Common Hour is a Melbourne daytime cafe group. Four venues, 35 staff, about $5m a year. In July 2026 it turned over $411,408 and returned $22,111.
Labour ran at 32.8% of sales, wages only. The ATO’s benchmark band for coffee shops turning over more than $600k is 27% to 35%. Comfortably inside. Loaded with on-costs it was 37.9%.
Revenue was up 3.1% on June and 32.3% on the same month last year. Cost of sales was 34.3% against a band of 33% to 38%. Read the group P&L top to bottom and nothing asks a question.
One of the four venues lost $3,697 that month.
The growth was one venue, and not the profitable kind
Start with the 32.3%. Group revenue grew $100,475 on July 2025. Glen Waverley, which opened in September 2025, accounts for $92,172 of it.
The three venues that existed a year ago added $8,303 between them. That is same-store growth of 2.7%.
So the headline number describes an opening, not a year of trading. A group that opens a site every two years can post a number like 32.3% indefinitely while the business underneath it moves 2.7%, and the two get read as one figure.
The move: split last year’s growth into same-store and new-site before you call the year good. They are different achievements and only one of them repeats.
The split a group P&L cannot show you
Contribution by venue, July: Hawthorn 9.0%, Yarraville 8.4%, Coburg 6.5%, Glen Waverley minus 4.0%.
In dollars: $11,427, $8,251, $6,130, and minus $3,697. Those four add to the group’s $22,111.
Glen Waverley against the other three combined is a gap of 12.1 points, because the three together run 8.1%. Labour with on-costs is 83% of that gap. Rent is another 12%. Cost of sales is almost none of it, which matters because cost of sales is where most owners look first.
This is the whole point. The group figure of 32.8% is inside the band because three venues are carrying one. An average is not evidence about its parts, and a four venue group has exactly four parts. Nothing in a consolidated P&L is capable of showing this, however carefully it is read.
The move: compute contribution per venue, not per group. If you cannot say today which of your sites made money last month, the group figure is not reassuring you, it is hiding the answer.
The hours are in the right shape. There are too many of them.
The obvious next question is what Glen Waverley is doing wrong, and the obvious guess is that the roster is pointed at the wrong hours.
It is not. Glen Waverley spreads its hours across the day in a pattern that already matches its own trade. The daypart shape is right.
The level is wrong. The site runs 975.6 labour hours, which is 91% of Hawthorn’s hours, on 73% of Hawthorn’s sales. It earns $94.48 per labour hour against $116.38 at Coburg, $118.57 at Hawthorn and $124.13 at Yarraville.
There is one exception worth naming, because it is the cheapest thing to fix. Weekend hours ran 44.6% above roster in July. Weekdays ran 1.7% under. The weekend overrun alone is worth about $2,933 a month, and it is call-ins drifting past plan rather than a staffing decision anyone made.
The move: check the level before you reshape the roster. Moving hours around a site that has too many of them produces a better looking timetable and the same result.
It stopped being a ramp in May
The reasonable objection to all of this is that new sites cost more and settle down. Glen Waverley opened eleven months ago. Give it time.
Its labour percentage by month since opening: 65.6, 58.3, 55.4, 52.8, 52.5, 47.1, 42.3, 41.3, 39.1, 40.3, 39.5. Eight consecutive falls to 39.1% in May, then June at 40.3 and July at 39.5.
The improvement was real and it has stopped. It is also worth understanding what produced it, because that decides whether waiting works. The site’s sales grew from $51,362 to $97,609 by May while the roster stayed roughly where it was set before opening. Every point of that improvement came from sales rising into fixed hours.
Sales then stopped rising. June and July both came in under May. So the mechanism that delivered eight months of improvement has run out, and the site is still 4.5 points above the top of the ATO band and 8.7 points above its three siblings.
The move: when a new site’s numbers improve, find out whether anyone did anything. Improvement that came from growth stops when growth does, and it looks identical to improvement that came from management until it stops.
What the gap is worth
Glen Waverley runs labour at 39.5% of sales. The other three run 30.8% of their combined sales. Applied to Glen Waverley’s July sales of $92,172, that 8.7 point gap is $8,019 a month, or $96,228 over twelve.
Closing it moves group contribution from 5.4% to 7.3%.
Name the cost, because there is one and it is not small. $8,019 a month is about 215 labour hours, roughly a fifth of that site’s total. That is a real reduction in what the venue can staff, and it is the kind of decision that is easier to defend with eleven months of daypart data behind it than without.
It is also directional rather than exact. It assumes Glen Waverley can reach the labour percentage its three siblings run at, and a newer site in a different catchment may not get all the way there. The number is the size of the prize, not a forecast.
Run it on your own numbers
None of this needs software. It needs one month, split five ways, per venue.
- Net sales by venue. Gross less discounts, voids and comps.
- Wage cost by venue, and the actual hours behind it, not the rostered hours. The 44.6% weekend overrun above is only visible in actual.
- Rent by venue.
- Cost of sales by venue.
- Everything else by venue, allocated however you already allocate it. Be consistent rather than precise.
Then two ratios per site: wage cost over net sales, and net sales over actual labour hours. Rank your venues on both.
Two numbers tell you whether you have a problem. If your best and worst venues sit within a point or two of each other on labour percentage, your group figure is a fair summary and you can stop. If they are eight points apart, the group figure has been describing a venue you do not have.
The honest limit: this compares your venues to each other, which is the comparison that matters and also the only one you can make cleanly. A benchmark band tells you whether the group looks normal. It cannot tell you which site is paying for which.
Common Hour’s owner, Dan, could see 32.8% and inside the band. He could not see minus 4.0%, because nothing he was reading was capable of showing it to him.