Nava / Benchmarks
Four tools got the same 12 TikTok briefs, written from real buyer requests. One judge graded every pick without knowing which tool found it. Nava found 30 creators who are the customer and still get watched; the next best, an AI agent working alone, found 7. It still missed the bar we set in advance.
Updated 2026-11-05 · Data: benchmarks.json · benchmarks.md
12 TikTok briefs · top 5 per tool · higher is better
Qualified: the judge says their own bio or posts show every must-have, they posted in the last 30 days, and their recent views are at least 0.3x their followers.
On 4 briefs Nava found no one: certified personal trainers who train clients; small business owners who answer their own phone calls; ER nurses who work nights in Boise, Idaho; people laid off in the last 60 days who are job hunting. Neither database did either; the AI agent alone found one creator on two of them.
Picks the judge called exact, and how many of them still post and get watched
The databases mostly return the right kind of person who has stopped posting or no longer gets views.
All-in, including the buyer's time to check picks · lower is better
Tool spend alone, all 12 briefs: Nava $22.75, AI agent alone $4.67, Influencers.club $3.39, Exa $0.32. Nava spends the most on data and AI. The rest of the all-in cost is the buyer's time checking picks by hand, for tools that don't vet.
Qualified creators in each tool's top 5, out of 5.
Brief 1
Nava won vs databases · won vs agent
Brief 2
No win vs databases · no win vs agent
Brief 3
Nava won vs databases · won vs agent
Brief 4
Nava won vs databases · won vs agent
Brief 5
Nava won vs databases · won vs agent
Brief 6
Nava won vs databases · won vs agent
Brief 7
No win vs databases · no win vs agent
Brief 8
Nava won vs databases · won vs agent
Brief 9
Nava won vs databases · won vs agent
Brief 10
Control: almost nobody fits, tests padding
No win vs databases · no win vs agent
Brief 11
Control: recent event, tests stale indexes
No win vs databases · no win vs agent
Brief 12
Control: the databases' home turf, Nava should lose
Nava won vs databases · won vs agent
Pre-registered on 2026-11-05, before any of the 12 briefs was run (one dry run on a practice brief came first and doesn't count). Nava wins a brief if its precision is at least the best of Influencers.club and Exa, and its cost per qualified creator is a third of that tool's or less (or it finds qualified creators and that tool finds none). The headline claim needs wins on at least 10 of the 12 briefs, including a control Nava is expected to lose. Against the AI agent alone, Nava must find more qualified creators at no more cost on at least 10 of 12; if not, search alone is not a moat. Each tool runs twice; a gap no bigger than a tool's own run-to-run swing is a tie. If Nava wins fewer than 7 of 12 against the databases, Nava stops claiming better matches.
Limits| # | Who | Must have | Shows up in | How many exist |
|---|---|---|---|---|
| 1 | dog or cat owners whose pets are in their videos | owns a dog or cat; the pet appears in their own videos | bio | many |
| 2 | certified personal trainers who train clients | is a certified personal trainer; trains clients now | bio | many |
| 3 | dermatologists or licensed estheticians | is a board-certified dermatologist or a licensed esthetician; practices now | bio | few |
| 4 | registered dietitians | is a registered dietitian (RD/RDN) | bio | few |
| 5 | people paying off debt and sharing their progress | is paying off their own debt; shares their own numbers or progress | posts | many |
| 6 | people aged about 25 to 50 who lift weights regularly | appears to be about 25 to 50; lifts weights regularly, shown in their own videos | posts | many |
| 7 | small business owners who answer their own phone calls | owns and runs a small business; handles customer phone calls themselves | posts | many |
| 8 | parents of kids with food allergies | is a parent; their own child has a food allergy | posts | few |
| 9 | people who talk openly about living with anxiety and going to therapy | says themselves that they live with anxiety; talks about their own therapy | posts | few |
| 10 | ER nurses who work nights in Boise, Idaho Control: almost nobody fits, tests padding | works as an ER nurse; works night shifts; lives or works in or near Boise, Idaho | posts | very few |
| 11 | people laid off in the last 60 days who are job hunting Control: recent event, tests stale indexes | was laid off within about the last 60 days; is looking for a job now | posts | many |
| 12 | US fitness creators with 100K to 500K followers Control: the databases' home turf, Nava should lose | makes fitness content; is based in the US | bio | many |
Brief 9 counts only creators who say it about themselves; no one is included on a guess about their health. No minors in any brief.
