There are roughly three ways of measuring food waste: with scales and forms, with a camera above the bin, or with a photo of every plate, serving dish or pan. Each strikes a different balance between precision, effort and detail. We put them side by side honestly, so you can choose what fits your organisation.
What you want to know when measuring food waste
A measurement should give you more than a number in kilos. Researchers at Wageningen University & Research wrote in their measurement method for hospitals that "simply measuring the amount of food thrown away is often not enough for a good analysis" (WUR, report 1637, 2016). You want to know where in the process things go wrong, which food it is, and how it relates to what you bought.
For Dutch care organisations working with the Milieuthermometer Zorg, the Dutch environmental certification for care organisations, some of those questions are already settled. Requirement 7.3 asks you to measure and record kitchen waste, plate waste and banqueting at least once a year, and holds waste against purchased weight. Whichever method you choose, you will always need purchasing data.
We compare the methods on three points:
- Precision: how close is the result to the real weight?
- Effort: how much time and equipment does it take, and how long can you keep it up?
- Detail: can you see per ingredient, per ward and per meal what comes back?
Method 1: scales and forms
This is the classic approach and the basis of the WUR method. Buckets per food category, for example vegetables, meat and starch, stand by the dishwashing area. Staff separate the leftovers, weigh each bucket and write the weight on a form. WUR's list of materials: "scales, pens, forms, a binder, buckets". The figures then go into a spreadsheet.
Precision. High. You actually weigh, so for the total this is the yardstick other methods are measured against.
Effort. Considerable. Someone has to sort, weigh and record the leftovers, and one person checks every day that the forms are filled in correctly. "The measuring itself can only be automated to a limited extent," WUR writes. That is why such a measurement usually lasts two weeks and then stops. WUR also names two pitfalls: kitchen staff who weigh their own waste want it to look good, and during the measurement weeks a team often cooks more carefully without realising it.
Detail. As fine as your buckets. Per food group works well, per ingredient quickly becomes unworkable. Per ward is possible if you keep each ward's trays separate.
Fits: a one-off baseline with a project team, or a sample to check other measurements against.
Method 2: a camera above the bin
Here a scale sits under the bin and a camera hangs above it. Whenever someone throws something away, the weight is recorded and a photo is taken. Software recognises what falls into the bin, sometimes helped by a choice the staff member taps on a screen.
Precision. Every throw is weighed, so the total is reliable. How well the system recognises what's in a mixed throw differs per system.
Effort. Little once installed, because throwing away and measuring happen at the same moment. On the other hand, you need a station at each waste point, with power and space.
Detail. Per station. WUR notes that automated weighing is simple "when all food waste arises at the same time and in the same place, as in company catering". When cooking happens somewhere other than where the returns come in, "this requires a lot of extra coordination". In a hospital or nursing home, plates from many wards often come together at the dishwasher. If you want to know which ward or which meal something came from, ask how a system records that: with a station per ward, or by staff tagging each throw.
Fits: a central production kitchen with one clear waste point, where you mainly want to track kitchen waste continuously.
Method 3: a photo per plate, serving dish or pan
With this method you photograph what comes back, at the moment it comes back: every plate with leftovers, and the ward's serving dishes and pans. The photo is used to estimate how much is left and what it consists of. A trained rater used to do that. Today, AI image recognition can.
Estimating from a photo is an established research method. In a hospital study presented at the Nutrition Society of Australia conference in 2025, researchers measured 1,763 plates two ways: an estimate from the photo on a seven-point scale from 0 to 100% left, and the scales. The results tracked each other very closely across all plates (correlation 0.99), with an average difference of 4.67 grams. The researchers call photography "a valid method" for measuring plate waste in hospitals (Ni et al., Nutrition Society of Australia, 2025). Two caveats: this is a conference abstract, and the estimates were made by people, not by AI.
It pays off in practice too. Geel hospital in Belgium photographed the plates that came back and weighed the leftovers. "We saw that of our 3 scoops of mash on the plate, only 2 were eaten," facilities manager Paul Corty said (our translation). It moved to two scoops, took beetroot off the menu and matched every dish more closely to demand over two years. That took it from 712 kg of waste a month in 2023 to 289 kg in 2025 (VRT NWS, 2026).
Precision. It remains an estimate. In that study a single plate could be off by some 50 to 60 grams either way, while the average over many plates was almost spot on. A photo measurement is therefore strong on totals, averages and trends, and less suited to judging one plate to the gram. Soup, drinks and a poorly lit photo are harder.
Effort. Low to moderate. You need a phone, not an installation. Empty plates don't need a photo, so the work is only in the plates with leftovers.
Detail. High. You know which ward and which meal each photo belongs to, so you see per ward, per meal and per ingredient what comes back. The photos are kept as well, so you can look back at what was actually on the plate.
Fits: organisations that want to measure on many wards, keep it going, and know which dish or ingredient comes back. How reliable an AI estimate is, and where its limits lie, is covered in How reliable is measuring food waste with photos and AI?
The three methods side by side
- Precision on the total: scales high, bin camera high, photo per plate an estimate that averages out well over many plates.
- Daily effort: scales a lot, bin camera little after installation, photo per plate low to moderate.
- Upfront investment: scales low, bin camera hardware per waste point, photo per plate just phones.
- Per ingredient: scales hardly, bin camera depends on the recognition, photo per plate yes.
- Per ward and meal: scales if you sort separately, bin camera with a station or tag per ward, photo per plate as standard.
- Keeping it up after two weeks: scales hard, bin camera good, photo per plate good.
Start from your question, not the device
- Do you want to know once where you stand? A two-week baseline with scales and forms is very doable, especially with a project team. Milieuplatform Zorgsector (MPZ) advises measuring for at least two weeks. Food waste baseline measurement: a step-by-step plan for care helps you get started.
- Is most of the waste in the production kitchen, in one place? Then a system at the bin fits, because measuring happens automatically as things are thrown away.
- Do you want to know which ward, which meal and which dish? Then a photo per plate, serving dish or pan gives you that detail without any sorting.
You can also combine methods. Weigh a sample of plates now and then, and hold it against what the photos estimate. That tells you how well the estimate works on your plates.
For all three, no purchasing data means no percentage. So record what was purchased for the meals you measure. What the Milieuthermometer asks exactly is in Milieuthermometer Zorg: how to measure food waste for requirement 7.3. The full overview is in our guide Food waste in healthcare: what it costs and how to measure it.
How the photo per plate works in Waste Warrior Monitor
Waste Warrior Monitor is built on the third method. After the meal, a staff member chooses the ward, the meal and the number of people in the app, and photographs every plate with leftovers and the serving dishes and pans that come back. The AI estimates how many grams are left in each photo and which ingredients they are. The app says so too: "naar schatting", an estimate. If something isn't right, staff adjust the ingredient. Drinks or other leftovers you enter yourself in grams, so everything you throw away counts.
You link purchasing with a scanned delivery note, with dishes you enter yourself, or through bulk purchase import (CSV import) from your supplier's export. The dashboard shows waste as a percentage of purchasing, per meal, per ward and per ingredient, with a photo gallery you can search by ingredient. Automatic report generation (PDF report) records the outcome.
Curious what that looks like for your kitchen and wards? Book a demo.
