Start with the uncomfortable truth: nobody outside a company's ad accounts knows what that company spends. Not the spy tools quoting a monthly figure to the dollar, not the agencies citing them, nobody. Google and Meta publish ads, not budgets. Every spend number you have ever seen for a competitor was modeled from indirect signals, and most were modeled badly.
That does not make the question worthless. It makes precision worthless. What you can build honestly is a range, wide where the evidence is thin and narrower where it is strong, and a range built from public signals is genuinely useful for the decision it usually serves: how much would it cost to compete here, and who is actually committed?
What is actually public
Four things, and only four. The existence and count of a competitor's ads, published in the Meta Ad Library and the Google Ads Transparency Center. How long each ad has run, via start dates on Meta and shown-date ranges on Google. The formats and regions. And, in two special cases, real numbers: spend ranges for political ads, and reach figures for ads shown in the European Union, where the law requires publication.
Everything else, including every point-estimate of commercial spend, is inference. So the method is to reason from the four public signals, carefully, and label everything downstream of them as the estimate it is.
Signal one: ad count, weighted by what ads cost to run
Count the live ads per platform from the transparency tools, using the resolution steps in our competitor ad analysis workflow so the counts are real. Then weight, do not just count. Ten search text ads can sit on a modest budget. A rotating slate of video campaigns across YouTube and Reels almost never does, because video programs carry production cost and are rarely run at trivial spend. Format mix is a budget signal in itself.
Be careful with raw counts in both directions. A large ad count can be one serious program or a pile of barely served variations. A small count of long-running ads can carry more spend than a large count of experiments. Count, then look at what the ads are.
Signal two: longevity as commitment
Advertisers do not pay for months for ads that lose money. An ad running continuously for a long stretch is evidence of sustained budget behind it, and a competitor whose oldest ads go back years is telling you their program survived every budget review in that period. Continuous history reads as committed spend. Stop-start history reads as small or experimental budgets, whatever the ad count says today.
Signal three: price the search side from its market
For search advertisers you can go one step further, because the cost side of search is partly public. Keyword planning tools publish the going cost-per-click range for commercial terms, and search volumes for them. If you know which terms a competitor's text ads clearly answer, you can multiply a plausible share of that click market by the published cost band and get a monthly figure range for their search program. Every input is a range, so the output is a range. That is not a weakness of the method. It is the method.
Build three scenarios, present a band
Combine the signals into three numbers: a floor, where the competitor buys cheaply and captures little; a central case; and a ceiling, where they pay top of the cost band for a strong share. Present all three. A band of this shape, grounded in counted ads and published cost ranges, will change decisions: it tells you whether competing is a small bet or a serious one, and against whom. A false point estimate changes decisions too, in whatever direction its errors point.
Every input is a range, so the output is a range. That is not a weakness of the method. It is the method.
The traps
False precision. The moment an estimate is written as one number, the assumptions vanish and the number hardens into fact by repetition. Keep the band visible everywhere the estimate travels.
Spy-tool point figures. Third-party tools quoting exact monthly spend are running the same inference you can run, with less care about your specific market, then hiding the uncertainty. Use their inputs where useful. Never inherit their confidence.
Cross-market comparison. Click costs differ enormously between countries and niches, so an ad count that implies a large budget in one market implies a small one in another. Estimate within a market, never across.
Answering the wrong question. Often the decision does not need spend at all. Who advertises, who has run the longest, and which messages are unclaimed are knowable exactly, from public sources, and they usually matter more than the budget guess. Estimate spend when the decision truly needs it, and read the market either way.
The shortcut
All of this starts with counted ads, resolved identities, and longevity data, which is exactly what a benchmark contains. Request a free ad benchmark for your market: Muffin Intel tracks the ads, we read them, the written benchmark lands inside one business day.
Frequently asked questions
Can any tool show a competitor's exact ad spend?
No. Spend for commercial advertisers is not published anywhere: not in the Meta Ad Library, not in the Google Ads Transparency Center, not through any API a third party can read. Tools quoting exact figures are modeling from signals like the ones in this guide and presenting the model's output as a measurement. The exceptions are political ads, where platforms publish spend ranges, and EU reach data, which is audience size rather than money.
Is spend even the right thing to estimate?
Often not. Spend answers "how big is their budget," but most competitive decisions turn on questions the public record answers exactly: who is advertising at all, whose ads have run longest, which platforms they trust, and which messages nobody has claimed. Those are counts and dates, not estimates. Reach for the spend band when you are sizing the cost of entering an auction. Reach for the exact signals for everything else.
How wide should the range be?
As wide as the evidence honestly forces, and it is common for the ceiling to be several times the floor. That still tells you what you need: whether a competitor's program is small, serious, or dominant, and what committing to compete would plausibly cost. If a range that wide feels unusable, the fix is more evidence, such as longer observation of ad turnover, never false narrowing.
Does a big ad count mean a big budget?
Not by itself. Platforms make it cheap to generate variations, so one advertiser's 200 ads can be a modest automated program while another's 12 long-running video campaigns carry far more spend. Weight the count by format, longevity, and turnover before reading money into it. The combination of many ads, heavy video, wide regions, and long continuous history is the pattern that reliably signals real budget.
Free
See what your market is advertising
Written by Sarah Mitchell, Muffin Media
Want this run on your brief?
