We Analyzed 27 Million API Requests: What Social Data Developers Actually Build On
There's a lot of research about social platforms and almost none about the people building on top of them. That gap exists for a boring reason: the data sits inside private billing systems.
We happen to run one of those. So we pulled twelve months of our own aggregate usage telemetry, 27,083,773 API requests from 4,499 developer accounts, to answer a question nobody seems to publish on: when developers pay per request for public social data, what do they actually ask for?
Every request in that dataset cost someone a credit. That makes it a more honest signal than a survey, where ranking a platform highly is free. Here's what surprised us.
The most popular platform isn't the most-used one
This is the finding that reframed the whole study for us.
More developers touched TikTok than any other platform: 2,271 of 4,499 accounts, just over half. But Instagram absorbed nearly twice the request volume, 35.2% of everything against TikTok's 18.2%.
Put per developer, the gap is stark:
| Platform | Share of requests | Developers | Requests per developer |
|---|---|---|---|
| 35.2% | 1,861 | 5,127 | |
| TikTok | 18.2% | 2,271 | 2,174 |
| Twitter/X | 13.2% | 884 | 4,033 |
| 10.3% | 671 | 4,162 | |
| 9.6% | 506 | 5,156 | |
| YouTube | 3.7% | 721 | 1,400 |
TikTok is where teams start. Instagram is where they commit. If you're deciding which platform to support first, those are two different questions with two different answers, and conflating them is how you end up building for breadth when your users needed depth.
The clearest illustration is the two profile endpoints. TikTok profile lookups came from 1,204 developers; Instagram profile lookups came from 1,216, basically the same audience. Instagram's endpoint got four times the volume. Same operation, wildly different depth of use.
At the other end, Threads and Pinterest show the inverse pattern: 226 and 133 developers respectively, but only 497 and 407 requests each. Plenty of teams tried them. Almost nobody built on them.
Identity is the primitive, content is the bill
Grouping every endpoint by what it's for separates adoption from consumption, and the two don't line up.
| Use case | Share of requests | Developers |
|---|---|---|
| Posts and content | 44.3% | 2,264 |
| Profiles and identity | 22.6% | 2,299 |
| Search and discovery | 16.1% | 1,716 |
| Comments and conversation | 9.2% | 616 |
| Ad intelligence | 2.5% | 243 |
| Commerce | 2.2% | 249 |
| Trends | 1.7% | 345 |
| Transcripts | 1.3% | 319 |
| Audience and followers | 0.1% | 354 |
Profile lookups reached more developers than anything else, 2,299 accounts, just over half the dataset. It's the atomic operation of this work: before you can rank, dedupe, analyze, or contact anyone, you resolve who they are. The single most-requested endpoint in the entire study is Instagram profile, at 3,511,306 calls.
But content fetching is where the money goes. 44.3% of all volume, and the highest intensity of any category at 5,301 requests per developer. The reason is structural: identity is one call per entity, content is a call per page, per account, forever. If you're costing out a social data pipeline, that's your line item, and it's why API cost optimization tends to be about pagination rather than clever caching of profiles.
The feature everyone tries and nobody keeps
Look at the bottom row of that table again. 354 developers made at least one follower or audience request. The category drew 87 requests per developer, the lowest intensity we measured by a wide margin.
That's the signature of a feature people sample once and quietly abandon. It reads as essential in a pitch deck ("we'll analyze their whole audience") and behaves as optional in production, because enumerating a follower list is expensive, slow, and usually answerable with a smaller sample.
If you're planning a roadmap around social data, that one line is probably worth more than the rest of the study.
Reddit is the quiet heavyweight
Only 506 developers used Reddit, one of the smaller groups for a major platform. Yet it posted the highest request intensity in the study, roughly 5,156 per developer, and its share of total demand nearly doubled over the year, from 6.5% to 13.3%.
