How we measured

Three numbers appear across this site: how many photos SmileCrop handles without intervention, how many it flags for review, and how long the flagged ones take to finish. Here is the whole procedure, the results, and what this measurement cannot tell you.

Measured August 2026 in the author's own clinic. Ten photos and 456 photos, one person.

1. How many photos go through untouched

One ordinary batch from the clinic: 524 files off the card, of which 68 were QR name-tag shots (one per patient, used to split the photos into folders) and 456 were clinical photographs. Those went through SmileCrop in a single batch on normal settings.

Outcome	Photos	Share
Saved without intervention	423	92.8%
Flagged for review	33	7.2%
Total processed	456	100%

We round that to 93% elsewhere on the site. It is not a measure of crop accuracy. It means the software was confident enough to save the photo — not that we went back through those 423 finished crops one by one. We did not, and we do not pretend otherwise.

2. How long a flagged photo takes to finish

This is the number that matters: flagged photos are the only work left for a person. We took ten photos SmileCrop actually flagged in that batch — its own worst cases, some of which needed the name corrected too — and processed them twice.

Method	Time	Per photo
SmileCrop batch editor — crop, rename and file	1:32	9.2 s
By hand in another program — total	5:33	33.3 s
— of which cropping and renaming	4:47	28.7 s
— of which filing into patient folders	0:46	4.6 s

SmileCrop took 27.6% of the manual time. Discount the filing entirely and it is 4:47 against 1:32, or 32%.

Order and tools

SmileCrop went first, the manual run second. That order favours the manual side: by the second round it was already settled how each photo should look.
The manual run used the Windows photo editor. The clinic normally uses Zoner Studio; in the author's experience it offers nothing that would speed this particular job up.
The author of the software did the timing. That is both the fastest possible SmileCrop operator and someone who already knows where every photo belongs.

3. Where “97% less manual work” comes from

Plainly: this number is arithmetic, not a measurement. Nobody timed 456 photos cropped by hand against the 33 flagged ones from the same pile — that would be three hours of work for the sake of one figure on a website. Two things were measured: how the batch split (423 / 33), and two timed runs over ten photos. The rest is multiplication:

Handling 456 photos	Hands-on time
All by hand — 456 × 23.7 s	3 h 0 min
With SmileCrop — only the 33 flagged × 9.2 s	5 min 4 s
Manual work remaining	2.8%

Hence 97% less manual work (97.2% precisely; on a richer manual baseline it reaches 98%, which is why we quote the lower figure). Two things to see in that calculation:

The manual side got the easier basis. 23.7 s is the time for an ordinary photo without filing it into a folder; SmileCrop's 9.2 s includes the filing.
It is manual work, not your day. Shooting, pulling the card, starting the batch and taking the QR frame are not in it — and neither is the time the batch spends running. That is not the same as “a 97% shorter day”.

Why the batch runtime is not counted

Deliberately. How long a batch takes depends on the computer: an older office machine behaves differently from one with a fast graphics card. Counting that time would measure the hardware rather than the work, and the figure would stop applying to anyone with a different computer.

More to the point, nobody sits and watches it. You put the card in, start the run and go and do something else — leave it overnight if you like and deal with the flagged photos in the morning. Time you do not spend at the screen is not your time. Every figure on this page is therefore hands-on human time, and nothing else.

4. Where these numbers are used

The calculator assumes 7.2% of your photos will be flagged, and that a flagged photo costs you 28% of your manual time. That 28% is costed against your time for an ordinary photo, even though flagged ones tend to be the more awkward kind — so the result understates rather than overstates.

What this measurement cannot tell you

Named here so you do not have to go looking for it.

The 97% is calculated. No whole batch was processed both ways and timed end to end; it is two measurements multiplied together, each with its own limits listed below.
Ten photos is not a sample. The timings come from one sitting by one person. They indicate an order of magnitude, not an average; repeating them would give different figures.
The author did the measuring. It is neither independent nor blinded.
The manual filing had it easy. There was no tree of hundreds of existing folders to hunt through, only the occasional new one to create. In a real archive that 0:46 would be longer — so the manual side came out better here than it would for you.
We did not measure crop accuracy. 93% is the share of photos that were not flagged, not the share that were cropped correctly. The unflagged ones were not systematically rechecked.
One clinic, one camera, one series. The flag rate depends on how a practice shoots. A practice that shoots differently will see a different number.
This is not a study. There is no controlled experiment, no control group, no statistics and no peer review. It is a working measurement from practice, honestly described — no more and no less.

Which is why SmileCrop comes with 14 days free. Measure your own numbers on your own photos; treat ours as a starting point.

← Back to the time calculator