Under the hood
How the 0–10 score is calculated
No machine learning, no black box, nothing that needs an account. Four weighted components over public data, computed in your browser — written out here in full so you can decide whether you agree with it.
The formula
Four components, weighted
Each component is scored 0–10 on its own, then combined. The weights were tuned specifically for AliExpress, where ratings cluster so tightly at the top that the raw star value alone separates almost nothing.
The star rating, normalised onto 0–10. It carries the most weight because it is the one signal every listing has — but on its own it is also the most inflated, which is what the other three components exist to correct.
The share of positive reviews, penalised when negatives climb past 10%, and nudged up when AliExpress's own structured "Quality" label is strongly positive. This is where a popular-but-complained-about listing loses its shine.
A logarithmic curve over the review count. It answers "how much should we believe any of this?" — a thousand reviews is meaningfully more evidence than a hundred, but not ten times more.
Rating measured against price on a log scale, plus credit for a genuine discount. Lowest weight on purpose: a cheap bad product is still a bad product.
Then: the confidence dampener
Below 10 reviews, the total is scaled down — a listing with one review keeps roughly half of whatever it earned. This is the single most important rule in the model. Without it, every brand-new listing with one glowing five-star review would render as "Excellent", which is precisely the trap the extension exists to stop you walking into.
Reading the number
What each band means
| Score | Verdict | What it tells you |
|---|---|---|
| 8.0 + | Excellent | Strong on every component with enough reviews to trust it. |
| 6.5 + | Great | A solid buy; one component is merely good rather than excellent. |
| 5.0 + | Good | Fine, but check the breakdown before committing. |
| 3.5 + | Average | Something real is dragging it down — read the warnings. |
| below 3.5 | Poor | Thin evidence, poor sentiment, or a store with no track record. |
The same boundaries drive every colour in the extension — one table, so a badge and a bar can never disagree about what "good" means.
The verdict on the pill
The tier says how good the listing is; the verdict says what to do about it. It is cut from the same 0–10 score, at different points, so the two can disagree on purpose — a "Good" listing can still be "Buy with care".
- 8.0 and up
- Safe to buy
- 6.0 – 7.9
- Worth considering
- 4.0 – 5.9
- Buy with care
- below 4.0
- Risky
The other score
Seller trust is judged separately, and strictly
"Is this product good?" and "is this store safe?" are different questions, so they get different scores. Store trust is 70% positive-feedback rate and 30% follower count.
Why 85% counts as zero
AliExpress stores cluster between 90% and 100% positive feedback, so a raw percentage separates nobody. The scale is stretched: 85% maps to 0/10 and 100% to 10/10. A store at 94% is genuinely mediocre for this marketplace, and the score says so instead of flattering it.
Followers as a history proxy
Follower count is scored on a log curve — it rewards an established store without letting a mega-store buy its way to a perfect trust score on volume alone.
Where the data comes from
Public data only, scored locally
- 1
Read the page
Rating, review count, units sold, price, store feedback and followers — all of it already visible on the page you opened.
- 2
Ask the public review endpoint
AliExpress's own public review-statistics endpoint returns the positive/negative split, the five-star share and the structured Quality and Value labels. On a listing page it is asked only about a product you hover or click — never about every card on screen.
- 3
Score in your browser
The formula above runs locally. Nothing about you is sent anywhere, and the score exists only in the tab you are looking at.
- 4
Fail soft
If a selector moves or the endpoint does not answer, the panel says "not enough data". A product we know nothing about never renders as a green "Excellent".
Review lookups go directly from your browser to AliExpress — the extension does not send anything to any server we operate, and contacts no third party. A lookup happens only when you ask for one, and the answer is cached on your device for up to 72 hours, so opening the same product again usually sends no request at all. That is spelled out in the privacy page.
See it for yourself
Run the formula on a listing you invent
Everything described above, executing live. The presets are worth a click — especially "The 5-star trap".
Pretend this is the listing
And this is the store
AliScore verdict
🔥 Excellent
Score breakdown
- Quality (35%)9.6
- Sentiment (25%)9.6
- Review volume (25%)10.0
- Value (15%)10.0
No warning signs
Nothing in this listing trips a threshold the score penalises.
This panel runs the extension's actual scoring code — the same calculateScore() that ships in version 1.0.7. Move a slider and you see exactly what a real listing with those numbers would score.