pay for the difference.
05
Advantage catalogue ads are only as good as the feed behind them. The same feed work that drives Google Shopping serves Meta, which is one of the reasons running both together costs less than running them apart.
06
Blended contribution as the primary number, with holdout tests where spend justifies them. Turning Meta off in a matched region for a fortnight tells you more than any attribution model will.
In week one we agree a baseline and the number we are judged on. It is not platform-reported ROAS, for every reason set out above.
We report blended contribution: total revenue across every channel, less cost of goods, less all advertising, less fees. One number that cannot be improved by moving budget between platforms, because it already counts all of them.
The test worth running once a year
Switch Meta off in one region and leave it on in a comparable one. Two weeks is usually enough. Compare total revenue, not platform-reported revenue.
Most brands never do it, because the answer is uncomfortable either way. It is also the only method that tells you what the spend is genuinely adding, and it costs nothing but nerve.
Margin-based bidding needs accurate conversion values arriving reliably. Most accounts are bidding on partial signal without knowing it: consent handling that drops conversions across UK and EU traffic, browser restrictions eating client-side tags, values passed without cost of goods attached.
Standard work is included. Enhanced conversions, server-side tagging through a cloud container, Consent Mode configured properly, conversion values carrying margin rather than revenue.
Where a business needs more than that, we build custom server-to-server tracking so conversions are sent from your systems rather than from a browser that may never fire. That is quoted separately, because it is engineering rather than account management and it would be dishonest to pretend otherwise.
Brands who have asked not to be named. Figures are their own platform data, from the year before we started to total revenue since.
Client since May 2021
Google Shopping rebuilt, Meta prospecting and retargeting added, site conversion rate doubled, seasonal budget scaling. Still running four years later.
Year before we started
First full year
Total since
Client since April 2022
Launched with no digital presence at all. Google Ads, Meta, Amazon and eBay built from zero in a single quarter, with seasonal inventory and budget forecasting from the start.
At launch
First full year
Total since
Client since January 2022
German market entry from a standing start. Shopping localised rather than translated, Meta EU built from audience research, Amazon.de listings and PPC, margin modelling behind the pricing.
Year before we started
First full year
Total since
Client outcomes under specific conditions, not a forecast for your brand. Figures are cumulative revenue since engagement, taken from the clients’ own platform reporting.
Week 1
Conversions API and event match quality checked first, because everything after depends on the signal being clean. Cost of goods collected, blended baseline agreed in writing before anything changes.
Weeks 2 to 4
Prospecting and retargeting separated, campaign structure simplified, testing moved into its own space. First batch of new creative concepts into production. Expect volatility while the algorithm relearns on cleaner data.
Weeks 5 to 8
Concepts that work scaled, tired ones retired, the production cadence settled into a rhythm. First contribution review against baseline at day 60, with a holdout test planned if spend justifies one.
Meta on its own. If other channels need running too, the programme is better value and the table shows where it lands.
We direct and brief creative, and we will edit and assemble from footage you have. Original video and photo shoots are a production cost quoted separately, because folding them into a retainer misprices both.
Custom server-to-server tracking is quoted separately. Everything else above is included.
It works when
It does not when
The creative point is worth taking seriously before you start. Meta rewards volume and variety, and a brand that can approve four new concepts a month will beat one that approves four a quarter, whoever is running the account.
Because the two numbers measure different things. Meta reports every sale it can reasonably attribute, including view-through conversions and sales that other channels are also claiming. Blended contribution reports what your business actually kept. Both can be accurate at once. Only one of them pays your suppliers.
You need a steady supply, yes, and it is the single biggest factor in whether Meta works for a brand now. It does not have to be expensive. Phone footage, customer content and product demonstrations often outperform polished production. What does not work is running the same three ads for six months and concluding the platform is broken.
Rarely, and less every year. Broad targeting with strong creative generally beats narrow interest stacks, because the algorithm has more signal than any manual segmentation can reproduce. Exclusions still matter, particularly for existing customers, and there are categories where narrower targeting still earns its place. We test rather than assume.
You switch a channel off in one region and leave it running in a comparable one, then compare total revenue rather than platform-reported revenue. It is the only reliable way to know what a channel is genuinely adding. It needs enough spend and enough geography to be readable, so it is not for everyone, and we will tell you if you are not there yet.
No. Percentage pricing pays an agency more for spending more of your money, which is the opposite of the job. Flat fee, so recommending you cut spend costs us nothing.
It changes the measurement significantly. Meta drives a meaningful share of marketplace sales it will never record, so brands routinely underrate it and cut budget that was working. We connect the two and judge it on the blended number, which is covered in the full programme.
Vaseem Shaikh has spent fifteen years running paid media for ecommerce brands. Before that he spent five years researching artificial intelligence, genetic algorithms and machine learning at IIM Ahmedabad, work he has published and now applies directly to how bids, budgets and creative tests are structured.
That background is the reason results here are read at significance rather than at a convenient moment, and why the reported platform number is never the one we are judged on.
He leads a team of specialists across paid media, creative, content and analytics. You speak to him during the audit, and he stays involved in the account afterwards.
Google Ads Advisors is an invited research community Google consults on products and reporting ahead of general release. It is not a partner tier or a certification.
Vaseem Shaikh
Founder, Signal Over Noise
Account structure, creative performance, tracking and signal quality, and an honest read on how much of your reported return is genuinely incremental. Written down, about five working days, no charge.
You keep it either way. Hand it to your current agency if you want to.
Start a project
Tell us about your brand. You get a written audit of your advertising, site and marketplace position, ranked by what each fix is worth.
No charge. About five working days. We reply within one working day, Monday to Friday, 9:00 to 18:00 UK.
London, UK
71–75 Shelton Street
Covent Garden, London, WC2H 9JQ
Ahmedabad, India
DTC & ECOMMERCE GROWTH AGENCY
Services
Agency
We usually reply within a few hours
Our team will be in touch with you shortly.
