SignalThe numbers
The final run was on 3 November 2026. It covered 170 API paths: 145 that match ScrapeCreators’ paths and 25 that match Outscraper’s core sources (Google Maps, app stores, Amazon, Google News).
| Result | |
|---|---|
| Paths tested live | 170 |
| Passing | 165 of 170 |
| Median response | 1.44 s (ScrapeCreators: 3.96 s) |
| Faster on | 102 of 119 endpoints both APIs answered |
| ScrapeCreators catalog covered | 198 of 205 (97%) |
| Response shape match | 100% of top-level keys on 120 endpoints with a reference sample |
| Repeat call served from cache | 152 of 152 |
How we got to 165
The test harness passed 152 paths on its first pass. We checked each of the 18 misses by hand. Thirteen were problems with the test inputs, not the endpoints, and they pass with explicit inputs. That leaves five real failures.
We don’t count those five as working.
What failed, and what changed
On launch day the failures were in these endpoints:
Facebook group posts, LinkedIn ad details, LinkedIn ad search and TikTok Shop product reviews. The platforms blocked the datacenter IPs we tested from.
Update, 9 November 2026: we added residential IPs as a fallback after our datacenter pool. All four now pass on production, along with Rumble. A request still tries the direct and cheaper routes first, so this doesn’t change prices.
What we chose not to build
Some ScrapeCreators endpoints we skipped on purpose. Creator email lookup and age/gender detection are personal-data enrichment, and we don’t do that. Reddit ads and TikTok popular songs were discontinued by the platforms. TikTok audience data and Instagram trending Reels have no public source.
That’s why coverage is 198 of 205 and not 205.
How we measured speed
We sent the same request to both APIs, end to end, one call per endpoint, from one client location. On the 119 endpoints where both returned data, our median was 1.44 s and ScrapeCreators’ was 3.96 s. We were faster on 102 of them.
That leaves 17 endpoints where we’re slower, and that’s our to-do list. One run from one location is a snapshot, not a guarantee: your numbers will vary by endpoint, region and payload size.
Two things we learned
Speed is mostly about what you don’t fetch. A warm store that keeps popular profiles fresh did more for the median than any proxy change. When the data is already there, the request is a database read.
Compatibility is the real feature. Matching another API’s paths and JSON exactly means switching is a one-line base-URL change. That makes it cheap for people to try us, which matters more than any single endpoint.
Try it
You get 1,000 free credits, no card. Point your existing ScrapeCreators client at SignalShelf, or start in the playground. If an endpoint breaks for you, tell us at support@signalshelf.dev.