Meet Tsuno at Sirha Lyon 2027 · Stand S.57Learn more
EN FR
Try the demo
Menu

Costs and losses

Measuring a pilot in 30 days: three numbers, not thirty

Before day one, choose three measures, record their baseline and write down the decision. At day 30, you continue, adjust or stop without a debate.

Published on

In a restaurant office, three sticky notes in different colours on a switched-off screen, with a mug next to it.

You measure a pilot with three numbers chosen before day one, not with thirty readings pulled together afterwards. For example: the manager's interruptions for a procedure question, the date of the first opening or closing done alone and signed off by a mentor, and voluntary use of the tool at day 14. Record their baseline the week before, compare with a site or routine without the tool if you can, and write down from the start what you will do at day 30: continue, adjust or stop.

Monday, 10 am, in front of a blank screen

Benjamin owns two restaurants. On Wednesday he starts a trial of a new tool in the first one. The vendor has sent him a demo dashboard: twenty-eight metrics, charts, percentages.

Océane, his assistant manager, has switched the screen off. She has stuck three sticky notes on it. A yellow one: "How many times a week you get interrupted for a procedure". A green one: "Date of the first closing done alone and signed off". A pink one: "At day 14, who uses it without being asked". Under the three, she has written in marker: "Day 30, we decide."

"We'll look at the rest," she says, "if these three move." Benjamin puts his mug down next to the keyboard. He does not switch the screen back on.

(Fictional scene and characters.)

Why thirty metrics kill a pilot

Nobody records them

In the first week, everyone takes notes. By the second, service takes over again. By the third, two columns out of thirty are still being filled in, and they are not necessarily the right ones.

There is always a green number

With thirty measures, at least one improves by chance. Whoever wants to keep the tool shows that one; whoever wants to drop it shows another. The pilot no longer settles anything: it feeds an argument.

Noise drowns the signal

In a restaurant, a month's numbers move for a thousand reasons: a rainy week, an extra pair of hands, a dish that changes. Across thirty metrics recorded in one site, a lot of the movement is noise. Three measures tracked closely are easier to read.

The team feels watched

Thirty readings means a team that feels scrutinised on everything. It starts working for the number, or it goes quiet. A pilot measure should never be about a person: you count routines, procedures, days.

The 30-day protocol

The 60-day test tells you whether a tool has become a habit. The first 30 days are for something else: deciding whether it is worth continuing. Here is how it runs.

  1. Write the pilot sheet before day one. One page: the chosen routine, the problem it causes today, the three measures, their benchmarks, the review date and the decision planned for each case. The owner and the manager sign it.
  2. Choose three measures, no more. Each one must be recordable without the tool under test, with a notebook or a sheet. If a measure only exists in the vendor's dashboard, you will not be able to compare it with the baseline.
  3. Record the baseline. In the week before day one, record the three measures under normal conditions. Date the sheet.
  4. Set up a comparison group if you can. A second site that keeps its usual ways, or a routine in the same site that is left as it is (opening with the tool, closing without). Failing that, write down each week what else has changed.
  5. Tell the team. Say what will be recorded, and what never will be: no names. In France, the Labour Code (Article L1222-4) forbids collecting personal information about an employee through any system they have not been told about beforehand; elsewhere, telling people up front is simply the honest way to run a trial.
  6. Conclude nothing before day 14. The first week measures curiosity. At day 7, you only fix practical problems.
  7. Hold the review at day 30. Record the three measures, compare them with the baseline and the comparison group, read the decision written on the sheet, and apply it.

Three possible measures

These are examples: choose the ones that match the problem the tool is meant to solve.

  • The manager's interruptions for a procedure question, over a typical week. You count topics, never names. The logging method is in how much time a manager loses answering the same questions.
  • The date of the first opening or closing done alone and signed off by a mentor, for someone who joined the role during the pilot. Compare it with the time it took for previous arrivals in that role: you are judging the onboarding path, not the person. See how to know when a new restaurant employee is actually ready.
  • Voluntary use at day 14. The number of days, out of the last seven, when the tool was used on the routine without a reminder. You count days, not people.

The pilot sheet, in one table

The benchmarks below are examples to discuss, not standards. They come from no study. The point is to write them down before day one.

MeasureHow to record itBaselineSuggested benchmark at day 30Who records it
Manager interruptions for a procedureLog of a typical week, by topicWeek before day oneDown on the tested routine, compared with the baseline and the comparison siteThe manager
First opening or closing done alone and signed offDate of arrival in the role, date of sign-off by the mentorTime taken by the last arrivals in that roleAt least the same, ideally shorterThe mentor
Voluntary use at day 14Days ticked without a reminder, out of sevenNot applicableThe tool is used on most days on the routineOne named person, recording no names

One rule sits on top of the three lines and is a deal-breaker on its own: if the team says it feels monitored and a clear answer has not brought that feeling down, you stop.

