Social data API benchmark: SignalShelf vs ScrapeCreators

Results

Run finished 2026-11-03 13:17 UTC. Medians use only endpoints where both APIs returned HTTP 200.

Method

Date: 3 November 2026. ScrapeCreators reference calls were made first (between about 10:48 and 12:26 UTC); our run finished at 13:17 UTC. Calls were sequential, not interleaved.

Client: one machine in India for both sides. Each endpoint was called once per API (n = 1).

SignalShelf side: our production app code booted on that machine with a fresh, empty database (no warm store, no cache), so every first call did real work: auth, validation, live fetch from the source, billing. Outbound traffic went through a datacenter proxy pool or direct HTTP.

ScrapeCreators side: their public API (api.scrapecreators.com) with our own key.

Inputs: one fixed public input per endpoint, for example handle=stoolpresidente for /v1/tiktok/profile. Where ScrapeCreators returned data, our call reused the exact same parameters.

Time: client wall-clock from sending the request to receiving the full JSON body.

“Faster” means our time was lower than theirs on that endpoint. Medians and the faster count only include endpoints where both APIs returned HTTP 200.

Shape match: the share of ScrapeCreators’ top-level response keys that also appear in ours (wrapper keys like success and credits_remaining excluded).

Each passing path was called a second time to confirm caching. Those repeat calls are excluded from all timings.

Caveats

Point in time: one run, one day. Latency on both sides varies by hour, input and source.

One client location (India) and one call per endpoint. This is not a load test.

Our app ran on the test machine itself, so our times don’t include the network hop to a hosted API that ScrapeCreators’ times do. A hosted deployment adds that hop.

We don’t know whether ScrapeCreators served any response from its own cache.

Shape match checks top-level keys, not every nested field or value.

Only ScrapeCreators was speed-tested. We have not timed Outscraper, Apify, Bright Data, EnsembleData, Data365, SerpApi or any other provider, and make no speed claims about them.

Check it yourself

Raw data: every row below, as JSON, at /compare/benchmark.json (internal cost fields removed; it contains no keys).

Script: scripts/smoke-guest-e2e.mjs in the SignalShelf codebase boots the real app, replays every path, and writes docs/guest/E2E_RESULTS.json, which this page is built from.

Without our code: time any endpoint on both APIs with your own keys, as below. Use 1,000 free credits + 5,000 to earn on our side.

P="/v1/tiktok/profile?handle=stoolpresidente"
curl -s -o /dev/null -w "SignalShelf    %{time_total}s\n" "$SIGNALSHELF_BASE_URL$P" -H "x-api-key: $SIGNALSHELF_API_KEY"
curl -s -o /dev/null -w "ScrapeCreators %{time_total}s\n" "https://api.scrapecreators.com$P" -H "x-api-key: $SCRAPECREATORS_API_KEY"

Every endpoint

One row per path. "—" means that side had no successful reference call to time. Error codes on our side are mostly test-harness inputs (for example a missing required parameter), kept here unedited.