Notes
Your café does not need an app
A rep tries to sell me one twice a year. Here’s what an app actually costs to build and keep running, and what customers asked for instead.
Twice a year, someone tries to sell me an app for the café. Usually it’s a rep with a lanyard and a laptop; occasionally it’s a customer who works in software and means well. The pitch is always the same shape. Your regulars will order ahead. You’ll send a push notification when the Kenyan lands. You’ll own the relationship instead of renting it from Instagram. It sounds like the future, and it is a genuinely bad idea for almost every café in Melbourne, including this one.
I say that as someone who spent fourteen years building and maintaining software before I started making coffee. I was the person who got the call at 2am when one fell over, so I know what keeping one alive actually costs.
The cost before anyone opens it
An app is not a thing you buy. It’s a thing you keep. The build is the cheap part, and the build isn’t cheap.
A modest cross-platform app — a menu, order-ahead, a loyalty screen — from a competent Melbourne studio starts around $18,000 and climbs past $40,000 the moment you want anything specific. I’ve sat in on those quotes for advisory clients. The middle of the last three I saw was $25,000, plus GST. That’s before a single person has downloaded it.
Then it costs you every year, whether anyone uses it or not.
| What it is | What it costs |
|---|---|
| The build, cross-platform, both stores | $18,000–$40,000 once, plus GST |
| Apple Developer Program | US$99 a year, every year |
| Google Play developer account | US$25, one-off |
| Backend hosting and push notifications | roughly $40–$90 a month |
| Fixing what the next OS update broke | a day or two of a developer’s time a year, at $150+ an hour |
The two store fees look trivial, and they are, but they’re the tell. Apple charges US$99 a year and it renews forever; miss the renewal and your app quietly vanishes from the store. Google’s is a one-off US$25. The real money is the last row. Apple and Google ship a major operating system version every year, and every year something your app relied on changes. By then the person who built it has moved to other clients, or gone in-house somewhere, or simply stopped answering. Now you’re paying a new developer to understand someone else’s code before they can even start on the thing that’s already broken.
One of my advisory clients — a two-site operator, lovely people — paid a small agency $30,000 for a booking-and-loyalty app back in 2022. It worked for about fourteen months. Then an iOS update changed how the calendar behaved, the agency had folded, and the app started double-booking Saturday nights. They came to me to fix it. There was nothing to fix. There was only a decision to make, which was to stop, and they had already spent thirty grand to arrive at it.
The install nobody does
Here is the arithmetic the pitch skips. For someone to use your app, they have to go to a store, search for you, download it, open it, and usually make an account. Every one of those steps loses people. And you’re asking all of it of someone who wanders in for a flat white and a chat.
Nobody installs an app for a place they visit twice a month. People install apps for their bank, for Myki or the train, for the two or three food delivery services they already argue with. Your café is not making that list, and it shouldn’t want to — that list is expensive real estate to hold.
A web page, by contrast, installs itself. Someone taps a link, it opens, they read it, they leave. No store, no account, no 180 megabytes sitting on a phone that’s always full. If they want it on the home screen they can add it in two taps, and most won’t, and that is completely fine. A single good page does everything an app menu does, works on every phone including the cracked Android in the tradie’s ute, and costs a rounding error to run.
What people actually want
In eight years of feeding people I have never once been asked whether we have an app. I have been asked, a great many times, three things: what time do you close, are you open on the public holiday, and can I book a table for six. That is close to the entire brief.
A menu that loads in under a second on a bad connection. Hours that are correct, including the Tuesday you’re shutting early for a funeral. A phone number that is tappable and rings a phone somebody answers. An address that opens in maps. Get those four things right on a fast page and you have beaten most of the hospitality apps in this city, a good number of which take six seconds to load a menu that turns out to be a photo of a laminated sheet from 2021.
Spend the app money on getting those right, and on decent photographs of the food. That is advice that pays for itself. The app is advice that bills you forever.
The loyalty-card argument
Loyalty is where I tend to lose people, because loyalty is the one thing an app looks obviously good at. Buy nine, get the tenth. Points. A little badge that fills up.
You do not need to spend $25,000 to give someone a free coffee. A paper card and a stamp costs about eleven cents, and it lives in the customer’s wallet rather than on the fourth screen of a phone behind a login they’ve forgotten. We use a paper card at Rosetti. It works. People lose them, which is frankly part of the economics, and the ones who keep them feel a small honest loyalty to a small honest object.
If you want it digital, the tool already exists and you’re probably paying for it. Square has a loyalty add-on that sits on the terminal you already own, and the customer joins with the phone number they tap in at the till — no download, no store, no account. That is the whole promise of the app, minus the app.
The one time it’s worth it
There is a real exception, and honesty requires naming it. If you are a high-volume, order-ahead operation — a CBD lunch counter doing four hundred coffees before ten, where a queue that moves faster is directly more money — then order-ahead genuinely earns its keep. At that volume the maths flips.
Even then, do not build it. Use a platform that already has the customers and the engineers. Square’s own online ordering, or a shared-order platform like me&u, means people order through software that is already on their phone and already maintained by a company with a whole team, rather than by whoever you can still afford in eighteen months when it breaks. You rent the machine; you don’t buy the factory.
And if the rep leans on card fees, as they sometimes do — “an app lets you take payments without the terminal surcharge” — know that from 1 October 2026 the card networks are removing surcharging on eftpos, Mastercard and Visa anyway. You won’t be passing that surcharge on either way, so it’s no longer a reason to build anything.
The last rep left me a glossy brochure. It’s under the short leg of the two-top by the window, doing the single most useful thing an app has yet managed for this café.