Where diaries die

Food diaries rarely die at home. At home there’s a scale, labels, and a food library of your regulars. They die on a Friday night, over a plate of pasta of unknown composition. The chain is always the same: “I can’t log this precisely” → skip the meal → the day has a hole → the day “doesn’t count” → neither does Saturday → the app quietly retires.

The mistake isn’t your estimating skill. It’s the standard. A diary’s job is a usable picture of your week, not a lab report.

The math of a bad guess

Say you misjudge a restaurant meal by 30%. On a 700 kcal plate that’s about 200 kcal — roughly 10% of a 2,000 kcal day, and 1.5% of the week. The weekly average barely notices.

A skipped day is worse than a zero, because skipped days aren’t random. You skip the birthday dinner, the beer-and-pizza night. The most calorie-dense days of the month systematically vanish from the record, and the diary starts flattering you — the average reads low exactly where it matters most.

Log the guess. Wrong beats missing.

Three tiers of estimating

Tier 1 — components you recognize. Grilled chicken with a rice side is countable: 150–200 g of meat, 150–200 g of cooked rice, plus oil. Oil is the part everyone forgets — assume 10–15 g (90–135 kcal) for anything fried, sautéed, or dressed. Restaurants are not shy with it. Drinks count too: a glass of dry wine is about 120–150 kcal, half a liter of beer about 200–250.

Tier 2 — the whole-dish guess. A plate of pasta: 600–800 kcal. A burger with fries: 900–1,100. A soup: 200–400. Round up in restaurants; portions and oil run bigger than at home.

Tier 3 — one number. Calsheet has quick-add for exactly this: plain calories, no product needed. Put the context in the entry note — “pizzeria, 3 slices” — so future-you knows what the number meant and can adjust it.

Which tier to use? The lowest one that doesn’t stop you from logging. Tier 3 exists because the meal where you argue with the app is the last meal you log.

Backfill the same day, not “someday”

Dinner at a friend’s is not a logging environment — don’t be the person typing into a food diary at the table. Log it the next morning instead: the diary in Calsheet flips back day by day, so you open yesterday and reconstruct it from memory in a minute over coffee.

Memory does this job fine at 12 hours and fails at 72. Same-week backfilling is accurate enough; “I’ll reconstruct the weekend on Wednesday” is fiction.

And if you’ve connected Calsheet to Claude (the claude.ai connector), backfilling is one sentence: “yesterday’s dinner — borscht, chicken with rice, a glass of wine” becomes diary entries with macros. It needs a signed-in account, and it shines on exactly this case: a multi-item meal you’d rather describe than itemize.

Your repeat spots deserve ten minutes

The business-lunch place you visit twice a week isn’t an estimation problem — it’s a one-time data-entry problem. Create a custom food for your usual order with your best component estimate, refine it over a few visits, and from then on it’s a two-tap log like everything else in your library.

How to know your guesses are good enough

You have a built-in auditor: the weight journal. If the 30-day stats say you’re on budget while the scale trends the wrong way for a month, your estimates run low — most people’s do. The fix is mechanical, not moral: bump your restaurant guesses by about 20% and keep going.

And if the numbers and the scale agree — congratulations, your eye is calibrated. That skill came from the boring weeks of scanning and weighing.

Nobody’s diary is exact. The good ones are merely complete.