| Brief | Tool | Exact in top 5 | Qualified | Spend | With checking | Cost per qualified |
|---|---|---|---|---|---|---|
| 1 | Nava | 5/5 | 5 | $1.13 | $1.13 | $0.23 |
| 1 | Influencers.club | 3/5 | 1 | $0.30 | $4.05 | $4.05 |
| 1 | Exa | 4/5 | 0 | $0.03 | $3.78 | none found |
| 1 | AI agent alone | 4/5 | 2 | $0.32 | $4.07 | $2.03 |
| 2 | Nava | 0/5 | 0 | $2.21 | $2.21 | none found |
| 2 | Influencers.club | 0/5 | 0 | $0.30 | $4.05 | none found |
| 2 | Exa | 2/5 | 0 | $0.03 | $3.78 | none found |
| 2 | AI agent alone | 0/5 | 0 | $0.37 | $4.12 | none found |
| 3 | Nava | 4/5 | 4 | $0.91 | $0.91 | $0.23 |
| 3 | Influencers.club | 5/5 | 0 | $0.30 | $4.05 | none found |
| 3 | Exa | 4/5 | 0 | $0.03 | $3.78 | none found |
| 3 | AI agent alone | 4/5 | 0 | $0.34 | $4.09 | none found |
| 4 | Nava | 5/5 | 5 | $0.68 | $0.68 | $0.14 |
| 4 | Influencers.club | 4/5 | 1 | $0.30 | $4.05 | $4.05 |
| 4 | Exa | 3/5 | 0 | $0.03 | $3.78 | none found |
| 4 | AI agent alone | 3/5 | 0 | $0.26 | $4.01 | none found |
| 5 | Nava | 5/5 | 5 | $0.71 | $0.71 | $0.14 |
| 5 | Influencers.club | 4/5 | 0 | $0.30 | $4.05 | none found |
| 5 | Exa | 3/5 | 1 | $0.03 | $3.78 | $3.78 |
| 5 | AI agent alone | 4/5 | 1 | $0.43 | $4.18 | $4.18 |
| 6 | Nava | 2/5 | 2 | $1.06 | $1.06 | $0.53 |
| 6 | Influencers.club | 1/5 | 0 | $0.30 | $4.05 | none found |
| 6 | Exa | 1/5 | 0 | $0.03 | $3.78 | none found |
| 6 | AI agent alone | 1/5 | 1 | $0.32 | $4.07 | $4.07 |
| 7 | Nava | 0/5 | 0 | $4.00 | $4.00 | none found |
| 7 | Influencers.club | 0/5 | 0 | $0.30 | $4.05 | none found |
| 7 | Exa | 0/5 | 0 | $0.03 | $3.78 | none found |
| 7 | AI agent alone | 0/5 | 0 | $0.34 | $4.09 | none found |
| 8 | Nava | 5/5 | 5 | $1.35 | $1.35 | $0.27 |
| 8 | Influencers.club | 1/5 | 0 | $0.30 | $4.05 | none found |
| 8 | Exa | 1/5 | 0 | $0.03 | $3.78 | none found |
| 8 | AI agent alone | 3/5 | 0 | $0.37 | $4.12 | none found |
| 9 | Nava | 3/5 | 3 | $3.24 | $3.24 | $1.08 |
| 9 | Influencers.club | 0/5 | 0 | $0.30 | $4.05 | none found |
| 9 | Exa | 0/5 | 0 | $0.03 | $3.78 | none found |
| 9 | AI agent alone | 1/5 | 1 | $0.37 | $4.12 | $4.12 |
| 10 | Nava | 0/5 | 0 | $3.24 | $3.24 | none found |
| 10 | Influencers.club | 0/5 | 0 | $0.09 | $2.34 | none found |
| 10 | Exa | 0/5 | 0 | $0.03 | $3.78 | none found |
| 10 | AI agent alone | 1/5 | 1 | $0.57 | $4.32 | $4.32 |
| 11 | Nava | 0/5 | 0 | $2.97 | $2.97 | none found |
| 11 | Influencers.club | 0/5 | 0 | $0.30 | $4.05 | none found |
| 11 | Exa | 1/5 | 0 | $0.03 | $3.78 | none found |
| 11 | AI agent alone | 4/5 | 1 | $0.42 | $4.17 | $4.17 |
| 12 | Nava | 1/5 | 1 | $1.25 | $1.25 | $1.25 |
| 12 | Influencers.club | 1/5 | 0 | $0.30 | $4.05 | none found |
| 12 | Exa | 1/5 | 0 | $0.03 | $3.78 | none found |
| 12 | AI agent alone | 0/5 | 0 | $0.57 | $4.32 | none found |
