Lisbon florist ratings and review counts

Lisbon florist ratings and review counts in this file are two columns that refuse to travel together. The unweighted mean of the 12 ratings is 4.775. The review-weighted mean is 4.6985, from a numerator of 8,142.5 divided by 1,733 reviews. The published rank still sorts by rating first. A 5.0 with 29 reviews outranks a 4.4 with 299 reviews. The Flores Island homepage, accessed 2 Nov 2026, describes “Saldanha & Parque das Nações studios” in its meta description. That sentence is the shop’s wording. The neighborhood cell on the row is still empty. Three rows are 5.0 and together hold 93 reviews. Five rows are 4.7 and together hold 749. The largest single count, 401, belongs to Florista Bella at 4.7.

Key takeaways

• Report a rating with its review count in the same breath. Either number alone misleads on this file.
• Use 4.775 when the question is “what is a row worth if each row counts once”.
• Use 4.6985 when the question is “what rating does a stored review sit on”.
• Do not promote the product column to a rank. The homepage key is unchanged.
• Shop pages fetched on 2 Nov 2026 contribute titles, not stars.

How are the 1,733 reviews distributed across the four ratings?

Each row contributes its review_count to exactly one band, because each row has one rating. The bands are 5.0 (93 reviews, three rows), 4.8 (592 reviews, three rows), 4.7 (749 reviews, five rows), and 4.4 (299 reviews, one row). Shares of 1,733, rounded to one decimal, are 5.4%, 34.2%, 43.2%, and 17.3%. Those four rounded shares add to 100.1 because of rounding, not because a review was counted twice. The unrounded shares are 93/1,733, 592/1,733, 749/1,733, and 299/1,733.

The weighted mean uses the same bands. 5.0 times 93 is 465. 4.8 times 592 is 2,841.6. 4.7 times 749 is 3,520.3. 4.4 times 299 is 1,315.6. The sum is 8,142.5. Dividing by 1,733 yields 4.6985. The unweighted mean adds the 12 ratings and divides by 12: 15 + 14.4 + 23.5 + 4.4 = 57.3, and 57.3 / 12 = 4.775 exactly. The weighted figure sits 0.0765 below the unweighted figure because the heavy counts sit on 4.7 and 4.4, not on 5.0.

A reader who says “these shops are about 4.8” is closer to the unweighted mean (4.775 rounds to 4.8 at one decimal) than to the weighted mean (4.6985 rounds to 4.7). Both roundings are ours. The file stores one-decimal ratings only. We keep the extra digits on the means so the two questions stay distinct. Collapsing both to “about 4.8” hides the 0.0765 gap, which is the whole point of weighting.

Shares in that table are each count divided by 1,733, rounded to one decimal. 34/1,733 is 1.96%, which rounds to 2.0%. 30/1,733 and 29/1,733 both round to 1.7%. The band shares above are a different rounding of the band sums. Do not add the row shares and expect the band share to match to the last tenth.

Why is the 5.0 band thin on reviews?

The three 5.0 counts are 34, 30, and 29. Sorted, they are three of the four smallest counts in the file. Only Flores Island Studio, at 60, sits near them from outside the band, and 60 is still about double 29. The next count above Studio is Finaflor at 73, which is a 4.8. So the entire top rating band lives below the median review count. The median of 12 counts is the mean of the 6th and 7th sorted values, 83 and 89, which is 86. 34, 30, and 29 are all below 86. A 5.0 in this file is a high rating on a small count, not a high rating on a large count. That sentence is a description of these 12 integers. It is not a general law about florists, and it is not a reason to discard the rows.

Inside the 5.0 band the shares of 93 are even. 34/93 is 36.6%. 30/93 is 32.3%. 29/93 is 31.2%. Those three shares add to 100.1 at one decimal, again a rounding residue. The counts sit close to each other. The gap that matters is not among them. It is between the whole band, 93 reviews, and Bella’s 401 on a single row.

Which three rows hold the bulk of the reviews?

