Apify Store Trends: demand, ratings, failing incumbents

Which Apify Store niches have demand, and which incumbents are weak. One JSON row per store actor and one per niche, for anyone deciding what to build, what to wrap, or which competitor to go after:

• Demand: users in the last 7, 30, and 90 days and lifetime, plus public runs in the last 30 days. Apify publishes no revenue, so users_30d is the demand proxy throughout.
• Reliability: the 30-day fail rate (failed + timed-out over total public runs), the maintenance notice, and days since the actor last ran.
• Quality: the store rating and review count, with ratings under three reviews ignored where a judgement is made.
• Price: the listed price in one readable label ($0.0027/Result, $0.005/item, $30/mo, FREE) and as a number.
• Gap signals per actor: leader_failing, unrated, under_maintenance, idle_14d, low_rated.
• Target rows per niche: total demand, actor count, the leader and its share, the leader's rating and fail rate, the user-weighted rating, the best rating anyone holds, how many actors are broken, and the A to G shortlist flags (poorly served, thin market, monopoly, and so on).
• Change over time: point every run at the same named dataset and each row carries what it looked like last snapshot: previous users, the change, previous rating and fail rate, and whether the actor is new.

This is the same scan the developer ran to pick their own niches on the store, turned into an actor. Two modes: search answers one question in seconds ("who sells Google Trends data and are they any good?"); full sweeps the whole catalog for a monthly snapshot.

Source: the public store endpoint api.apify.com/v2/store, no key, no login. Nothing is guessed; missing values are null.

Quick start

Everyone selling Google Trends data, with a niche summary row on top:

{ "query": "google trends" }

A niche with only serious incumbents, sorted by demand:

{ "query": "bizbuysell", "minUsers30d": 20, "maxActors": 50 }

A monthly snapshot of the whole store with deltas against last month (schedule it; about ten minutes a run):

{ "mode": "full", "maxActors": 5000, "minUsers30d": 5, "snapshotDatasetName": "store-snapshots" }

Input

Field	Type	Notes
mode	search | full	search runs the store's own search for query: a few requests, seconds. full sweeps the whole catalog across every category and pricing model: tens of thousands of actors, about ten minutes, meant for a scheduled monthly snapshot. Default search.
query	string	Required when mode is search. Ignored in full mode.
minUsers30d	integer	Drop actors with fewer users than this in the last 30 days. Default 5.
maxActors	1 to 5000	Keep at most this many actors, the most used first. Default 200. Set 5000 for a complete snapshot in full mode.
includeTargets	boolean	Also emit one aggregate row per target bucket, and in search mode one for the query itself. Default true.
snapshotDatasetName	string	Optional named dataset in your account. Every actor row is appended to it; before writing, the newest snapshot already there is read to fill delta. Use the same name every run. Not charged again.
requestDelaySeconds	number	Politeness delay toward the store API. Default 0.2.

Output

Rows come in two shapes, told apart by row_type. Target rows are pushed first, then actor rows sorted by users_30d descending (in search mode, actors whose title or name match the query come before the rest). The dataset has an actors view and a targets view.

Actor rows (row_type: "actor")

Field	Meaning
actor, actor_id, title, url	username/name, the store's internal id, the display title, and the store page.
target	The niche bucket the actor falls in, matched by regex over title and name (google trends, zillow, government/permits, about 230 buckets). (unmatched) when nothing fits.
query_match	search mode only: whether every word of query appears in the actor's title or name. The store's search also matches single words and descriptions, so google trends returns every Google actor; the ones that are about the query come first and carry true. null in full mode.
categories	The store categories the author chose, as a list.
users_7d, users_30d, users_90d, users_total	Distinct users who ran the actor in each window. users_30d is the demand proxy.
runs_30d	Public runs in the last 30 days.
fail_rate_30d	(failed + timed-out) / total public runs in the last 30 days. null when there were no runs.
rating, reviews	Store rating (0 to 5, two decimals; null when unrated) and review count.
bookmarks	Bookmark count.
price, price_usd, pricing_model	A readable label ($0.0027/Result for pay-per-event at the free tier, $0.005/item, $30/mo, FREE), the same number alone, and the raw model (PAY_PER_EVENT, PRICE_PER_DATASET_ITEM, FLAT_PRICE_PER_MONTH, FREE).
notice	The store's status notice, e.g. NONE, UNDER_MAINTENANCE.
days_since_last_run	Days since the actor last started a run.
agentic_payments	Whether the store lists the actor as enabled for agentic payments.
gap_signals	Why this incumbent might be beatable, any of: leader_failing (fail rate 30%+ with 10+ monthly users), unrated (under three reviews with 20+ monthly users), under_maintenance, idle_14d (no run for two weeks), low_rated (under 3.6 with three or more reviews). Empty when nothing stands out.
snapshot_at	When this run read the store (UTC). Every row in a run shares it; it is the key that groups a snapshot.
delta	Change since the previous snapshot in snapshotDatasetName: users_30d_prev, users_30d_change, rating_prev, fail_rate_prev, is_new (the actor was not in the previous snapshot). All null when no snapshot dataset is given or it is still empty.

