Restaurant Inventory Software: What to Automate and What to Count by Hand

Good restaurant inventory management splits the work in two. Software should do the arithmetic. People should do the counting that checks it.

The mistake is to buy inventory software hoping it removes the count. It does the opposite: it makes the count more valuable. Software can only tell you what stock should be. A person with a clipboard, or a tablet, tells you what is.

What can software do better than people?

Software is better at repeated sums, and worse at anything that needs eyes on the shelf. Four jobs are worth automating early because they are repetitive, easy to check and rarely go wrong in an interesting way.

• Invoice entry. Reading supplier invoices and updating ingredient prices saves hours and keeps your costs current.
• Recipe costing. When a supplier price changes, every dish that uses the item should update.
• Par levels and order lists. Software can compare stock with the level you want and suggest what to order.
• Reports. Expected stock, usage by period and cost percentages should appear without a spreadsheet.

Operators are already moving this way. In Deloitte's 2025 survey of 375 restaurant executives in 11 countries, 55% of respondents said they use AI for inventory management daily. That figure covers AI applications specifically, not inventory software in general, and it reflects executives' own reports, so read it as a sign of interest, not proof of results.

What can only a person do?

Only a person can see, taste, weigh and doubt. Four jobs should stay with people because they are the checks on everything else.

1. The physical count. A system computes expected stock from deliveries, sales and recipes. If any of those is wrong, expected stock drifts. The count is the reality check.
2. Receiving. Someone must weigh the fish, check the date codes and count the cases against the delivery note. A scanned invoice only proves what was billed.
3. Judging quality and waste. Software cannot tell that the herbs are tired or that a station is over-portioning.
4. Deciding what a variance means. The number tells you where to look. A person finds out why.

Software tells you what you should have. Only a count tells you what you have.

What should you count, and how often?

Count by value and risk, not equally. A simple tiering keeps the effort proportionate.

Treat the rhythm as a starting point and adjust it to your business. The goal is to spend most counting time on the few items that drive most of the money, which is the same logic as a menu where a handful of dishes carry the profit.

A few habits make counts more reliable:

• Count in shelf order. Set up the count sheet in the same order as the storeroom, so nothing is missed or counted twice.
• Count blind where you can. If the counter sees the expected figure, they tend to confirm it. Hide it and compare afterwards.
• Use two people for tier A. One counts, one records. It is quicker, and it discourages shortcuts.
• Record units the same way every time. A case, a kilo and a portion are not interchangeable. Many early variances in a new system turn out to be unit mistakes.
• Count at the same time. After close, before the next delivery, on the same day each week.

How do you turn the count into a useful number?

Compare what you expected with what you counted, and look at the gap by item. That gap is variance, and it is the most useful number inventory work produces.

Here is how it works, in an illustrative case. Imagine a bistro whose system expects 12 kg of a certain cut of beef on Monday night, and the count finds 10.5 kg. The shortfall might be a wrong recipe quantity, extra trim waste, a delivery that was short, or portions running large. Start with the cheapest explanations to check: was the delivery weighed, and was the recipe quantity right? Then look at waste logs and portion checks.

Do this for the tier A items each week and a pattern appears. One item is always short. Fix that, and the effect on your margin is clearer than any broad drive to cut costs. Our guide to food cost percentage shows how the overall figure is built, and seven numbers to know every week puts variance in context with the rest of your dashboard.

Two cautions. Small variances are normal, so set a tolerance and do not chase noise. And a variance without a recorded reason teaches nothing, so write one line next to every large gap.

Why is waste the part you should still log by hand?

Waste is easy to see and hard to automate, so a simple hand log pays off first. The Champions 12.3 study of 114 restaurants in 12 countries, published in 2019, found that kitchen food waste fell 26% on average within a year, and that on average each restaurant saved $7 for every $1 invested. Every site kept its total investment below $20,000. The report credits a food waste inventory with helping restaurants see how much they wasted and where.

Those are results from sites that chose to take part, so do not treat them as a promise. The useful point is the method. A waste log with a reason code (trim, spoilage, plate return, error, staff meal) turns an invisible loss into a list. If you can log it for two weeks, you can decide whether you need a scale or a camera. If you cannot, a tool will not rescue the habit. For ideas on what to do with the findings, see ten ways to cut restaurant food waste.

How do you choose inventory software without regrets?

Choose for fit with your counting habits, not the length of the feature list. Ask these questions during a trial:

• Does it connect to your POS and suppliers' invoices, and what does it do when they disagree?
• Can you count on a phone or tablet in the storeroom, offline if needed?
• Can you set units and pack sizes your way, including conversions between case, kilo and portion?
• Does it show variance by item and by week, with room for a note?
• Can you export your recipes, items and count history if you leave?

Several of these echo the checks in our guide to choosing a restaurant POS system. Start small: items and invoices first, recipe-level tracking once your recipes and units are tidy.

Key takeaways

• Automate the arithmetic: invoices, costing, par levels, order suggestions and reports.
• Keep the physical count by hand, because it is the only check on what the system believes.
• Count by value: weekly for expensive and fast-moving items, monthly for low-cost ones.
• Count blind, in shelf order, with consistent units, and at the same time each week.
• Track variance by item, add a one-line reason, and chase patterns rather than noise.
• Log waste by hand first, then decide whether a scale or camera is worth it.