Methodology: how gool.li computes its figures
The gool.li methodology rests on one rule: a figure is published only with its sample size, its date and its nature. Every price is an asking price, taken in Morocco from de-duplicated online listings; none is a sale price.
Sources
gool.li reads listings published online in Morocco, by private sellers and by professionals. The number of sources followed is part of the service’s definition, at the top of the About page; it changes when a source is added or suspended.
One row per real item
The same car is often published several times, on several sources or by several intermediaries. Listings that describe the same item are grouped, and only one counts in each computation. Without this grouping, a seller who reposts ten times would weigh ten times in the median.
Asking prices, not sale prices
A listing price is what the seller asks. The price paid after negotiation is written nowhere, and gool.li does not estimate it. Every published figure must therefore be read as an asking price.
Median and range
The median is the price that splits the sample into two halves: it barely moves when one listing is very high or very low. The published range runs from the first to the third quartile: half of the listings sit inside it. Plainly erroneous prices are left out before the computation.
Minimum sample sizes
Under n = 10, a group is not shown. Between n = 10 and n = 30, it is shown with its size and the mark “small sample”, and it is not used in any summary sentence. A change over time is given as a number only when both periods hold at least n = 100 each.
Comparing two periods
Two months do not hold the same cars: a month rich in recent models shows a higher median without any price having moved. A change is therefore published only like-for-like: both periods are recomputed over the same year and mileage bands, weighted the same way. Failing that, the two medians are shown side by side, with no percentage.
Update frequency
Price pages are recomputed every day; public selections are recomposed several times a day. The computation date is on every page. A Barometer edition, once published, is no longer changed: a correction is added to it as a dated erratum.
Risk signals
Seven automatic checks run on a car listing, in five families: the price against comparable listings, the plausibility of the mileage, the consistency of the seller type, reposting, and certain wordings in the text. These are automatic signals, not an inspection: a signal invites a question, it proves nothing.
Known limits
The sources followed do not cover the whole market. Prices are asking prices. In small towns and for rare models, the sample is often too thin to publish a figure. A poorly filled-in listing may land in the wrong band.
The Lhamza and Nqiya selections
Every hour, gool.li composes public selections from listings first seen in the last 48 hours. Listings that describe the same car are grouped, a listing without photos enters no selection, and a listing with a significant risk signal is left out. Two selections carry a name; each is defined by its composition rule.
Lhamza
Lhamza is the name of gool.li’s “Below market price” selection: used cars listed in Morocco at least 10% below the median of comparable listings.
Nqiya
Nqiya is the name of gool.li’s “No risk signal found” selection: used cars listed in Morocco on which none of the 7 automatic checks found a signal, with at least 3 photos and a description.
A selection reports facts and automatic checks at a given date; it is not a recommendation. An automatic check is no substitute for inspecting the car.