There is a story worth telling at every meeting on developing a coffee business. In an audit of one large Russian chain, 500 coffee machines were checked. Not one fully matched the chain's own recipe card. Not even two machines had the same settings. This was told by Alexander Kuzmin, head of the RusHolts group, in an interview with RBC Saint Petersburg.
Five hundred machines means five hundred different drinks under one logo. A casual customer will not notice the variation. A regular will notice it by the third visit and stop thinking of your coffee as predictable.
Why settings drift
Nobody spoils coffee on purpose. The divergence builds up by itself, from small things.
What it costs in money
Direct losses from taste variation do not show up in reporting, so the problem goes untouched for years. It has to be calculated indirectly.
Purchase frequency. Coffee is a habit product. A person comes back for a specific taste. If the drink differs between neighbouring points of the chain, the habit forms not for the brand but for one good point, and the rest of the fleet lives on random traffic.
Ingredient consumption. A machine with an overstated bean dose uses a gram or two more on every serving. At 60 cups a day that is an extra 2-4 kg of beans a month from one point. Across a hundred points the sum is large enough to justify a regulation.
Mismatch between cups and receipts. A separate story from taste, but a neighbouring one. In some chains the difference between drinks prepared and drinks paid for reaches 30%. Eliminating that mismatch alone pays for any control system.
How to hold the standard
One recipe card per model, not per chain
A universal card for a fleet of four different models cannot work by definition. The right structure is target drink parameters (volume, strength, temperature, milk ratio) and a separate setup for each model that delivers those parameters.
Centralised settings upload
Settings should arrive at the machine from a system, not be keyed in by hand on site. Any manual tweak at a point is a future divergence. If the equipment allows the settings menu to be password-locked, it should be locked.
Telemetry that serves the business, not every parameter
The processor of a modern coffee machine records 400-600 parameters. Running a business takes about twenty: bean level, milk level, a missed cleaning, a machine error, serving counter, mismatch with the till. The rest is for the service team's diagnostics and should not end up in operational reports.
Water as a separate process
A filter is replaced by volume, not by calendar. The cartridge's life is known and the machine has a water counter, so replacement can be planned automatically. It is the cheapest upgrade of all those listed and the most noticeable in taste.
Who gets the notifications
At a fuel station it is simpler: there is a station manager's workstation, and the message goes there. In retail it is harder: phones on the sales floor are often banned and staff change. A working scheme here goes through regional managers, who receive the notifications and pass the task to a specific store.
Raw material quality control
Large players have already gone beyond equipment settings. Plekhanov Russian University of Economics and Perekrestok created a coffee quality control laboratory for the chain and the Select format: every batch of green beans is tested for botanical species, arabica-to-robusta ratio, defects and moisture, and after roasting is analysed for colour profile and aroma with an "electronic nose". A digital quality passport is generated for every batch. The Select format now brings together more than 440 points in supermarkets, and in the first half of 2026 they prepared more than 8 million drinks.
Not everyone needs this level of control. But the logic applies to any chain: the more points, the fewer decisions can be left to the discretion of the location.
Where to start if the fleet has already diverged
The work is dull. But it is the kind that requires no new investment in equipment and delivers an effect across the whole fleet at once.
Conclusion
Inconsistent settings do not show up in reporting, so they go untouched for years. Meanwhile they prevent a habit for your brand from forming and quietly eat beans on every serving.
Bringing the fleet to a standard is dull work and needs no new investment in equipment. But it delivers its effect immediately at all points.
Frequently asked questions
How many coffee machine parameters need to be tracked
Two dozen of the 400-600 the processor records are enough to run the business. The rest belong to engineering diagnostics.
How can you tell that settings have diverged
The quickest way is to export the actual profiles from the fleet and compare them with each other. A second sign: bean consumption per serving differs from point to point with the same menu.
Can quality be controlled without telemetry
On a fleet of up to ten machines, yes, through regulations and site visits. Beyond that manual control stops keeping up, and deviations are found from complaints rather than from data.
What to do about different water in different districts
Choose filtration to match the actual hardness and replace cartridges by volume, not by calendar. An identical filter across the whole city gives different results.
More on fleet control: coffee machine monitoring and telemetry, solutions for retail chains, coffee at fuel stations.
How this works on the under-counter station — on the home page; short answers on installation, water and service are in the Short answers section.
What a station would earn for you
We’ll run the payback on your numbers: traffic, cup price, rent.