Target rows (row_type: "target")

Field	Meaning
target, target_kind	The bucket name. site rows aggregate the actors whose title or name matched that bucket's regex; in search mode one query row (first in the dataset) aggregates the actors with query_match: true. No query row when nothing matched.
hostility	high, med, or low: a hand-labelled guess at how hard the target site fights scrapers, which sets the on-call load. med for query rows.
actors, active_actors	Actors in the bucket (after minUsers30d), and how many had at least one user in 30 days.
users_30d, users_total	Summed over the bucket.
leader, leader_url, leader_share	The most-used actor, its page, and its share of the bucket's 30-day users.
leader_rating, leader_reviews, leader_fail_rate, leader_price, leader_signals	The leader's rating, review count, 30-day fail rate, price label, and gap signals.
weighted_rating	User-weighted mean rating over actors with three or more reviews. null when none qualify.
best_rated, best_rating	The best-rated actor with five or more reviews, and its rating.
broken_actors	Actors with 10+ monthly users failing 40%+ of runs.
signals	The A to G shortlist rules: A_poorly_served (150+ monthly users and weighted rating under 4.0, or the leader fails 30%+, or nothing with five reviews rates 4.2+), B_big_actor_low_rated (an actor with 500+ lifetime users rated under 3.6), C_actor_failing_in_use (an actor with 25+ monthly users and 100+ runs failing 40%+), D_incumbent_idle (an actor with 20+ monthly users under maintenance or idle two weeks), E_thin_market (40+ monthly users and six or fewer active actors), F_low_hostility_demand (low hostility and 60+ monthly users), G_monopoly (300+ monthly users and the leader holds 80%+).
snapshot_at	Same as on actor rows.
delta	users_30d_prev, users_30d_change, actors_prev, leader_prev, is_new, computed from the previous snapshot's actor rows in the same bucket. All null without a snapshot dataset.

Pricing

Pay per event. Free-tier credit covers thousands of rows.

Event	Price	When
actor	$0.001	One actor row.
target	$0.01	One target row (a niche summary, or the query summary).

A search for one niche is typically 20 to 60 actor rows and a handful of target rows: a few cents. A full monthly snapshot at maxActors: 5000 is about $5 plus the target rows.

If maxTotalChargeUsd is set, the actor works out up front how many rows fit: target rows first, then as many actor rows as the remaining budget allows. Rows past the cap are dropped, never charged. A run is never charged for rows it did not return.

Limits and honesty

• Demand is totalUsers30Days from the store; Apify publishes no revenue, so this is a count of people, not dollars. Runs and fail rates are the store's public run statistics.
• Target buckets come from a regex over title and name only (descriptions drag generic actors into every bucket they mention), so a few actors land in the wrong bucket; check the actor rows before acting on a target row. Hostility is a hand-labelled guess.
• The store's search matches single words and descriptions, so google trends also returns every Google actor. Those rows are kept (they are what a buyer browsing the store sees) but ranked after the ones whose title or name carry the whole query, and flagged with query_match. The query target row counts only the matches.
• full mode unions the catalog across the global popularity sort, every category, and every pricing model, because the global sort is unstable past about 12,000 rows. Expect tens of thousands of items and about ten minutes; set maxActors to what you want kept.
• Deltas compare against the newest snapshot in the named dataset, whatever its age; if you run weekly, deltas are weekly. Rows with the same snapshot_at are one snapshot.
• If the store API changes shape (no items list, or an item without stats.totalUsers) the run fails loudly with Source drift detected and nothing is charged, rather than returning empty rows.

Support

Open an issue on the actor page. Changes at the source are fixed within days.