How we calculate the verdict
The Dobeo verdict follows fixed, public rules, the same for every store. We don't use artificial intelligence to decide and we don't make predictions.
Rules v3, published on September 29, 2026.
This is a translation. If it differs from the Portuguese version, the Portuguese version prevails. Read the original in Portuguese.
The price we compare
We compare the upfront price (Pix or boleto, in Brazil) with the history of that same upfront price. The card total is shown alongside, but it does not enter the verdict: on the same page, a product can cost R$ 3,999 via Pix and R$ 4,776.67 in 12 installments, and mixing the two would invent drops and rises that never happened.
How we build the history
Each reading holds until the next one, for at most 36 hours; periods without readings don't count, and sold-out products have no price. The typical price is the median weighted by how long each price was in effect. That way, reading the price many times during a sale that lasts a few hours does not change the typical price.
The verdicts
| Verdict | When it appears |
|---|---|
| Still learning | Less than 7 days of history or fewer than 4 readings. We can't tell yet whether the price is good. |
| Good price | The price is among the lowest 10% of the last 90 days and at least 3% below the typical price. With less than 30 days of history, it has to be at the lowest value we have recorded. |
| Normal price | Neither among the lowest nor 5% above the typical price. |
| Above normal | At least 5% above the typical price of the last 90 days. |
| Unavailable | The last reading found the product sold out. |
The badges
The store advertises a discount of at least 5% and there is at least one piece of evidence that it is bigger in the ad than in reality:
- the “was” price did not appear at any point in the 90 days we tracked (we only use this evidence with 21 days of history and 60% coverage);
- the real discount is less than a third of the advertised one (we use this evidence only with at least 7 days measured before the current price);
- the price rose 10% or more in the 30 days before the sale.
We call it a smaller discount than advertised, not false advertising: we describe the number, not the store's intent. Whether there was false advertising is for Procon (Brazil's consumer protection agency) or the courts to decide.
The current price is the lowest since we started tracking the product, with at least 14 days of data.
Real discount
We measure the discount against the lowest price of the 30 days before the current price started. It is the rule the European Union has required from stores since 2022 and the one bill 2849/2015 proposes for Brazil. If the store advertises 33% and the real discount is 4%, we show both numbers.
Until we have measured the full 30 days, we compare with the lowest price we recorded and say since which date. We write “the previous 30 days” only when we measured those 30 days.
When we alert you
An alert only goes out after we confirm the price with a second reading, taken at least 10 minutes later. We send at most one alert per product every 24 hours, unless the price drops another 3%.
Where prices come from
- Direct readings from the store, made by our server, several times a day for the products someone follows.
- The official Mercado Livre API, when connected.
- The Chrome extension, when the people who use it agree to share the prices they see, anonymously. An extension reading never triggers an alert on its own.
Limits
We don't know individual coupons, loyalty program prices or the shipping cost to your postal code. Prices seen while logged in may be different. Prices change at any moment: always check the store before you buy.
Independence
No store pays to change a verdict. If we ever use affiliate links, we will say so on the page, and the verdict will stay exactly the same. Each verdict keeps the version of the rules that produced it; when we change a rule, it is published here.