Almost every study of the social ecosystem measures platforms. This one measures the people building on top of them — 27 million API requests from 4,499 developers, and what they reveal about the public social data teams actually pay for.
More developers touched TikTok (2,271) than Instagram (1,861), but Instagram absorbed nearly twice the request volume — 35.2% against 18.2%. TikTok is where teams start; Instagram is where they go deep, at 5,127 requests per developer versus 2,174.
Profile lookups reached more developers than any other category (2,299, just over half the dataset), and Instagram profile is the single most-requested endpoint. But posts and content consumed 44.3% of all volume. One call resolves an identity; content is a call per page, forever.
Just 506 developers used it, yet it recorded the highest request intensity in the study at roughly 5,156 per developer. Its endpoint mix is topic-first rather than person-first — market research and community listening, not influencer discovery.
Comparing February–April 2026 with July–August 2026, Instagram's share of all requests rose from 25.4% to 40.4%, a gain of 15.0 percentage points and the largest shift measured. Facebook added 6.1 points.
From 6.5% to 13.3% across the same windows, a gain of 6.8 percentage points — the second-largest of any platform.
354 developers made at least one audience or follower request, but the category drew only 87 requests per developer — the lowest intensity by a wide margin. It reads as essential and behaves as optional.
This study analyses first-party, aggregated, anonymised usage telemetry from the SociaVault API. Every figure describes what developers requested. Nothing here measures any social platform, its users, or its content.
| Total requests analysed | 27,083,773 |
| Distinct developer accounts | 4,499 |
| Observation window | 30 Sep 2025 – 7 Sep 2026 |
| Platforms with recorded demand | 17 (up from 3) |
| Overall success rate | 97.88% |
| Credits consumed | 27,237,568 |
September 2026 covers 1–7 September only and is never compared like-for-like against a full month. September–November 2025 was early access with a handful of accounts and is excluded from trend interpretation.
The interesting column here is the last one. TikTok reached the most developers but is used comparatively lightly. Instagram has fewer developers and more than twice the intensity. Threads and Pinterest show the opposite pattern to both: widely sampled, barely built on.
| Platform | Requests | Share | Developers | Req / dev |
|---|---|---|---|---|
| 9,541,375 | 35.2% | 1,861 | 5,127 | |
| TikTok | 4,937,497 | 18.2% | 2,271 | 2,174 |
| Twitter / X | 3,565,294 | 13.2% | 884 | 4,033 |
| 2,792,561 | 10.3% | 671 | 4,162 | |
| 2,609,117 | 9.6% | 506 | 5,156 | |
| YouTube | 1,009,119 | 3.7% | 721 | 1,400 |
| 834,554 | 3.1% | 277 | 3,013 | |
| TikTok Shop | 621,401 | 2.3% | 435 | 1,429 |
| Facebook Ad Library | 432,670 | 1.6% | 205 | 2,111 |
| Google Ad Library | 225,169 | 0.8% | 71 | 3,171 |
| Google Search | 193,640 | 0.7% | 258 | 751 |
| Facebook Marketplace | 137,351 | 0.5% | 107 | 1,284 |
| Threads | 112,386 | 0.4% | 226 | 497 |
| 54,128 | 0.2% | 133 | 407 | |
| LinkedIn Ad Library | 15,345 | 0.1% | 56 | 274 |
| TikTok Ad Library | 2,163 | <0.1% | 42 | 52 |
| Twitch | 149 | <0.1% | 21 | 7 |
Shares are of 27,083,773 total requests. Developer counts overlap — most accounts use more than one platform. Green marks the highest developer count; amber marks the highest intensity.
Grouping every endpoint into use-case categories separates adoption from consumption. Identity reached the most developers; content consumed the most volume.
| Category | Requests | Share | Developers | Req / dev |
|---|---|---|---|---|
Posts & content Default category — read as an upper bound | 12,002,389 | 44.3% | 2,264 | 5,301 |
Profiles & identity Reached the most developers | 6,123,284 | 22.6% | 2,299 | 2,663 |
Search & discovery | 4,350,269 | 16.1% | 1,716 | 2,535 |
Comments & conversation Small group, heavy usage | 2,500,766 | 9.2% | 616 | 4,060 |
Ad intelligence | 675,352 | 2.5% | 243 | 2,779 |
Commerce | 606,305 | 2.2% | 249 | 2,435 |
Trends | 447,793 | 1.7% | 345 | 1,298 |
Transcripts | 349,792 | 1.3% | 319 | 1,097 |
Audience & followers Widely sampled, rarely kept | 30,660 | 0.1% | 354 | 87 |
Shares are of 27,086,610 requests, the total from the category query, which ran minutes after the volume query. The 0.01% discrepancy is documented in the methodology rather than reconciled artificially.