The decision, written in advance

  • Continue: all three benchmarks are met, or two out of three with a written reason for the third. You extend the follow-up to the second month.
  • Adjust: only one measure moves. You change the chosen routine, rewrite the procedures or train differently, then run another thirty days.
  • Stop: no measure moves, or the deal-breaker applies. You tell the team and the vendor, with no regrets.

Common mistakes

  • Choosing the measures after the pilot. You then keep the ones that suit you.
  • Measuring what the tool measures itself. The vendor's dashboard has no baseline in your restaurant. Keep measures you can record in a notebook.
  • Counting per person. The pilot turns into a check on people, the team changes its behaviour, and the numbers no longer say anything about the tool.
  • Forgetting employee representatives. In France, from 50 employees upward, a pilot can already count as implementation (Nanterre court, interim ruling of 14 February 2025); elsewhere, check local rules. See should you consult employee representatives before using a voice assistant.
  • Pushing back the review. "Let's give it two more weeks" is a decision not taken. The day 30 date is on the sheet.
  • Testing where there is no problem. A tool gains you nothing on a routine the team already has down.

What to measure

The three numbers on the sheet are enough, read from three angles.

  • The gap between the baseline and day 30 for each of the three measures.
  • The gap between the tested site or routine and the comparison group, where there is one.
  • Whether the timetable itself was kept: did the review happen at day 30, and was the written decision applied?

What we offer for a Tsuno San pilot

Tsuno San is the smart voice assistant for operational teams, connected to the Tsuno environment. It evaluates no one, which suits a pilot measured without names. San Voice is available as a demo; San Audio and San 360 are still being explored. On how actions run, see the two-step rule.

For a pilot, here is what we offer as a basis for discussion, not as a contractual offer: write the sheet and its three measures together before day one, on your own routines; work with a team that has volunteered and been informed; hold the review on the planned date; keep the freedom to stop. Some dates already exist in Tsuno: in Tsuno Compétences, skill levels are validated and dated by a manager, and opening and closing checklists are done on the tablet with their history.

What a good tool should do, and what we are working on: record usage per routine and questions per procedure, without ever tying them to a name. It is not an established feature today. The sheet and the three sticky notes are enough to run the pilot, with San or with any other tool.

To start smaller, the trial based on three non-sensitive operational documents runs in an isolated environment that is deleted on the agreed date: the steps are on the setup page. Our data rules are on the trust page, and the assistant is presented on the Tsuno San page.

Further reading

To put a money figure on what the tool saves, go back to how to measure restaurant software ROI in 90 days. To then track how the team is learning, see independent, not dependent: measuring that the team is learning. And on the resistance a pilot brings to the surface, read why teams reject new tools. The framework is in the pillar page AI for frontline managers. For an honest baseline, see restaurant staff turnover: which French figures to use, and for the mindset, why we show our failed trials.

Key takeaways

A pilot is decided on day one: three measures, a baseline, a review date and a written decision for each case. At day 30, you read the sheet and do what it says, without hunting for the thirty-first number that would suit everyone.

Frequently asked questions

How many metrics should you track during a restaurant software pilot?

Three. Beyond that, nobody records them after the first week, and there is always a favourable number to justify any decision. Three measures chosen before the start, with their baseline, are enough to decide at 30 days.

Which measures should you pick for a 30-day pilot?

Work measures you can record without the tool: the manager's interruptions for a procedure question, the date of the first opening or closing done alone and signed off by a mentor, and voluntary use of the tool at day 14. Adapt them to the problem the tool is meant to solve.

What is the baseline of a pilot?

It is the record of your three measures in the week before day one, before the tool is switched on. Without it, you compare against a memory. Write it on a dated sheet and keep it with the pilot sheet.

Do you need a comparison group for a pilot in a single restaurant?

It helps when you can have one: another site that does not use the tool, or a routine in the same site that is left as it is. Without one, compare against the baseline and note what else changed (season, headcount, menu) so you do not credit everything to the tool.

What should you decide at the end of a 30-day pilot?

One of the three answers written before the start: continue if the benchmarks are met, adjust the routine or the procedures if only part of it moves, stop if nothing moves or if the team feels monitored. The decision is read off the sheet, not negotiated on the day.

Is a 30-day pilot enough to adopt a tool?

It is enough to decide whether to continue. It is not enough to know whether the tool will become a habit: the curiosity of the first weeks can still carry usage. If you continue, extend the follow-up to the second month.