How do I track protein accurately when I have no appetite and no reliable sense of portion size?

I have been logging for eleven weeks and I do not trust my own log. I thought I was averaging 110 g of protein. Then I weighed everything for four days out of curiosity and the real number was 71 g. Nothing about my eating had changed; my estimation was just wrong, consistently and in one direction.

Two things I think are going on. One, I am eating tiny portions and my mental model of a portion was built when I ate normal ones, so a "chicken breast" in my head is 180 g and the one I actually finish is 90 g. Two, I abandon food halfway through now, and I log what I served rather than what I ate, because I serve it and log it and then twenty minutes later I am done.

How do people handle this without weighing every gram forever? I am not going to sustain a kitchen scale for the next two years, and I would like a system that fails gracefully rather than one that is accurate for six weeks and then gets dropped.

Your diagnosis is correct and both errors are systematic, which is good news, because systematic errors can be corrected once and then estimated against. You do not need permanent weighing. You need periodic calibration plus a structure that removes the estimation from the high-error items.

Fix the plate-waste error first, because it is the bigger one

Weigh the plate, not the food. Put the plate on the scale before you eat, note the number, put it back on afterwards, note the number, and the difference is what you ate. This costs about eight seconds per meal, requires no food-by-food weighing, and eliminates the abandonment error entirely. If most of your protein is in one item on the plate, weighing that item served and remaining is even simpler.

This alone probably accounts for a large part of your 39 g gap. Someone leaving 40% of a 150 g portion is over-logging by 25 g of protein from a single item.

Then remove estimation from the protein items specifically

You do not need to track anything accurately except protein. Vegetables, fats and carbohydrates can be estimated badly with almost no consequence for your stated goal. So build your day out of protein units with known, fixed values and stop estimating the ones that matter:

• Pre-portioned protein: tinned fish (whole tin, known value), a scoop of isolate (known value), a pot of Greek yoghurt (known value), a fixed number of eggs, a pre-weighed 120 g bag of cooked chicken portioned when you cook.
• Batch-weigh once, portion into containers. You weigh on Sunday for the week, not at every meal.
• Keep a written list of five to eight protein units with their gram values. Your daily log becomes "which units did I finish", which is countable rather than estimable.

A worked example of what that looks like for a 115 g target:

1 scoop whey isolate               25 g
170 g pot 0% Greek yoghurt          17 g
1 tin tuna in water, drained        24 g
120 g cooked chicken portion        37 g
2 boiled eggs                       13 g
                                   ----
                                   116 g

Five countable items. No estimation. If you finish four of the five you know you are at roughly 90 g rather than guessing.

Calibration schedule

Weigh everything for four consecutive days, ideally spanning both a high-suppression and a low-suppression phase of your dosing week. Compare your estimated log to the weighed log and compute your personal bias:

estimated average 110 g, weighed average 71 g
bias factor = 71 / 110 = 0.65

Now you know your untrained estimate runs about 35% high. Repeat the four-day calibration once a quarter or after any change in dose, appetite, or eating pattern, because the bias drifts as portions drift. Between calibrations, estimate freely and multiply.

The thing your log cannot tell you

Track something objective alongside the food log, because logs decay and bodies do not lie. Weekly average body weight, a monthly repeated strength benchmark, and periodic bloodwork. If your protein log says 115 g and your strength benchmark is falling for eight weeks, believe the benchmark.

Something the answer above implies but does not say directly: a 39 g measurement error is worth fixing, but the more urgent finding in this post is that OP's actual intake is 71 g.

Without a body weight we cannot compute the g/kg, but 71 g is below the floor for almost any adult who is losing weight rapidly. If OP is 80 kg that is 0.89 g/kg, barely above the population minimum requirement and far below anything defensible during rapid loss with the intention of retaining lean tissue.

So the priority order should be:

1. Raise actual intake, immediately, using the highest-tolerance vehicles available.
2. Get accurate logging in place, so you know whether step 1 worked.
3. Optimise distribution and sourcing, eventually, if at all.

Most people arriving at this problem reverse the order and spend three weeks building a tracking system while continuing to eat 71 g. The tracking is instrumentation, not treatment.

The fastest single intervention if you are at 71 g: two protein shakes a day, one on waking and one in the evening, at 25-30 g each. That is 50-60 g of protein that requires no appetite, no chewing, no cooking and no estimation, and it more or less closes the gap on its own. Whether shakes are optimal is a question for later. Right now they are the difference between adequate and inadequate.

One database caveat that produces surprisingly large errors and no warnings: raw versus cooked entries.

Chicken breast loses roughly 25% of its mass to water during cooking, so 100 g raw becomes about 75 g cooked, and the protein concentration per gram rises correspondingly. Most tracking apps contain both a raw and a cooked entry with almost identical names, and users select whichever appears first. Pick the wrong one and you have a 25-33% error on your single largest protein item, in either direction.

Rules that avoid it:

• Decide once whether you weigh raw or cooked, and only ever select the matching entry. Raw is more accurate because cooking loss varies with method and duration; cooked is more practical if you batch cook.
• Prefer entries that come from a national food composition database over user-submitted ones. User entries in the popular apps are frequently wrong by a factor of two and there is no correction process.
• Sanity-check any entry against the protein-density heuristic: almost no unprocessed animal protein exceeds about 25 g of protein per 100 kcal, and lean cooked meat and fish land between 18 and 24 g per 100 g of food. If an entry claims 40 g of protein per 100 g of chicken, it is wrong.

Same trap applies to dry versus cooked weight for rice, pasta and oats, which matters less for protein but will wreck your energy estimate, and an energy estimate that is 300 kcal wrong is how people end up believing they have a metabolic problem.