You search a screenshot, close the tab, come back a week later and search the same screenshot again. Different shops. Different prices. Maybe the jacket is now described differently.
Nothing went wrong. Three things move underneath the search, and your photo is not one of them.
Did the shops change?
This is almost always the answer.
Snagfit runs a live search against current shopping listings every single time. It does not keep a copy of what it found last week. So a listing that sold out is gone, a listing that was added yesterday is now there, and a retailer that pulled a product has taken it out of your results.
Snagfit returns up to six listings per garment. If two of the six from last time have sold out, two different ones fill those places.
Did the prices move?
Retailers change prices constantly. Sales start, stock runs low, seasons end, markdowns land.
Snagfit shows the price exactly as it appears in the listing at the moment you search. It never stores an old price and shows it to you later, because a stale price is worse than no price. You would arrive at the shop and find a different number.
Did the wording of the search change?
This one is smaller, and it surprises people, so it is worth being straight about it.
The step that reads your photo and names each garment runs a vision model. A model can phrase the same garment differently on two separate runs. "Chunky retro runner sneakers" one time, "chunky retro trainers" another.
Snagfit runs that step at a low temperature setting, which is the control that reduces this kind of variation. The category and the colour stay stable. The exact wording can shift, and a shifted phrase runs a different shopping search.
What does Snagfit actually save?
The photo and the detected items, so you can reopen a fit you found before and see what was in it.
The product matches are deliberately not saved. They are fetched fresh each time you tap Snag it. A stored list of shops and prices from three weeks ago would be wrong more often than right, and a wrong price sends you somewhere on false information.
What are the three reasons, in short?
The shops changed, the prices moved, and the wording shifted. In that order of importance, and the first one accounts for most of what you notice.
Snagfit searches live shopping listings every time you tap Snag it, so what comes back reflects what is actually for sale at that moment. Between two searches, listings sell out, prices move, and new sellers appear.
None of these are settings you can turn off, and none of them indicate anything is broken. They are what "live" means when the thing being searched is real retail stock rather than a fixed database that never changes.
Why search live every time?
It would be technically easier to run the shopping search once and keep the answer. Snagfit does not, and the reason is worth spelling out because it is the cause of the variation people notice.
A stored result is a photograph of a shop shelf. It is accurate for a moment and then it decays. Within days some listings have sold out, some prices have moved, and a few products have been pulled entirely.
If Snagfit showed you that stored answer, everything would look wonderfully consistent and a good share of it would be wrong. You would tap a listing and find a dead page, or arrive at a shop expecting one price and see another. The consistency would be a lie about the state of the world.
Running the search fresh means what you see is true at the moment you looked. The cost of that honesty is that two looks can differ.
Is this a caching problem?
People sometimes assume a stale cache is behind the differences. It is the opposite: there is no cache on the shopping side at all.
Caching would make results look more consistent, not less. What you are seeing is the absence of one. Each search goes out and asks what is for sale right now, and retail genuinely changes between two Tuesdays.
The only thing held between visits is your photo and the garments found in it, because those describe your picture rather than the state of any shop.
Which parts of the result stay the same?
It helps to know which parts of the result are stable, because then a change tells you something instead of just being noise.
Stable. The photo itself, obviously. The garments Snagfit found in it, since those are saved. The category each item was filed under. The rough colour.
A little variable. The exact wording of a garment name if you re-run detection, because a model can phrase the same thing two ways.
Genuinely variable. Which shops appear, how many listings come back, the prices, and the order they are in. All of this is the shopping search, and all of it reflects live retail.
So if the garment names changed a lot, look at your photo: a vague description is less stable than a specific one. If only the shops changed, nothing is wrong at all.
Why do two phones show different shops?
If you search the same photo on two devices and see different shops, nothing is wrong.
Each search is a separate live request, so two searches a few seconds apart can still land on different results. The comparison people usually intend to make, whether one device gets better results than another, is not a comparison at all.
What does change results is the photo, not the device.
Why do results swing most during sales?
There are stretches of the year when the same search looks completely different from one day to the next.
End of season markdowns, big sale weekends, and the run-up to a holiday all move prices and stock hard. A garment that had six listings between 180 and 220 in one week can have six listings between 90 and 200 the next, with three different shops in the list.
This is a good time to re-run an old saved fit rather than a reason to distrust the results. If you saved something in September because it was too expensive, searching it again during a sale period is one of the more practical uses of the saved list.
Why does the same photo sometimes find fewer items?
Occasionally a re-run of detection finds seven garments where it previously found eight, or files something differently.
This is the same low-level variation as the wording, and it shows up most on items that were marginal in the first place: a partly hidden accessory, a small piece of jewellery, something at the edge of the frame.
If it matters, the fix is not to keep searching. It is to give the detection more to work with, which means a tighter crop or a better frame. A garment that is clearly visible gets found every time.
What does this mean when you are deciding to buy?
One practical consequence worth stating.
Do not treat a price you saw on Snagfit as a commitment from anyone. It is what the listing said at that moment. Prices, stock and shipping terms are the retailer's, and their own product page is the only place any of it is authoritative.
Check the price on the shop's page before you buy, every time. That is true of any shopping search, and it matters more when the listing you are looking at came from a live result that was fetched seconds ago from somewhere else.
When is it worth searching the same photo again?
When the first search did not turn up anything you liked.
This is genuinely worth doing for a piece that was recently sold out. Stock comes back. New sellers list things. A search a week later is looking at a real different set of shelves, not the same set reshuffled.
How do you get steadier results?
Give the description step more to work with.
Most of the variation people notice comes from a vague description matching a huge pool of listings. "Long dark coat" can match thousands of things, so any reshuffle looks dramatic. "Camel double breasted wool overcoat, below knee" matches a narrow set, and that set stays much the same from week to week.
The way to get a specific description is a tight crop of a still, well-lit garment. There is a whole post on cropping, because it is the part you control.