Food & Drink

Pastel de Nata quality score

A 4-axis scorecard that turns "is this one any good?" into a number you can compare across bakeries.

What this solves

Every guide to Portugal tells you to try a pastel de nata. None of them tell you how to tell a great one from a sad one — except vaguely by bakery name. But the same bakery can serve you a perfect one at 9am and a slightly tired one at 4pm, and the tourist-facing "famous" bakery right next to Belém Tower is not always the best one.

This tool gives you a 0–10 score across four axes that actually distinguish a great pastel from an okay one. Use it on the spot in any pastelaria, and you'll know within two bites whether to commit to a second one or move on.

Score the one in front of you

Adjust each slider after your first bite. The total updates live.

1. Top crust (blistery & dark)0

The classic sign is dark caramelised spots with slight burn marks. If the top is uniform yellow, it's underbaked. If it's a sheet of brown paint, it's overbaked.

2. Crust (flaky & shatter)0

A great pastel shatters when you bite through the side. If it's doughy, soft, or bread-like, the lamination failed or it sat too long. Pastel de Belém machines it; hand-laminated is a different texture.

3. Custard (set, not runny)0

It should wobble but not run. Bright eggy yellow, not pale. Sweetness moderate — not "pastry-shop frosting sweet". A small vanilla note (some bakeries skip it) is a positive sign.

4. Serving temperature (warm, not cold)0

The best ones are served warm, around 35–40 °C. Cold pastel de nata is a missed opportunity — the custard firms up and the crust softens. Many bakeries reheat on request.

Total (0–40, normalised 0–10) 0.0
Adjust the sliders to score your pastel.

How to read the result

A single high score can hide a flaw — if crust is 10 but custard is 3, it's a beautifully baked bowl of vanilla pudding. Use the per-axis numbers, not just the total, to compare surfaces.

Lisbon pastelaria reference (2026)

Indicative scores based on multiple visits. Expect ±1 point depending on time of day and season.

PastelariaAreaTop crustCrustCustardTempTotal /10Note
Pastéis de BelémBelém88888.0The "original" 1837 recipe. Served fresh from the oven, near the entrance. Skip the takeaway boxes near the door — those are the driest.
ManteigariaChiado / multiple98898.5Made in front of you, served warm. The "factory" ones taste slightly less individual than Belém.
AlomaCampo de Ourique89968.0A neighbourhood classic. Ask them to warm it — they usually do.
Pão Pão Queijo QueijoVarious77777.0Reasonable chain, decent price, zero queue.
Fábrica da NataMultiple66666.0Often aimed at tourists. Service is OK; pastry is "fine".
Hotel pastry shopAny44534.0Often the worst on the buffet. Skip if the brand isn't the point.
Continente / Pingo DoceSupermarket55524.3Surprisingly OK when warm from the supermarket bakery, average 4 a packet. Won't change your life.

These are not "official" scores. They are based on 4-axis scoring across multiple visits. Expect ±0.5 between visits, ±1.5 between branches of the same chain.

Limitations

FAQ

Is Pastéis de Belém really the best?

It's the most famous, and on a busy day it's a great batch. On a quiet morning it's still very good but other bakeries match it. The famous 1837 "secret recipe" is a marketing asset; the actual pastel quality is consistent rather than transcendent.

Why do some have cinnamon, some don't?

Cinnamon on top is optional and traditional but not universal. Manteigaria serves them without cinnamon by default; Pastéis de Belém offers a shaker. Both are correct. The cinnamon doesn't change the score — it's a personal choice.

Are airport pastel de nata any good?

Usually no — they're rewarmed. The crust is softer, the custard is set, the score drops by 2–5 points. Eat one before security if you can.

Can I take them home?

Yes, but they don't travel. Pastéis de Belém has a worldwide shipping scheme that uses partial baking — it's so fine it isn't the same product. Best eaten within 30 minutes of baking.

Related tools

Sources & methodology