How Filtizy scores a product: the factors, the weights and what a 72 actually means

A product score is only useful if you can take it apart. Most shopping sites give you stars that summarise what other buyers felt; Filtizy gives you a number that summarises what the pack declares. Those are different questions. Stars tell you whether a delivery arrived on time and whether people liked the taste. A Filtizy score tells you how one product compares with every other product in the same category on a fixed list of measurable factors.

This post is the plain-language version of what the scorer does. If you want the formal version it lives on our methodology page, and the factor list with its weights is a table in our database that admins can see and change - not a secret model.

One score, two profiles

Not every product has a nutrition panel, so the scorer runs one of two profiles depending on what the category can actually supply.

The food profile is used for anything with a declared nutrition panel - groceries, snacks, drinks, breakfast cereals, sweets, nuts. It scores the panel itself, plus what the ingredients list implies.

The generic profile is used everywhere else - skin care, apparel, home, electronics. It scores quality signals, value for money, how much verified data we hold on the listing, and how widely available it is across retailers.

The food profile, factor by factor

Eleven factors, one hundred weight points. The direction column matters as much as the weight: for some factors a lower printed value scores better, for others a higher one does.

Added sugar carries the single largest weight because it is the value most often high and least often necessary in packaged food, and because Indian dietary guidance is explicit about it. Protein and fibre together carry 26 points, which is why a plain roasted-nut pack outscores a flavoured one that lists sugar third.

How a factor becomes a number

Each factor is scored on ten rungs rather than a smooth curve: 100, 90, 80 and so on down to 10. Rungs were a deliberate choice. A smooth formula implies a precision the source data does not have - a pack that declares 4.9 g of sugar and one that declares 5.1 g are not meaningfully different, and a scoring system that ranks them apart is pretending to know something it does not.

The factor scores are then multiplied by their weights and summed, which gives the number you see on the card and on the product page. The per-factor breakdown is on every product page, so you can see which factors pulled a score up and which held it down.

The missing-data problem, and what we do about it

Here is the failure mode every scoring system has to answer for. If a pack does not declare added sugar and the scorer treats the blank as zero, the product gets full marks on the heaviest factor in the profile for the simple act of not telling you. That is backwards, and it rewards exactly the labels that deserve the least trust.

A blank is not a zero. When a product gives the scorer no real panel data at all, it does not score well - it scores nothing, and the product page says the data is incomplete instead of showing a flattering number.

Counts the scorer works out for itself - how many ingredients are listed, how many of them are flagged additives - do not count as declared data for this test. Otherwise a pack with nothing but an ingredients line would qualify as scored.

What a 72 means, and what it does not

A 72 means: on the eleven published factors, weighted as above, this product sits comfortably in the upper half of its category. That is the whole claim.

It is not a health verdict on you. It does not know your calorie needs, your training load, your allergies or your budget. It does not know that the 84-scoring option tastes like cardboard to you, and it does not know that you are buying biscuits for a child who will eat two, not the whole sleeve.

The number is a shortlisting tool. It is built to cut a 200-product category down to the eight worth actually reading - after which the per-factor breakdown, the ingredients list and your own preference do the rest.

Comparing scores fairly

• Compare within a category. A 78 on a green tea and a 78 on a breakfast cereal were computed against different peers.
• Check the data completeness marker before you trust a high score - a thin panel means a thin score.
• Open the factor breakdown when two products are within a few points; they are usually strong in different places.
• Check the per 100 g column, not the per serving column, when you read the panel yourself.
• Remember that price is not in the food profile at all. A great score can still be a bad buy.

If you think a score is wrong

Sometimes it is. A pack gets reformulated, a panel is transcribed wrong, a variant gets attached to the wrong listing. Every product page has a way to flag it, and corrections go to the same queue our own catalogue team works from. Tell us the product, the value you think is wrong and where you read the correct one - a photo of the back of the pack is the single most useful thing you can send.

That is the deal we are trying to hold up: the factors are published, the weights are published, the data is checkable, and when it is wrong you can say so.