Start from who they are, not from follower counts. Look for creators whose own bio or posts say it: a nurse who posts about her night shifts, an account executive who films his prospecting block. Then check that they still post (the last 30 days) and still get watched (typical views on recent posts at least 0.3x their followers), and read the profile yourself before you reach out. Nava does this for you: tell it who your customer is and it sends the creators who fit, each checked against their own bio and posts.
In this benchmark, Nava: 30 qualified creators across the top 5s of 12 TikTok briefs, against 7 for an AI agent working alone with the same data, 2 for Influencers.club and 1 for Exa. Nava won 8 of 12 briefs against the databases and 8 of 12 against the agent. The bar set before the run was 10 of 12, so it was not met.
Most of their picks are the right kind of person but have stopped posting or no longer get watched. The judge called 19 of Influencers.club's top picks and 20 of Exa's exact, but only 2 and 1 of those passed the activity and views check.
On 4 briefs Nava found no qualified creator: certified personal trainers who train clients; small business owners who answer their own phone calls; ER nurses who work nights in Boise, Idaho; people laid off in the last 60 days who are job hunting. Neither database found one either. The AI agent alone found one each on two of them, the near-empty Boise control and recent layoffs.
Nava searches TikTok for the brief, then reads each creator's own bio and recent posts and grades them against every must-have: exact, close or no. It keeps only creators who posted in the last 30 days with typical views at least 0.3x followers. In two earlier blind checks (ER nurses who work nights; people who sell for a living), every creator Nava's grader called exact, 13 of 13, was also called exact by a person. That is a small sample, so Nava says "checked against their own bio and posts", not "verified".
Four tools (Nava, Influencers.club, Exa and an AI agent working alone with the same TikTok data) got the same 12 briefs on the same day. Each tool's top 5 per brief were pooled, shuffled and graded by one judge, Jev (TypeSafe), a model none of the tools uses, without knowing which tool found each creator. Every creator got the same activity and views check. The bar for a win was fixed before the briefs were run.
To test whether Nava adds anything an agent can't do itself with the same raw TikTok search, profile and video tools. The bar was 10 of 12 briefs; Nava won 8, so by the rule set in advance, search alone isn't yet a moat.
Yes, under CC BY 4.0. Cite it as: Nava (2026). Nava UGC creator sourcing benchmark, run 1. https://meetnava.com/benchmarks.json. The JSON file has every brief, definition, the bar and every result row.
Yes. Nava has an MCP server at https://meetnava.com/mcp (no key) and the same tools as REST, described at https://meetnava.com/openapi.json. find_creators returns creators Nava already checked at once and queues a fresh graded search.
Nava (2026). Nava UGC creator sourcing benchmark, run 1. https://meetnava.com/benchmarks.json
License CC BY 4.0. Machine-readable: benchmarks.json (every brief, definition and result) · benchmarks.md · llms.txt