> ## Documentation Index
> Fetch the complete documentation index at: https://docs.9pic.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Test Photo Search
> Verify FaceFind selfie search and BibTrack BIB search on your trial gallery
## Why This Matters
Photo search is the feature most organisers come to 9Pic for. Before you upgrade, you should be confident that:
* **9Pic FaceFind** correctly returns a participant's photos when they upload a selfie.
* **9Pic BibTrack** correctly returns a runner's photos when they enter a BIB number (running events only).
This page walks you through both, end-to-end, on your trial event.
Search runs on the shared `trial.9pic.ai` domain, which is **rate-limited for anti-abuse**. Run a handful of searches yourself (5-10 is plenty to judge quality) — **don't** crowdsource the test by sending the link to a list of real participants. Anti-abuse will block them. After you upgrade, your domain moves to `app.9pic.ai` and the limits drop to normal production levels.
## Before You Begin
* You finished [Run Your First Event](/trial-guide/first-event) and the event is **Live**.
* The event has at least **20-50 photos** processed. Search quality is harder to judge with fewer photos.
* For BibTrack: the event was created with **Event Type = Running** and the photos contain visible BIB numbers.
* For FaceFind: the photos contain clear, mostly front-facing faces.
* You have a **selfie** of someone you know is in the photos, and a **BIB number** you can verify (running events only).
If your photos do not have clear faces or BIB numbers, the test is not meaningful. Don't grade FaceFind or BibTrack on a gallery they were never going to work on.
## Test FaceFind (Selfie Search)
From your event's **Details** page, open the configuration tab and confirm **Selfie Search** is **enabled**.
If it wasn't enabled when you created the event, turn it on now and **re-run Start Processing** so 9Pic can build the face index.
From the event header, click **Open Link**. Or use **Copy Link** and open the URL on your phone (selfie search is easier to test on a phone with a fresh selfie).
On the participant page, choose **Selfie Search** (or the equivalent search-by-photo option). Upload a clear, front-facing photo of someone you know is in the gallery.
Look at what comes back. Good results look like:
* The first 3-10 results are clearly the same person.
* There may be a few low-confidence near-misses further down.
* The person's photos that are visibly clear in the gallery are returned.
Failure modes to recognise:
* **No results**: face index hasn't been built. Re-check Selfie Search is on and re-run Start Processing.
* **Wrong person at the top**: usually means the test photo isn't front-facing enough, or the gallery has very few faces. Try a clearer selfie.
* **All matches missing**: confirm the gallery photos actually show the test person's face.
Repeat with a different participant's selfie. Search quality should be **consistent** across people, not just lucky for one face.
## Test BibTrack (BIB Number Search) — Running Events Only
Open the event's **Details** page. The event must have been created with **Event Type = Running** (or another running-related type) for BIB search to be available.
If the event type is wrong, the cleanest fix is to create a fresh test event with the right type. BIB indexing is set up at creation time.
From the event header, click **Open Link**.
On the participant page, choose **BIB Search** and enter a BIB number you can verify — ideally one whose photos you've already eyeballed in the gallery.
Good results look like:
* The matched runner's photos are in the top results.
* The BIB number is clearly visible in those photos.
* Photos where the BIB is occluded or blurry may be missing — that's expected.
Failure modes to recognise:
* **No results**: usually the BIB was never visible enough in the photos. Pick a runner whose BIB you can clearly see.
* **Wrong runner returned**: often a digit confusion (e.g. 8 vs 3). Try a BIB with very distinct digits.
* **Search not available on the page**: confirm the event was created with the running type. Re-create as a test if needed.
## What Counts as "Good Enough"
Search is probabilistic. A useful question for your decision is not *"is search perfect?"* but *"is search better than what my participants would do manually?"*
| Signal | What it means |
| -------------------------------------------- | ------------------------------------------------------- |
| Top-3 contains the right person | Strong. This is what most participants experience. |
| Top-10 contains all the obvious photos | Good. Participants will scroll a little. |
| Some near-misses with similar-looking people | Normal. Selfie quality usually fixes this. |
| Top result is a stranger every time | Investigate. Re-check the photos and re-run processing. |
## Common Questions
The face index is built during **Start Processing**. After turning Selfie Search on, run **Start Processing** again so 9Pic can index the existing photos.
Usually it's input quality, not the model. Try a clearer, front-facing selfie. For BIB search, pick a BIB you can clearly read in the photos. If problems persist, see [Search not working](/dashboard-guide/troubleshooting/search-not-working).
No. Search is free for participants. Credits are consumed when 9Pic processes photos (during **Start Processing**), not when participants search.
Yes. FaceFind works on any event type. Only BIB search is gated to running event types.
The flow works the same, but your "score" isn't meaningful. Add more photos before judging search quality.
Search on `trial.9pic.ai` is rate-limited for anti-abuse. If you're hitting limits during legitimate testing, slow down to a handful of searches at a time. If you've shared the link with real participants, anti-abuse is doing exactly what it should — stop sharing externally and continue evaluation internally only.
## Next Steps
Tick the search items off the checklist and continue.
Deeper reference for how AI analysis and rotation work.
Search works on every paid plan — pick the one that fits your scale.
If search is empty or wrong, walk through the fix path.