A competitor ad that is scaling cannot be identified from spend, because retail ad spend is never published. It is identified by what scaling leaves behind: an unusually long run compared with its own launch cohort, many variants of one concept rather than many concepts, a widening placement and aspect ratio list, localised versions, a hook that keeps getting refreshed while the core creative stays, and a landing page built specifically for it. Any one of those can be a coincidence. Two or three together, confirmed a fortnight apart, is a brand protecting something that works. Read the argument the ad is making, not the execution.
- Spend is private for retail ads, so every usable signal is about persistence and footprint.
- Run length only means something measured against the same advertiser own launch cohort.
- Count concepts rather than ads: a high ad count often means nothing has worked yet.
- Maintenance is the strongest signal, because new sizes and localisations cost real hours.
- One snapshot proves nothing. The Persistence Test needs a second look two weeks later.
A competitor ad is scaling when the brand running it is deliberately increasing its budget and its placement footprint, and the thing you need to know first is that you cannot observe that directly. Spend data is not published for retail advertising. What is published is a public record of which ads are live, when each started, and where each is being shown, and scaling leaves marks in that record that a careful reader can find. Last updated: September 2026.
Omniconvert has scored storefronts on their Creative and Ads dimension against real competitors in the same category and country, drawing on CROBenchmark-scale data covering 70,000+ experiments and 13 years in eCommerce. The consistent finding is that teams overrate what they can see in an ad library and underrate what they can infer from it. The spend number they want does not exist. The behavioural record they ignore is sitting there in plain view.
This piece sets out the six signals worth reading, how reliable each one is, and the Persistence Test that stops a coincidence being presented as a finding. For the wider method this sits inside, see the DTC competitor research stack, and for the mechanics of the library itself, our guide to how to Reverse-Engineer competitor ads with the Meta ad library.
What scaling actually means, and why you cannot see it
That constraint is worth sitting with, because it disqualifies most of what gets presented as competitor ad intelligence. A tool showing you a rival monthly ad spend for a clothing brand is producing an estimate from a model, and the model cannot see the account. Treat those figures as what they are, a guess with a confident interface.
What you can see is a record of decisions. Every ad in a public library carries a start date, a live or inactive status, and a list of the platforms and formats it runs on. That is enough, because scaling is not a quiet decision. It consumes budget, it consumes the creative team hours, and it produces artefacts.
The useful mental model is that you are reading maintenance rather than money. A brand does not commission four new aspect ratios, a localised voiceover and a dedicated landing page for an ad it is about to switch off. Those artefacts are the receipt for a decision you cannot see directly.
How to read a competitor ad without over reading it
Three habits keep the read honest.
First, always compare within the advertiser. Brands have wildly different creative operating rhythms. One ships fifty ads a month and kills them in a fortnight. Another runs eight ads a year. The same observation carries opposite meanings in those two accounts, and a benchmark drawn across advertisers will mislead you in both directions.
Second, record before you interpret. Keep a simple sheet with one row per concept: the concept name, the launch date, the number of live variants, the placement list, and the landing page it points at. The signals below are all differences between two readings of that sheet, and a difference cannot be computed from a single visit.
Third, separate the concept from the asset. Platforms generate multiple assets from one upload, so a naive ad count inflates. What you want is the count of distinct arguments, not the count of files. Statista consumer survey work on advertising recall consistently shows that repetition of a single message is what builds recognition, and it is that concentration you are trying to detect.
The six signals that a competitor ad is scaling
1. An unusually long run against its own cohort. Note every ad the advertiser launched in the same week, then watch which ones survive. An ad still live when its siblings have gone dark has passed an internal decision the others failed. This is the cheapest signal to collect and the easiest to overweight, because a brand that never prunes anything produces long runs by neglect rather than by choice.
2. Concentration, not proliferation. Count distinct concepts rather than assets. Twenty ads expressing one argument is concentration, and it means the argument is earning budget. Twenty ads expressing twenty arguments is a test programme, and it often means nothing has worked yet. Teams routinely read a high ad count as strength when it is the signature of a brand still searching.
3. A widening placement and format list. An ad that started in one feed and now appears across stories, reels, in stream and on a partner network has been rebuilt for each surface. That rebuild is a deliberate act. A static ad sitting in its original single placement for months is more likely evergreen than scaling.
