Choose five measures to run a college canteen week by week
A canteen manager can count dozens of things, yet only a few of them change what gets cooked, bought or staffed.
A measures sheet with five chosen measures, how each is counted, the decision each one informs and a threshold that would trigger action, plus a one-screen mock dashboard in slides using made-up data.
This is a dashboard project. Turn a table into a few charts that answer the questions someone asks every week.
The approach: List the three questions the dashboard must answer → Prepare one clean table → One chart per question → Add a line saying what each chart shows.
Think first
A canteen manager decides things every day: how much rice, when to open a second counter, what to stop buying. You are choosing just five numbers that would change those decisions, with a clear rule for when each one says 'act'.
- Which decisions did you actually see staff make during your lunch-hour visit?
- For each decision, what single number would tell them to act, and how could it be counted with a pen and a register?
- Who would count it, when, and how long would it take?
- Which measures will you drop, and what is the reason for each?
A measures sheet with five rows, each linking one measure to one decision, how it is counted, who counts it and an action threshold, plus a dropped-measures list and a one-screen mock dashboard clearly marked as using made-up data.
Picking measures that sound impressive, like customer satisfaction scores, that nobody in a busy canteen will ever record. If a worker cannot count it daily without extra tools, it does not belong.
Gather
- Notes from one lunch hour at a canteen: every decision you saw or heard staff make
- A short chat with one staff member about what they check and what goes wrong, with their permission
- Your own made-up data for the mock dashboard, typed by you and labelled as invented
AI must not invent decisions the staff make, or present any number as the canteen's real sales or waste. Decisions come from your visit; dashboard numbers are yours and labelled as made up.
Build with AI
- 3.1 Turn visit notes into a decision listCLAUDE CHAT
Why this tool: Claude can sort rough notes into a clean list of decisions without adding new ones.
Paste your visit notes and ask for a list of distinct decisions only. Delete anything you did not actually observe.
- 3.2 Propose and stress-test measuresCLAUDE CHAT
Why this tool: Asking what could go wrong with counting each measure exposes the ones nobody will record.
For each decision, ask for candidate measures and their counting problems. Keep only measures a worker could record daily.
- 3.3 Save a measure-review SkillCLAUDE SKILLS
Why this tool: A saved checklist applies the same tests to every candidate measure, so your choice is consistent, not a mood.
Write a short Skill that checks each candidate measure for: one linked decision, a hand-countable method, a named counter, a threshold. Run it on your full candidate list and read every verdict before accepting it.
Before you let it act: Before running it, read the Skill's instructions and make sure the four tests are the ones you want. Check its verdict on two measures you already judged yourself.
- 3.4 Build the one-screen mock dashboardGOOGLE SHEETS
Why this tool: A sheet with a few small charts is enough to show how the five measures would look on one screen.
Type two weeks of made-up daily values yourself, label them as invented, and draw one small chart per measure with the threshold drawn as a line. Copy the charts onto one slide.
Check the AI's work
- Each of the five measures links to a decision you actually observed
- Each counting method needs only a register, a pen or a simple tally
- Every threshold says what action follows, not just 'review'
- The dropped-measures list gives a reason for each drop
- The dashboard data is labelled as made up
- Totals on the dashboard match the raw table
- Every axis is labelled with its unit
AI likes generic restaurant metrics such as average ticket size or table turnover that do not fit a college canteen. Check each suggestion against what you saw at the counter.
Finish and show
- Five measures with decision, method, counter and threshold
- A list of dropped measures with reasons
- A one-screen mock dashboard with threshold lines
- A clear label that the dashboard data is invented
Page one is the measures table; page two is the dashboard slide. A canteen manager should see which number to watch for each decision in under a minute.
Describe one measure you dropped even though it sounded good, and why a busy canteen worker would never have counted it.
Which tool fits
- Claude chat: Claude can sort rough notes into a clean list of decisions without adding new ones.Free plan works; Pro gives longer conversations.
- Claude Skills: A saved checklist applies the same tests to every candidate measure, so your choice is consistent, not a mood.Pro or above, with code execution switched on.
- Google Sheets (free alternative): A sheet with a few small charts is enough to show how the five measures would look on one screen.Free with a Google account.
- Power BI (free alternative): an interactive dashboard with filtersFree desktop app (Windows).
- Claude for Excel: shaping the data and checking totalsPaid plans; needs Excel.
Turn your measure-review Skill into a reusable kit for any small food outlet, and test it on a tea stall or a mess to see which of your five measures survive.
AI helps you think, check and polish. The work, and the judgement, stay yours.
This is a free sample. Full Access opens every project in full and lets you submit your finished work.
See Full Access →