Search "ordering from memory restaurant" or "restaurant keeps running out of ingredients" and the answers mostly sell you inventory software: get the order into a system and the problem stops. Sometimes that is the right last step. But owners who live with this rarely describe two separate problems. On one day the kitchen runs out of an ingredient in the middle of service, and a dish comes off the menu. On another day the walk-in holds more perishables than the week can use, and some of it goes in the bin. This page is the cost audit version of that problem: why the two failures are one failure, what the published sample behind these Notes shows about it, what can honestly be priced, and the count you can run on your own orders to get your own number.
The diagnosis: gut feel ordering fails in both directions
Owners have their own words for this, and the words are better than the consultant version. The first is ordering from memory: the order is built from what the person ordering remembers — what ran low last time, what felt short, what the last busy day looked like — instead of from a count of what is on hand and what the coming days actually need. The second is gut feel ordering: the quantities are a judgement call made standing in front of the shelves, with no written par, no usage count, and no record of what last week's order turned into. Neither phrase describes carelessness. For a small kitchen the person ordering usually knows the food better than anyone. The failure is that the knowledge lives in one head, on one day, and the day changes what the memory says.
Now the part that makes this problem hard to see, and it is the whole point of this page: the pattern is the same in both directions, because under-ordering loses sales and over-ordering becomes spoilage. Under-order by a little and the cost shows up the same day, in service: an ingredient runs out, a dish you cannot serve comes off, a customer is told no, or someone leaves the kitchen mid-service to buy what is missing. Over-order by a little and the cost shows up days later, quietly: perishables bought for a busy week that never came are thrown away, and the invoice for them was paid as if they had been sold. Because the two failures happen on different days, owners never see them as one problem. The shortage day gets blamed on a busy service. The spoilage day gets blamed on a slow week. Neither day gets blamed on the order, which is the one thing both days have in common.
That is also why "how to stop over ordering stock" is only half a search. The owner who tightens the order to cut spoilage gets more shortage days. The owner who pads the order to protect the menu gets more spoilage. Ordering harder in either direction does not fix a method that runs on memory, because memory has no record of its own misses: last week's spoiled case is forgotten by the time this week's order is written, and last week's stockout is remembered as a reason to order more of everything, not more of the one thing that ran out. What fixes the seeing comes first, before any fix for the ordering itself: count both sides of the same order, on paper, for a few weeks. That count is the audit, and it is at the end of this page.
The sample problem, in the business's own words
The published sample these Notes are built on is Taste & Thyme Catering, a synthetic example whose sample plan is published on this site. Its weekly catering operations workflow names four problems, and the second of them is this one, written the way the business talks about it: Ordering from memory, which produces ingredient shortages on some days and overbuys on others.
Read that sentence the way it is built. It does not name two problems. It names one habit — ordering from memory — and two results, shortages and overbuys, as products of the same habit. A caterer feels this version sharply because the order is written against events, not against a steady menu: each event has its own dishes, its own guest count, and its own venue, so what last week needed is a poor guide to what this week needs, and a memory trained on last week's event orders confidently for the wrong week. A restaurant with a fixed menu feels the same habit more slowly, as a standing order that never quite matches the week: the same ingredient short on Fridays, the same perishable over-bought every Monday. The shape is identical. The order is a recollection, and recollections do not come with counts attached.
The sample also shows where this problem lives in the workflow's time. The ordering work — deciding, writing, receiving, checking, fetching what was missed — sits inside the blocks of the week priced in the next section, alongside the booking, staffing, and invoicing work. The problem sits inside those blocks, which is exactly why pricing the problem on its own is harder than it looks.
What ordering from memory costs, and what this page will not invent
It would be tidy to put a price on the ordering problem itself: so much for a shortage, so much for a spoiled case. This page does not do it, for the same plain reason as the other diagnostics in these Notes. The sample does not break out a separate cost per problem, it does not price one stockout, and it does not price one spoiled case. Its published figures price the blocks of the workflow, and the ordering sits inside those blocks alongside the rest of the work. Splitting out a per-problem ordering price would mean guessing at a split the sample never made.
What can be priced exactly is what the sample does publish, and it is still useful:
- The whole workflow — booking in, event staffed, food ordered, invoice out — takes 18.62 hours a week and costs $722.31 a week, which is $37,560 a year, priced block by block from the hours and rates stated in the sample.
- The coordination block — 12 hours a week at $30 an hour — is $360 a week, and $18,720 a year. That is the largest block of the week, the work inside which orders are written, received, and chased.
- The owner's own block — 5 hours a week at $60 an hour, $300 a week, $15,600 a year — is where the worst misses land, because when an ingredient runs out mid-event, the owner is the one who decides what comes off the menu.
Then there are your own numbers, and there are two of them, because this problem has two sides. For the time side, the formula is the whole method, in public: cost = your hours x your rate x how often it happens. Count the hours a week spent ordering and fixing orders — the emergency runs, the mid-service substitutions, the checking — and price them at the rate of the person doing them. For the spoilage side, no formula is needed: count what you threw away, at what it cost you, and add it up. Here is that line worked through with example inputs, so the shape is clear. The inputs are examples — replace them with your real ones:
- Say that in one week you throw away $40 of perishables that were over-ordered.