4. Localisation. Translated copy, a re recorded voiceover, currency and delivery terms changed for another market, a new set of on screen captions. Localisation is the most expensive thing a brand can do to an ad and the least likely to be done speculatively. When you see it, someone has already decided the concept works and is buying reach with it.
5. A refreshed hook on an unchanged core. The first three seconds change, the middle stays. This is the specific fingerprint of a team fighting creative fatigue on a winner rather than testing a new idea. Nielsen long running work on advertising wear out describes exactly this pressure: effectiveness decays with exposure, so an ad being pushed harder needs its opening rebuilt to keep working. A brand only pays that cost for an asset it does not want to lose.
6. A dedicated landing page. Follow the click. A scaling ad frequently earns its own page, with the ad claim repeated in the same words, a matching offer, and sometimes its own reviews block. Building and maintaining that page is a cross team cost, and it is the clearest single indication that the ad is now a channel rather than a test.
None of the six is proof. Read them as a portfolio: one signal is an observation, two or three that agree is a finding, and the ordering above tells you how much weight each deserves.
How reliable each signal is
| Signal | Cost to the rival | Reliable read | Tempting misread |
|---|---|---|---|
| Long run vs cohort | None | It survived a pruning decision | Long run means high spend |
| Concentration on one concept | Low | The argument is earning budget | Any high ad count is strength |
| Wider placements and formats | Medium | Rebuilt deliberately per surface | Platform auto placement counted as intent |
| Localisation | High | Reach is being bought with it | A translated caption read as full localisation |
| Refreshed hook, same core | High | Fatigue being managed on a winner | A new hook read as a new concept |
| Dedicated landing page | Highest | The ad is now a channel | A generic campaign page counted as dedicated |
Work the table from the bottom up when time is short. The two highest cost rows can usually be checked in ten minutes each, and they will tell you more than a morning spent counting ads.
The Persistence Test, and what a growth lead should do this week
The test exists because a single library visit cannot distinguish a scaling ad from an ad that happened to be live the day you looked. Run it before any finding reaches a planning meeting.
- Pick three rivals and record one sheet per brand. Concept, launch date, live variant count, placement list, landing page. An hour each. This is the baseline every signal is measured against, and without it the next visit produces impressions instead of differences.
- Diarise the second look for a fortnight from now. Same fields, same order. The comparison is the deliverable, not the first sheet.
- Check the two high cost signals first on your single strongest candidate. Follow the click to see whether the page is dedicated, and look for localised versions. Ten minutes, and it usually settles the question.
- Write the argument down in one sentence, then test it yourself. Not the creative, the claim. Get your own free leaderboard score on the Creative and Ads dimension to see where your current creative sits against real competitors in your category and country before you commit budget to matching theirs.
Then treat the result as a hypothesis with a price attached. A rival can profitably run an ad that would lose you money, because their margin, delivery promise and repeat rate are not yours. Deciding which of several competitor derived hypotheses to fund is a prioritisation problem rather than a research one, and Nexus by Omniconvert is an AI for eCommerce growth engine that unifies commerce data, ranks experiments by True Profit, and generates campaigns and creative you approve before they go live. The controlled testing half of that work belongs to Omniconvert Explore.
What the library will never tell you
The money, as established, is not published for retail advertising. Accept it rather than buying an estimate of it.
The margin is invisible and decisive. An ad driving a high volume of a low margin product can be a strategic loss leader, a clearance exercise, or a mistake nobody has caught yet. You are watching the top of a funnel whose economics you cannot see, and the brand may not be winning at all.
The motive is invisible too. A brand may be scaling an ad to defend a category position, to hit a quarterly target, to support a retail listing, or because a founder likes it. None of those reasons is visible in the library, and only the first two would be worth matching.
What survives all three caveats is still valuable: a public record of which arguments a rival has decided are worth protecting. That is a genuinely useful input to your own creative pipeline, provided it enters as a hypothesis to test and not as a conclusion to copy.
The bottom line
Stop trying to estimate a rival ad spend and start reading the maintenance. The six signals in order of value are a dedicated landing page, a refreshed hook on an unchanged core, localisation, a widening placement list, concentration on one concept, and a long run against its own cohort. The last of those is the one most teams collect and the one worth least. Run the Persistence Test before any of it reaches a plan: compare within the advertiser, look twice a fortnight apart, and ask who is still doing work on this ad. Then take the argument, not the execution, and price it against your own margin before you fund it. A rival scaling decision is evidence about their business, and it only becomes evidence about yours after you have tested it.