Note the contrast between the two profile endpoints. TikTok profile was requested by 1,204 developers, Instagram profile by 1,216 — nearly the same audience, four times the volume. Same operation, wildly different depth of use.
| Endpoint | Requests | Developers |
|---|---|---|
| Instagram profile | 3,511,306 | 1,216 |
| Instagram post info | 2,534,935 | 494 |
| Twitter/X search | 1,578,630 | 350 |
| Instagram comments | 1,373,282 | 276 |
| Facebook profile posts | 1,282,859 | 316 |
| TikTok video info | 1,061,427 | 381 |
| Instagram reels | 1,059,916 | 322 |
| TikTok videos | 1,036,563 | 572 |
| Reddit subreddit posts | 1,010,531 | 246 |
| Instagram posts | 946,646 | 777 |
| TikTok profile | 891,977 | 1,204 |
| Facebook post | 881,424 | 199 |
| Twitter/X profile | 868,781 | 480 |
| Reddit search | 643,808 | 366 |
| TikTok keyword search | 543,938 | 544 |
Reddit accounts for 9.6% of requests from just 506 developers — 11% of the dataset — at roughly 5,156 requests per developer, the highest figure we measured. Its share also nearly doubled over the year, from 6.5% to 13.3%.
The endpoint pattern is distinctive: subreddit listings (1,010,531 requests) plus search (643,808), rather than profile lookups. Reddit users in this dataset are monitoring communities and topics, not people. That is a different shape of work from the creator-centric pattern on Instagram and TikTok, and it points at a different job to be done — market research, product validation, and community listening rather than influencer discovery.
Caveat: the per-developer figure is a mean and we did not publish the distribution, so a small number of heavy monitoring operations could be driving it. The combination of rising share, high intensity, and a topic-first endpoint mix is consistent across three independent cuts of the data, but the intensity number alone should be read carefully.
Two multi-month windows, compared on share rather than absolute volume — total requests grew roughly 3.3× between them, so absolute counts rose almost everywhere. A platform can grow in requests while losing share, and several did.
| Platform | Feb–Apr 2026 | Jul–Aug 2026 | Change |
|---|---|---|---|
| 25.4% | 40.4% | +15.0pp | |
| 6.5% | 13.3% | +6.8pp | |
| 4.7% | 10.8% | +6.1pp | |
| Google Ad Library | 0.1% | 1.4% | +1.3pp |
| Google Search | 0.3% | 1.0% | +0.7pp |
| Facebook Marketplace | 0.0% | 0.7% | +0.7pp |
| 0.6% | 0.1% | -0.5pp | |
| Twitter / X | 13.7% | 11.6% | -2.1pp |
| YouTube | 6.2% | 2.7% | -3.5pp |
| TikTok Shop | 4.8% | 1.1% | -3.7pp |
| Facebook Ad Library | 5.2% | 0.5% | -4.7pp |
| TikTok | 20.5% | 15.4% | -5.1pp |
| LinkedInsupply artefact | 11.5% | 0.7% | -10.8pp |
Early window: 4,570,426 requests. Late window: 15,020,019 requests.
This is the finding we most want to flag rather than bury. LinkedIn endpoints had real availability and reliability problems during the later window — at one point we temporarily removed a LinkedIn endpoint from our own documentation navigation because its data source was failing. A platform that is unreliable gets fewer requests regardless of how much developers want it. Read that row as a supply story, and as a reminder that any usage-based dataset measures availability multiplied by desire. Smaller declines, such as TikTok Shop and Facebook Ad Library, may carry similar effects we have not isolated.
| Month | Requests | Active developers | Platforms in use |
|---|---|---|---|
| Oct 2025 | 31,559 | 6 | 7 |
| Dec 2025 | 121,239 | 71 | 11 |
| Feb 2026 | 1,346,219 | 440 | 14 |
| Apr 2026 | 1,774,350 | 842 | 14 |
| Jun 2026 | 3,126,008 | 1,174 | 17 |
| Aug 2026 | 8,570,864 | 1,094 | 17 |
Selected months shown; September 2026 omitted as partial. Full monthly series is in the published dataset.