- That is $40 a week, which is $2,080 a year — on the spoilage line alone, before any shortage day has cost anyone a dish they could not serve.
Notice what that shape does not need: anyone else's average. A spoilage tally is your own count of your own bins, priced at your own invoice costs. Your number might be smaller. It might not. The audit below is how you find out with your counts instead of a guess, and the shortage side of the count matters just as much, because a week with no spoilage and three stockouts is not a good ordering week — it is the same problem, showing its other face.
What the named fix costs
The sample plan names a fix for ordering from memory, and its shape follows the diagnosis: the fix is not a better memory, it is an order built from a count. The sample names MarketMan, listed at $2,988 a year.
Read that price with the diagnosis in mind. What the price is weighed against is not a per-problem cost this page refused to invent — it is your two counts from the section above: the spoilage tally, priced at cost, and the ordering-and-fixing time, priced with the formula. What a counted order buys, used this way, is that the quantity for each ingredient is built from what is on hand and what the coming events need, instead of from memory of what the last event needed. Whether that is worth its price in your business depends on how large your two counts are — and on whether they are large on both sides, which is the signature of this problem rather than of a single bad week.
One more honesty note: this page does not claim any tool will end shortages or spoilage in your kitchen. A counted order ends the guessing. It does not stop a guest count changing the day before an event — which is why the count matters more than the shopping.
The audit: count both sides of the same order
You do not need a new system to run this audit. You need your next few orders and three tallies, kept separately, because they price differently:
- The shortages. Every time an ingredient runs out before the service it was ordered for — the dish that came off, the substitution that was made, the item someone had to go and buy. Note which ingredient, and which day of the order it ran out on: shortages late in an order usually mean the quantity was short, not the week.
- The spoilage. Every perishable thrown away at the end of its life, written down at what it cost when it was bought. This tally is already money — add it up weekly, and do not round it into a general sense of waste.
- The emergency runs. Every trip or rush delivery to replace something the order missed, with the time it took and who took it. This is the time side of the formula above, and it is the tally owners most often leave out, because the run feels like solving the problem rather than being part of its cost.
At the end of a few weeks you have the numbers this page is built on: a spoilage total priced at cost, a shortage count with the ingredients named, and the fixing time priced with the formula. If the tallies come back near zero on both sides, ordering from memory is working in your kitchen and the honest answer is to keep it. If one side is heavy, the order is skewed, and the count says which way. If both sides are heavy in the same few weeks, you have the finding this page started with — one problem, not two — in your numbers, not in a vendor's average.
The worked examples behind this page
Four pieces in these Notes show the rest of this method with every number visible. The group-text scheduling audit, "Scheduling staff by group text: count what the chaos costs", at https://www.agenticartists.com/blog/scheduling-staff-by-group-text works through the sample's third problem, the shift swaps that never reach the event sheet. The double booking diagnostic, "How to stop double booking: fix the record, not the calendar", at https://www.agenticartists.com/blog/how-to-stop-double-booking works through the sample's first problem, the booking record that drifts. The teardown, "What a catering workflow costs: the Taste & Thyme teardown", at https://www.agenticartists.com/blog/what-a-catering-workflow-costs pulls the sample workflow apart block by block, including the fixes named for all four problems. The method guide, "How to audit one workflow", at https://www.agenticartists.com/blog/how-to-audit-one-workflow walks through running the same audit by hand, on a workflow of your own, step by step.
Doing this with Where AI Fits
Where AI Fits does this audit with you, for one workflow: a guided conversation reconstructs the workflow, you confirm the steps, the hours, and the rates, and the paid AI Integration Plan, at $49, writes up the time and cost block by block, names the problems in your words, and puts a named fix and its catalog price next to each one. If your open question is the ordering chain — how the order is built, what it is built from, and where its misses show up — the sample plan is published on this site, so you can read the finished document and check its numbers before deciding.
Sources
Every number on this page traces to the studio's own published sample or catalog, or to a worked example whose inputs are labelled as examples. No external statistic is used on this page, and no per-problem ordering cost, no cost for one stockout, and no cost for one spoiled case are used, because the sample prices none of them. Taste & Thyme Catering is a synthetic example, labelled as such on the sample plan itself; its figures are the sample's own, not measurements of a real business.
- Taste & Thyme Catering sample plan — $722.31 a week, $37,560 a year, 18.62 hours a week, and the problem recorded in the business's own words as ordering from memory, which produces ingredient shortages on some days and overbuys on others — https://www.agenticartists.com/products/where-ai-fits/sample-plan — checked 2026-10-30
- The Taste & Thyme teardown in these Notes — block pricing (coordination 12 hours a week at $30 an hour, $360 a week, $18,720 a year; owner 5 hours a week at $60 an hour) and the sample's named fix for ordering from memory — https://www.agenticartists.com/blog/what-a-catering-workflow-costs — checked 2026-10-30
- Where AI Fits plan catalog prices, as published in the shipped catalog — MarketMan $2,988 a year — checked against the catalog on 2026-10-30
- Where AI Fits — $49 for the AI Integration Plan for one workflow, published on this site at /products/where-ai-fits — checked 2026-10-30
Ready to look at one workflow?
Start with the Where AI Fits sample plan, or learn more about the product and its workflow-first approach.