Florista Bella (401), Pequeno Jardim (386), and Fleurs à Lisbonne (299) sum to 1,086. That is 1,086 / 1,733 = 62.7% of stored reviews, on three rows, at three different ratings: 4.7, 4.8, and 4.4. Their homepage ranks are 7, 4, and 12. A reader who wanted “where the reviews are” would start with those three names and would not start at rank 1. The rank guide already shows the alternate order. This guide adds the share: those three rows are 62.7% of the review mass and 3 of 12 rows, which is 25% of the rows. The mass and the row count do not match. That mismatch is the reason a mean of ratings and a mean of reviews answer different questions.

Pequeno Jardim’s 386 is 22.3% of all reviews by itself, on a 4.8. Bella’s 401 is 23.1%, on a 4.7. The 0.1 rating gap between them is smaller than the gap from either of them down to 4.4, and the review counts are in the same hundreds. Fleurs à Lisbonne’s 299 is the third mass, and the rating is the file’s minimum. Treating 4.4 as “an outlier to drop” would drop 17.3% of the reviews. We do not drop it. The row stays rank 12 under the published key, and it stays visible in this table.

Flores Island Shop, at 133 and 4.8, is the next count down and the only “Designer de flores” category string. 133 is above the median and well below the three heavy rows. The Finaflor by Danni homepage, accessed 2 Nov 2026, says “Florista em Alvalade, Lisboa” in the meta description. Finaflor’s row is 4.8 and 73 reviews, category “Florista”. Alvalade is their sentence. It is not written into the neighborhood cell. Saudade Flores, 4.7 and 116, has a homepage title, fetched the same day, “Saudade Flores - Florista em Lisboa com Entrega no Mesmo Dia”. The title asserts same-day delivery in the shop’s words. We do not adopt the assertion as a test. Fleur de Paris titles itself “Fleur de Paris | Bouquets Exclusivos de Luxo com Entrega em Portugal”. Again the title is theirs. The row remains 4.7 and 89 reviews. Saturday on that row is truncated, which the weekend guide handles. The rating does not depend on the Saturday characters.

The quote is 24 words. It names two studios. Our neighborhood column names none. Both statements can be true at once: their page said that on 2 Nov 2026, and our record’s neighborhood string is empty. Filling the cell from the quote would mix a fetched sentence into a field the file left blank. The fields guide keeps that boundary.

Why is the product column not a rank?

Multiplying rating by review count gives Bella 1,884.7, Pequeno Jardim 1,852.8, Fleurs à Lisbonne 1,315.6, then Flores Island Shop 638.4, and on down to Flow Flowers at 145.0. The order of the products is close to the order of the review counts, because the ratings only range from 4.4 to 5.0. A factor between 4.4 and 5.0 cannot reorder 401 above 386 (5.0 would be required in a way the file does not have: 4.8 × 386 = 1,852.8 and 4.7 × 401 = 1,884.7, so Bella’s product still leads). It can reorder a mid-size count against a small 5.0: Flor e Pão’s product is 170, below Studio’s 282, even though Flor e Pão is rank 1 and Studio is rank 11. The product therefore punishes small counts and barely notices the 5.0. Publishing it as “the real ranking” would be a third policy, not a correction of the homepage. We print it so the arithmetic is checkable, and we leave the homepage on rating then review count.

What does Nils write down before he orders?

The useful habit is the one Nils uses: write the count beside the rating, and do not let the rank number erase either. Rank 1 and “401 reviews” are both in the file. A page that mentions only one of them is editing the record. This guide mentions both, and it points at the hours checker only when the next constraint is a clock time. A rating does not say whether Monday 21:00 is inside a stored range. Flor e Pão’s Monday string happens to cover 21:00, and that fact lives in the weekday guide, not in the 5.0.

How should a rating cell be quoted?

1. Copy the rating to one decimal, as stored: 5.0, not 5.
2. Copy the review count as an integer. Do not round 386 to “about 400” in a table.
3. If you cite a mean, say whether it is 4.775 (per row) or 4.6985 (per stored review).
4. If you cite a share, show the numerator. “5.4%” without “93 of 1,733” is incomplete.
5. Leave the product column out of any sentence that claims to restate the homepage rank.
6. If a shop title names a neighborhood, quote the title and leave our neighborhood cell empty.

The rating column is a file field dated 2 Nov 2026. When the file changes, both means have to be recomputed from the new integers. Until then, 4.775 and 4.6985 are the two honest summaries, and 401 remains a larger count than 34.