Platforms drawing real demand grew from 3 to 17 across the window, and that count kept climbing even as request volume concentrated on Instagram. The head got heavier and the tail got longer at the same time. For anyone building here, that is the operationally relevant finding: demand is not consolidating onto two or three platforms. Teams increasingly expect one integration to cover a widening set of surfaces, including commerce and advertising archives that barely registered a year earlier.
Across 27,083,773 requests, 574,104 returned a status of 400 or above — a 97.88% success rate, or roughly 1 failure in every 47 requests.
That deserves a qualification: this measures request completion, not data usefulness. A response that succeeds and returns an empty result counts as a success, and the figure reflects our own retry and error handling rather than raw platform reachability. Still, it is worth stating, because the prevailing assumption about public social data is that it is fragile and constantly breaking. Plan for failure as a normal condition, but not for constant instability.
They are not the same bet. TikTok gets you the broadest share of developer interest; Instagram absorbs the most volume once teams commit. The volume platform is not the popular one.
Half your users will need profile lookups, and it is usually the first thing they integrate. Make it fast, cheap, and dependable before optimising anything else.
Content fetching is 44.3% of volume and the highest-intensity category. Pagination behaviour and per-page cost will dominate your users' bills, so that is where efficiency work pays off.
354 developers tried it; the category drew 87 requests each. It reads as essential and behaves as optional.
Low headcount, high intensity, topic-first rather than person-first. If you serve them, they will use you heavily.
The LinkedIn story in this dataset is the cleanest evidence we have: when an endpoint became unreliable, requests followed the reliability, not the interest. Uptime is product-market fit.
Ordered by how much they should change your reading of the results. This section is not boilerplate — at least one headline trend above is substantially explained by our own supply problems rather than developer behaviour.
Everyone in this dataset chose a commercial public-data API. Teams using official platform APIs, in-house scrapers, or enterprise vendors are invisible here. The sample is self-selected by definition.
We can only observe requests for platforms and endpoints that exist on our service. More Instagram endpoints are available than Pinterest endpoints, which mechanically raises Instagram's ceiling. Shares reflect available demand, not latent demand.
LinkedIn's share fell from 11.5% to 0.7% between windows. Our LinkedIn endpoints had availability and reliability problems during the later window, to the point that one was temporarily hidden from our own documentation while its source was failing. That decline reflects supply at least as much as demand, and should not be read as developers losing interest in LinkedIn data.
They are sensitive to a small number of heavy users, and we did not publish the distribution. Treat them as a rough intensity signal, never as a typical developer's usage.
Monthly active developers grew from single digits to over 1,100. Some share shifts reflect who joined rather than existing users changing behaviour, and aggregate data cannot separate the two.
A response that succeeds and returns an empty result counts as a success. The figure describes request completion, not the quality of the payload.
Use-case categories are a rule-based classification of endpoint names, and 'Posts & content' is the default branch, so it absorbs anything unmatched. Someone else classifying the same endpoints could reasonably produce different buckets.
This study is authored by the operator of the API it measures, which creates two obvious incentives: to make the dataset look large, and to make the platforms we support look important. We mitigate that by publishing the complete aggregate dataset and the exact read-only queries, so every share can be recomputed independently — including limitation 2, which cuts directly against our own framing.
No measurement, characterisation, or evaluation of any social platform, its users, or its content. Every figure describes requests made to our own service by our own customers. Nothing here should be read as a claim about any platform's conduct or quality.
Every percentage in this report can be recomputed from the aggregate dataset behind it. The full aggregates — totals, platform, endpoint, monthly series, and the use-case rollup — are available on request, along with the read-only SQL used to produce them.
For the underlying aggregates, a specific cut of the figures, or methodology questions, contact labs@sociavault.com.
Every endpoint in this report is available on the SociaVault API. Start with 50 free credits, no card required.