What a scraping API will really cost you

Every scraping API publishes a number that looks like the answer: $1.88 per 1,000 calls, $1.50 per 1,000 records, $0.001 a request. None of them is what you’ll pay for the data you actually use. That depends on four things the headline hides: what counts as one unit, what happens when a request fails, whether unused credit disappears, and what’s billed on top.

This guide goes through each, then works an example from list price to cost per useful record.

What you’re actually billed for

Before comparing prices, find out what one unit buys. The providers in our Instagram and TikTok rankings use six different models:

Weighted credits need the most care. ScrapeCreators charges one credit for most calls, which is where its $1.88 per 1,000 ($0.99 on the $497 pack) comes from. But Instagram comments with replies cost 15 credits, download_media costs 10, and the TikTok audience endpoint costs 26, which makes that last one $48.88 per 1,000 profiles on the $47 pack. EnsembleData does the same with units: an Instagram profile costs 3, a detailed one 10, and every Instagram post 1.

Price the result, not the call

A cheap call that returns one item can cost more than an expensive call that returns fifty. Work out the cost of 1,000 of the thing you need, using each provider’s own page sizes:

• Comments. A HikerAPI comments call returns 15 comments, so 1,000 comments cost about $0.07 at $1 per 1,000 requests. EnsembleData’s TikTok comments cost one unit per page of 30, about $0.08 per 1,000 on its entry plan. Bright Data charges per record, so 1,000 comments cost $1.50, the same as 1,000 profiles.
• Recent posts. An Influship Instagram profile call returns up to 50 recent posts for $0.0018, so recent posts across 1,000 creators cost $1.80. On EnsembleData the same job is 3 units for the profile plus 1 per post.
• Followers. EnsembleData charges 2 units per 100 TikTok followers, about $0.04 per 1,000 on its entry plan. Per-record pricing would make the same list cost dozens of times more.

Where a provider doesn’t publish its page size, test it. A per-page price is only as good as the number of items on the page, and a page can come back short or empty.

Check what failures cost

This is the line item most buyers miss. Every scraping job hits accounts that have been deleted, renamed or made private, plus some temporary errors. Whether you pay for those depends entirely on the provider:

HikerAPI and LamaTok are the ones to watch. Both bill 403 and 404 responses, so a list full of dead or private accounts costs the same as a list of live ones. Bright Data markets “pay only for success”, but its docs FAQ says records that fail because the input was incorrect are still billed. TikHub doesn’t say, on its pricing page or in its terms.

At the other end, Influship doesn’t charge for any 4xx or 5xx response, and every response shows what it cost. EnsembleData doesn’t charge for its own internal errors. ScrapeCreators doesn’t charge for many “confirmed unavailable” profiles, and since 13 October 2026 empty TikTok comment pages are free.

Then there are successes you can’t use. TikTok sometimes returns HTTP 200 with an empty body instead of an error, and a provider may pass that on as a billable success. ScrapeCreators itself says to expect about 90% success on Instagram comments. On X, SocialData and twitterapi.io both charge for some empty results, so a monitoring job pays for “nothing found”.

Credit that expires

Prepaid credit is only worth its headline price if you use it before it disappears.

• Never expires: ScrapeCreators credits, and HikerAPI and LamaTok balances. Influship bills after use, so there’s nothing to expire, and its 500 free starter credits don’t expire either.
• Resets daily: EnsembleData’s units reset at 00:00 UTC and don’t roll over. The $100 Wood plan is 1,500 units a day, not 45,000 a month to spend when you like.
• Resets monthly: Apify’s plan credit doesn’t roll over, and Data365 is a monthly subscription.

The maths is simple and unforgiving. EnsembleData’s Wood plan works out at $2.22 per 1,000 units if you use every unit every day. If your job averages 500 units a day, you’re paying $6.67 per 1,000 units you actually use. Daily allowances suit steady monitoring; spiky or one-off work belongs on credit that never expires or on billing after use.

Minimums, entry tiers and add-ons

What it costs to start

• No minimum: Influship (billed after use, 500 free credits with no card) and Bright Data pay as you go (5,000 free records a month).
• Small top-ups: TikHub takes a $5 minimum, but all sales are final. ScrapeCreators’ cheapest pack is $47; the $10 pack it sold in 2025 has gone.
• Steep entry tiers: LamaTok’s $20 Start tier charges $0.02 a request, $20 per 1,000, while a $50 deposit drops that to $2.50 and $300 to $0.60. Buying the smallest pack is the most expensive way to use it.
• Subscriptions: Apify from $19 a month, EnsembleData from $100, Data365 from €300 for one network. Bright Data’s pre-collected datasets have a $250 minimum order.

What’s billed on top

Apify is the clearest case. Its Actors charge per result, but many add other events: $0.001 per run start on some TikTok and Instagram Actors, transcripts at $0.041 a minute on the Starter plan, and $0.0013 per filter on Clockworks’ TikTok scraper. Which Actor you pick matters even more: Apify’s own Instagram Scraper costs $2.30 per 1,000 results on Starter, while API Dojo’s third-party one costs $0.49.

Elsewhere, extras tend to be priced per call: ScrapeCreators, for example, charges 10 credits per started minute for opt-in AI transcripts over two minutes.

A worked example

Say you need to enrich 10,000 Instagram handles from a creator list. The rates below are illustrative, not measured: we haven’t tested any provider, and your list will behave differently. They’re there to show how billing rules change the result.

• 10% of handles no longer exist or are private, so 9,000 are live.
• 5% of lookups on live handles hit a temporary server error and succeed on one retry: 450 extra requests, 10,450 in all.
• 4% of successful responses come back missing what you need, leaving 8,640 useful records.

At the same list price of $1.00 per 1,000 calls, here’s what three billing rules cost per 1,000 useful records:

With 10% dead handles, a provider that bills 404s ends up 16% above its list price per useful record. On an older list where 30% of handles are dead, that becomes 49%. If the provider only bills successful responses, you pay 4% over list either way, however stale your list is; the only overhead left is empty successes. Clean your input list first if your provider bills 404s.

Estimating your own cost

1. Take 200 to 500 real inputs from your own data, including old and small accounts. A clean sample flatters every provider.
2. Run them through two or three free tiers. ScrapeCreators gives 100 credits, HikerAPI and LamaTok 100 requests, Influship 500 credits, EnsembleData 50 units a day, Bright Data 5,000 records a month and Apify $5 a month.
3. Log every request: status code, what you were charged, and whether the response had the fields you need.
4. Divide what you spent by the number of useful records, then multiply by your monthly volume, allowing for retries and refreshes.
5. If the plan has a daily or monthly allowance, check you’d use all of it. If not, divide by what you’d actually use.

# One row per request you sent during a trial run.
# "charged" comes from the provider's response or usage dashboard:
# ScrapeCreators returns credits_charged, EnsembleData's SDK returns
# units_charged, and Influship shows the cost on every response.
spent  = sum(r["charged"] for r in rows) * PRICE_PER_UNIT
useful = sum(1 for r in rows if r["ok"] and r["has_what_you_need"])

print(f"Cost per 1,000 useful records: ${spent / useful * 1000:.2f}")

Where to start

Our Instagram ranking and TikTok ranking list every provider’s billing unit, minimum and what it charges for failures. For Instagram, the top of our ranking:

And for TikTok:

If you’re still deciding whether to pay for an API at all, read our guide to building your own scraper.

Questions