I Measured 900 Apify Actors and Got the Price Wrong by 100x. Here's the Corrected Data

I set out to answer a boring question with data instead of opinion: on the Apify Store, who actually makes money, and from what?

I got an answer, published it internally, and then discovered my central example was wrong by a factor of a hundred. This is the corrected version, including the mistake, because the mistake is the most useful thing in here for anyone doing the same measurement.

The method, and its two limits

Two public endpoints do all the work:

• GET /v2/store?limit=1000&offset=N — the catalogue
• GET /v2/acts/{username}~{name} — pricing, creation date, monthly users, monthly run counts

Revenue estimate per actor:

runs_30d × price_per_primary_event × 0.8

The 0.8 is the developer's share. This is a lower bound: it assumes one billable result per run, and most runs produce more.

Limit one: you cannot crawl the whole store. The endpoint reports a total of about 56,000 and Apify's own store page advertises more than 67,000 actors. But it stops returning records past an offset of roughly 16,000, and every page comes back short of the limit you asked for. Deduplicated, I ended up with 13,190 actors — under a quarter of the catalogue, and biased toward whatever the default ranking puts first. So: no store-wide survival rates from me. What follows is about the 900 most-used actors, enriched individually, which is where essentially all the revenue is anyway.

Limit two is the interesting one.

The pricing trap that makes every number wrong

Pay-per-event actors can price each event in tiers, keyed to the customer's Apify plan. The API returns them like this:

"eventTieredPricingUsd": {
  "FREE":     { "tieredEventPriceUsd": 0.1 },
  "BRONZE":   { "tieredEventPriceUsd": 0.001 },
  "SILVER":   { "tieredEventPriceUsd": 0.0008 },
  "GOLD":     { "tieredEventPriceUsd": 0.0006 },
  "PLATINUM": { "tieredEventPriceUsd": 0.0006 },
  "DIAMOND":  { "tieredEventPriceUsd": 0.0006 }
}

Read the first key and you conclude this actor charges $100 per 1,000 results. It doesn't. FREE is a deterrent price for people on the free plan — who, as Apify's own docs state, generate no developer revenue at all. Every paying customer is on BRONZE or better: $1 per 1,000, a hundred times less.

My original write-up leaned on exactly this actor as proof that expensive actors win. I had it at $8,600 a month. Corrected, it's about $86.

60% of the top 900 actors use tiered pricing, and 467 of them have a FREE tier priced above their paid tiers. If you take the first key, your revenue total comes out 21% too high and individual actors can be off by up to 100×. The fix is one line — take the maximum across the non-FREE tiers — but you have to know the trap is there.

Everything below is computed the corrected way.

Who takes the money

Five accounts — apify, compass, clockworks, junglee, lukaskrivka — are Apify's own and Apify-affiliated. Among the 900 most-used actors they hold 48% of the users and 44% of the revenue.

That's the single most important correction to make to any niche research. Instagram and Google Maps look enormous until you exclude those accounts, and then most of the money vanishes, because it never reached a third-party developer.

Of roughly $400,000 a month across these 900 actors, about $224,000 is left for everyone else.

The finding that changed how I build

Third-party actors only, grouped by what they charge per 1,000 results:

Price per 1k	Actors	Total/month	Median actor	Median runs/month
under $1	193	$11,409	$5	18,876
$1–3	267	$24,940	$9	5,748
$3–6	199	$82,982	$24	7,153
$6–15	97	$25,018	$23	3,221
over $15	71	$79,725	$111	2,799

Read the last two columns together. Actors priced over $15 per 1,000 do one seventh the traffic of the sub-$1 group and make twenty-two times more money.

The cheap end is a logistics business. 193 actors grinding through nearly nineteen thousand runs a month each, to earn five dollars.

The top earner in the whole sample is a LinkedIn profile search priced at $100 per 1,000 search-page results — genuinely, at every paid tier — doing 651,000 events a month for about $52,000. Four of the top twelve earners are LinkedIn actors, three of them third-party.

The mechanism is not clever. A buyer prices your scraper against the cheapest scraper doing the same thing. A buyer prices your enrichment tool against ZoomInfo, Hunter and Snov, which charge in dollars per contact. Same engineering, different comparison set.

What $500 a month actually requires

Pure arithmetic, after the 20% platform cut:

Price per 1,000	Rows needed per month
$1	625,000
$4	156,000
$20	31,000
$100	6,200
$200	3,100

A 200× spread in what you have to build and operate for the same money.

How many actually get there

Among the 829 third-party actors in the top 900 by usage — the most-used quarter of the store's most-used slice, so this is the optimistic end of the distribution:

Threshold	Actors clearing it
any revenue at all	827
$100/month	175
$500/month	62
$1,000/month	35

Median monthly revenue among those that earn anything: $14.

Remember these are the best-used actors in the store. The tail below them is far worse, and I can't measure it, because the catalogue endpoint won't page that far.

Runs per user, not users

Median runs per user across the sample sits between 15 and 90 depending on the niche.

Hundreds of runs per user means the actor is wired into a pipeline or called in a loop by an agent. Under ten means people tried it once and left, however good the user count looks on the store page. It's the cheapest health check you can compute and almost nobody looks at it.

The mechanics that gate all of this

Worth knowing before you start:

• 80% of revenue, minus platform usage costs. The split is confirmed in the API as a 20% margin, but Apify's terms deduct platform costs from your share, not from the gross. Compute per run.
• Only paying customers count. Free-plan usage earns nothing, which is exactly why the FREE pricing tier exists and exactly why reading it as your price is wrong.
• Payout minimum: $20 for PayPal and Wise, $100 for other methods. Below the threshold, the balance rolls over month to month.
• KYC is mandatory before any payout: ID, proof of address, tax data.
• Money is slow. Invoices generate on the 11th, auto-approve on the 14th, payment follows.
• Significant price changes take 14 days to reach existing users — a switch of model, a price increase, a new paid event. Reductions apply immediately.
• Rental pricing dies on 30 September 2026. Anything not migrated to pay-per-event collects nothing after that date.

The trend to plan against

Apify has published its developer payouts periodically: about $563,000 a month in September 2025, and $1.5 million a month as of August 2026. That's roughly a 2.7× gain in eleven months.

I'd be careful with any tighter extrapolation than that — the earlier figures I originally used came from banner text and undated tweets, and the banners sat unchanged for months at a time, so a smooth curve fitted through them is fiction. The direction is solid and the driver Apify names is explicit: actors used as tools by AI agents.

Meanwhile new actors keep arriving faster than the pie grows. Which brings it back to the price table — that race isn't won on volume.

What I'd tell myself at the start

Check demand before feasibility. I built several actors for sources that were technically satisfying and commercially empty — open government data is the sharpest version of this, where the API answers instantly, there's no anti-bot, nobody has taken the niche, and the niche leader has three users.

Price against the incumbent SaaS, not against the cheapest scraper.

Pick one vertical and stay in it. My own profile still shows the opposite — travel, Southeast Asian marketplaces, contact verification, B2B lead feeds — which is three themes too many, and cross-linking them is a patch, not a fix.

And verify your own measurement before you build a strategy on it. Mine survived, but only after the correction: the shape of the price table is the same, the magnitudes were not.

The actors are at apify.com/lergassy. The measurement is two endpoints and an afternoon — go compute it for your own niche, and take the maximum of the paid tiers.