Should we add a fourth shift?
A cafe manager pastes the situation and attaches twelve weeks of roster data. One question turns into three numbers and a decision the owner can check.
Everything in the grey boxes is what Learning Nodes returned in a real run, selected and unedited. The cafe and its roster file are made up for the example.
INTERACTIVE VERSION · TWO MINUTES
Click through a cafe manager's session in the Workstation, one step at a time. A short note explains each screen. No sign-up, and nothing is saved.
The situation, as typed
The decision Learning Nodes proposed
Choose between adding the shift, trialing it for four weeks on the two busiest days, and leaving the roster as it is
- Add the shift
- Trial it for four weeks on the two busiest days
- Leave the roster as it is
You have to tell the owner by Friday whether to add the shift, trial it for four weeks on the two busiest days, or leave the roster as it is. It turns on how the close shift takings compare with open and lunch and on which days show the biggest gap.
What would make each option right
Add the shift
Close shift takings are lower than open and lunch on more than two days a week and the two staff who cannot do afternoons can be kept off those shifts.
Trial it for four weeks on the two busiest days
Close shift takings are lower than open and lunch on exactly two days a week and a four-week trial on those days would show whether the extra shift lifts takings enough to cover 260 dollars a week.
Leave the roster as it is
Close shift takings are not lower than open and lunch on any weekday and the two staff who cannot do afternoons would have to be replaced.
The questions in the way
- What the close shift takes against the open and lunch shifts
- Which days are busiest
- Which two days show the largest gap between close shift takings and the other shifts?
Claims nobody had checked
The owner wants a fourth, shorter shift from 3 to 6 pm because she believes we lose afternoon customers when only one person is on.
If the owner is wrong about lost customers, the case for adding or trialing the shift disappears.
Cheapest test: Ask the owner to name the afternoons where she saw customers leave because only one person was on.
Two of the staff have told me they cannot do afternoons at all.
If those two cannot be kept off the new shift, adding it forces either overtime or new hires.
Cheapest test: Confirm with the two staff in writing which shifts they can and cannot cover.
Adding the shift would cost about 260 dollars a week in wages.
The trial and add options both rest on whether the extra takings cover this weekly cost.
Cheapest test: Check the roster file for the exact hours and rates that produce the 260 dollar figure.
Answers from the data
The manager attached the roster file and pressed Answer from data on the two questions that needed numbers. Learning Nodes wrote the queries from the column names, ran them in the browser, and showed each answer next to the queries that produced it.
What the close shift takes against the open and lunch shifts
Your close shifts average 521 in takings against 855 for lunch and 603 for open across the 252 shifts. Adding the shift would expand the lowest-performing option. Trialing it for four weeks on the two busiest days would place it on Friday and Sunday which average 717 and 715. Leaving the roster as it is keeps the current balance where close trails the other two. Every shift type appears exactly 84 times so the averages rest on equal counts, yet the file cannot address costs or demand changes beyond these takings figures.
SELECT AVG(takings) AS close_avg_takings FROM data WHERE shift = 'close' SELECT shift, AVG(takings) AS avg_takings FROM data GROUP BY shift ORDER BY avg_takings DESC LIMIT 10 SELECT weekday, AVG(takings) AS avg_takings FROM data GROUP BY weekday ORDER BY avg_takings DESC LIMIT 7 SELECT shift, COUNT(*) AS shift_count FROM data GROUP BY shift
roster_last_12_weeks.csv, 252 rows
Which days are busiest
Friday leads all days with average takings of 717.17 across the 36 shifts recorded for each weekday. Adding the shift would place it on the single highest day by that measure. Trialing it for four weeks on the two busiest days would cover Friday at 717.17 and Sunday at 715.39. Leaving the roster as it is would retain the existing pattern where the next days sit at 711.19 and below. Every weekday carries an identical shift count so the averages rest on equal samples.
SELECT weekday, ROUND(AVG(takings),2) AS avg_takings FROM data GROUP BY weekday ORDER BY avg_takings DESC LIMIT 1 SELECT weekday, ROUND(AVG(takings),2) AS avg_takings FROM data GROUP BY weekday ORDER BY avg_takings DESC LIMIT 7 SELECT weekday, COUNT(*) AS shift_count FROM data GROUP BY weekday
roster_last_12_weeks.csv, 252 rows
Bring your own file: attach a CSV to any situation and press Answer from data.
What the manager committed (illustrative)
Trial the afternoon shift for four weeks on Fridays and Sundays
Cafe manager
- Close shifts average 521 against 855 for lunch and 603 for open, over 252 shifts
- Friday (717.17) and Sunday (715.39) are the two busiest days
- The trial has to lift takings by more than 260 dollars a week to pay for itself
- Open: the owner has not yet named the afternoons where customers left
Frozen when committed. If the trial changes the picture, the next decision is a new record that points to this one.
See the same case in the product, one step at a time.
The owner asked for a shift. The manager came back with a smaller, cheaper test, the two days to run it on, the number it has to beat, and the one claim still unchecked.