What makes it interesting is the endpoint mix. Reddit users in this dataset pull subreddit listings (1,010,531 requests) and search (643,808), not profiles. They're monitoring topics and communities, not people. That's a completely different job from the creator-centric pattern on Instagram and TikTok: market research and product validation rather than influencer discovery, which lines up with how teams actually use Reddit to find real pain points.
Few teams build on Reddit. The ones that do build seriously.
Public data is more reliable than its reputation
Across all 27.1 million requests, 574,104 came back with an error status. That's a 97.88% success rate, or about one failure in every 47 calls.
The honest caveat: that measures request completion, not usefulness. A call that succeeds and returns an empty result counts as a success, and the number reflects our own retry handling rather than raw platform reachability. But it's worth stating, because the assumption in this space is that public data collection is perpetually breaking. Plan for failure as a normal condition, not as constant instability. The mechanics of how these APIs work explains why that number is higher than people expect.
What this study can't tell you
I'd rather front-load this than bury it, because at least one number above is substantially explained by our own failures rather than by developer behavior.
This is one API's customer base, not the market. Everyone here chose a commercial public-data API. Teams using official platform APIs, in-house scrapers, or enterprise vendors are invisible to us. It's a self-selected sample by definition.
Demand is capped by what we offer. We have more Instagram endpoints than Pinterest endpoints, which mechanically raises Instagram's ceiling. These shares reflect available demand, not latent demand.
One trend is our fault, not a signal. LinkedIn's share fell from 11.5% to 0.7% between our early and late windows. Our LinkedIn endpoints had real reliability problems during that period, to the point that we temporarily pulled one from our own documentation while its upstream source was failing. Requests followed the reliability, not the interest. Read that row as a supply story.
That last one turned out to be the most useful lesson in the dataset, and it's not about platforms at all: uptime is product-market fit. When an endpoint got unreliable, demand didn't wait around.
Per-developer figures are means, not medians. They're sensitive to a handful of heavy users, and we didn't publish the distribution. Treat them as directional.
The full report
The complete study has the per-endpoint tables, the month-by-month series, the full share-shift analysis, and the methodology, including the exact read-only SQL and the aggregate dataset so you can recompute every percentage yourself.
Read the full report: The State of Social Data Demand 2026
Worth saying plainly: we operate the API we measured, which is an obvious conflict of interest. That's exactly why the underlying aggregates and queries are published. Check the numbers.
Frequently Asked Questions
Which social media platform do developers use most?
It depends on what you mean by "most." More developers used TikTok than any other platform (2,271 of 4,499 accounts), but Instagram absorbed nearly twice the request volume, 35.2% versus 18.2%. TikTok has the broadest adoption; Instagram has the deepest usage.
What social media data do developers request most often?
Content (posts, videos, reels) accounts for 44.3% of all requests, the largest category by volume. But profile and identity lookups reached the most developers, just over half the dataset, making identity resolution the most widely adopted operation.
Is follower data worth building on?
Based on this data, less than people assume. 354 developers made at least one follower or audience request, but averaged only 87 requests each, the lowest intensity of any category. It's widely sampled and rarely operationalized.
Why is Reddit data usage growing?
Reddit's share of requests nearly doubled over the year, from 6.5% to 13.3%, and it had the highest per-developer intensity in the study. Its endpoint mix is topic-first (subreddit listings and search rather than profiles), which suggests market research and community listening rather than creator discovery.
How reliable is public social media data collection?
In this dataset, 97.88% of 27.1 million requests completed successfully, roughly one failure in 47 calls. Note that this measures request completion, not data quality: a successful response that returns no results still counts as a success.
Can I see the underlying data?
Yes. The full report publishes the aggregate dataset and the read-only SQL used to produce it, so every figure can be recomputed independently. It contains no customer identities and no request contents, only platform-, endpoint-, and month-level aggregates.
Want to build with the same endpoints in this study? Start free with 50 credits, no card required. If you'd rather compare options first, our rundown of the best social media data APIs is a fair place to begin.
Found this helpful?
Share it with others who might benefit
Ready to Try SociaVault?
Start extracting social media data with our powerful API. No credit card required.