Café Nova:
A Margin Collapse

The brief

Café Nova is a Portuguese coffee chain in Lisbon and Porto. Over 18 months it made two big bets: it grew from 33 to 45 locations, 12 of them new, and it put €800,000 into a mobile ordering app for delivery and pre-order. Revenue grew 15%, but operating margin fell from 25% to 7%, and the board wants an explanation. Four executives each have a theory.

The data on the table

Falling revenue per location, not runaway costs

EBIT margin is revenue minus four cost lines, each as a percent of revenue, and all four ratios went up. The easy read is that costs got out of control. Before I buy that, I want to know whether the ratios rose because costs grew, or because revenue per location fell and dragged them up mechanically. So I check the per-location numbers first.

Rent per location is basically flat, around €42K in both years, so the +2pp on rent isn't rising rent. It's the revenue per store shrinking underneath it. Labour per location rose about 7%, but the labour ratio jumped a full 7 points, from 24% to 31%. Had revenue per location held at last year's level, that same wage bill would read around 26%. So of the 7-point jump, only about 2 points is labour costing more; the other 5 is the revenue-per-location fall.

So only the app is clearly new cost. The rest is revenue per location collapsing, down 16%, against a cost base that barely moved per store. A back-of-the-envelope check: if the 45 stores had each done last year's €327K, the margin would sit near 22% even carrying all of today's costs, app included. That holds the cost lines flat, so it flatters the counterfactual a little, but the direction holds. The drop from there to 7% is the revenue-per-location shortfall.

The categories follow from that. On revenue: locations split into new versus established, revenue per location, ticket, and channel mix. On cost: the app's return specifically, and whether any line is high once you adjust for the revenue shortfall.

The core is shrinking, and it's the ticket doing it

Revenue is locations times revenue per location, and revenue per location is daily customers times ticket times operating days. I'd also split it by channel, in-store versus delivery. Walking through where it breaks:

So the same-store decline and the ticket aren't two problems, they're one. Back out the maths: an 8% lower ticket on roughly flat footfall accounts for almost all of the 9% same-store fall. People aren't staying away, they're spending less per visit. That shifts the question from "why did customers leave" to "why is each visit worth less", and the case doesn't say: it could be discounting, a cheaper delivery mix, or trading down. The distinction matters, because the fix differs: trading down on worse service is something fixing the operation can reverse, while a shift to low-ticket delivery isn't. I'll assume trading down, since it lines up with the NPS fall, but a delivery-heavy mix is what would change the call, and the delivery versus pre-order split would settle which it is. New-store dilution is real, but I discount it, since some is expected. On the cost side the app at +5pp is the only new cost; the rest is the revenue-per-location fall feeding through fixed costs.

Neither waiting nor cutting: fix the operation

The CEO is half right. New builds do ramp over 12 to 18 months, so some of the €217K shortfall is normal, and honestly I can't tell from this data whether €217K is bad for a one-year-old store, because there's no ramp curve. What I can tell, on that same new-store assumption, is that the established stores look down about 9%, and time doesn't fix that. So "needs more time" covers the new stores but not the core estate, which is the bigger problem.

The CFO is right that it's urgent, with EBIT down to €0.9M, but the fix is aimed at the wrong target. The cost base per store is roughly normal; the problem is revenue per location. Cutting costs in a café usually means cutting staff hours, which makes service worse when turnover's already 40% and NPS is already down 19 points. That deepens the very problem it's meant to solve.

Turnover traces back to the management ratio

I'd cut it four ways:

• The job itself (workload): span of control collapsed to about 1:11, against a healthier 1:6 or so, and understaffing means everyone carries more.
• Pay and progression: I can't assess this from the data. Labour per location rose only about 7%, but that's a store-level number, not per-person pay, and it's tangled up with the revenue fall. I'd want wage benchmarks and headcount first.
• Onboarding and supervision: 12 new stores in 18 months with too few managers means thin training and weak day-to-day supervision.
• Outside the company: the case puts the industry norm around 22% (real hospitality runs higher), so some of this is just the market.

I'll be honest that the first and third overlap, since both come back to too few managers, so this isn't perfectly MECE. The practical point is they share one root. Most likely it's the management ratio: turnover nearly doubled to 40%, 18 points above the 22% norm, and that gap is too big to pin on the market or pay alone. The COO names it directly.

A symptom that's turning into a cause

It's a symptom, and it's starting to feed back as a cause. The chain: the expansion outran the ability to staff and manage it, managers got stretched to about 1:11, staff ended up undertrained and left at 40%, service turned inconsistent, and NPS fell from 61 to 42 with complaints rising. So far that's shown up in a lower ticket more than in lost footfall, but a 19-point NPS drop rarely stays contained to spend per visit.

On whether NPS leads or lags, it's both, depending on what you point it at. It's a lagging read on service quality, since it reflects what already went wrong. But it's a leading indicator for revenue, because today's NPS predicts tomorrow's repeat business.

Keep the app, stop running it as a delivery business

My call is to keep it, but stop running it as a delivery business, and I'll give a clear trigger for reversing that. I lead with the decision rather than a number because, on the data given, the contribution can't be pinned to a single figure and it lands near break-even either way. The number isn't what should decide this.

First, the €800K build is sunk. Ignore it. The only question is what the app contributes from here. Three things would settle it, and none are in the case:

Put those together and the contribution lands around break-even, with a wide band: modestly negative if it's all 28%-fee delivery and the full €600K is forward cost, modestly positive if a good share is owned pre-order and part of the €600K is sunk. That's too soft to shut the app down on, in either direction.

The real issue is the 28% fee against the 12% planned. At 28%, third-party delivery is structurally low margin, and that won't change on its own. What's worth keeping is the owned channel: pre-order and loyalty pay no fee and build a direct line to the customer, which is what protects repeat visits and ticket. The 25,000-user target was over-ambitious; about 10% of revenue in year one is a slow start, not a failure.

Stabilise the operation first, then grow

What's missing would change the call

A real interview pushes on two things this case under-tests: handling missing data, and resisting a false-precision number on the app. The data I don't have but would most want:

• The delivery versus pre-order split inside the app's €1.2M. The single biggest gap. It decides how much of the 28% fee actually applies, and therefore whether the app is viable.
• Product cost on delivery, and how much delivery is incremental versus cannibalising in-store. Same theme: without it I can only guess at the app's real contribution.
• A same-store sales series, and the ticket drop broken into price, promotions, and mix. The series separates "new stores just need time" from "the core estate is eroding"; the ticket breakdown tells me why each visit is worth 8% less, which is the actual bleed. Right now the 9% rests on a single assumed new-store number.

I'd also want headcount and wage benchmarks for the turnover question, to settle whether pay is part of the story.

Around break-even, with the assumptions named

Starting from €1.2M of revenue, the costs I'd charge against it:

• Product cost at about 38% (the company's own rate): roughly €456K.
• Commission at 28%, but only on the delivery slice. If all €1.2M is delivery, €336K. If half is owned pre-order, closer to €168K.
• App operating cost, up to €600K, but some of that may be depreciation of the sunk build, so the true forward cash cost may be lower.

So contribution runs from clearly negative (all 28%-fee delivery, full €600K forward) to roughly break-even or slightly positive (a good share owned pre-order, part of the €600K sunk).

What I'd challenge about this case

A real handout wouldn't be this tidy, and part of the job is saying so. Five things I'd flag before trusting the conclusions.