Telemetry Is Not an IT Project

QUICK ANSWER. Coffee machine telemetry projects rarely fail on technology. They fail because they begin as data plumbing — connect the fleet, collect everything, build a dashboard — instead of beginning with a business question and a baseline. Connecting machines is the easy part. Deciding what the data must change is the project.

Every operator who has tried this recognises the pattern.

A connectivity vendor is chosen. Gateways are ordered. Somebody in IT owns the rollout. Eighteen months later the fleet is connected, a dashboard exists, and the service manager still runs Monday morning exactly the way they did before.

Nothing broke. Nothing changed either.

Why do coffee machine telemetry projects stall?

Because they are scoped as infrastructure, and infrastructure has no opinion about how you run your business.

A telemetry rollout scoped by IT produces a working data pipeline and a screen. Neither of those is a decision. The gap between “we can see the fleet” and “we act differently because of what we see” is not a technical gap — it is an operating-model gap, and no gateway closes it.

Connectivity is a prerequisite. It has never been the point.

The projects that work invert the order. They start from a decision somebody in the business is currently making badly, and work backwards to the minimum data required to make it well.

What should you measure before you connect anything?

A baseline. Not a comprehensive one — a usable one. If you cannot state today’s numbers, you will never be able to prove the project worked, and the initiative dies at the first budget review.

Four things worth writing down before a single gateway is ordered:

1. How many service visits you make per machine per year, and roughly what share are preventive versus reactive.
2. Your first-time fix rate — how often a technician resolves the issue on the first visit. Most operators estimate this generously; measuring it is uncomfortable and useful.
3. How long a machine typically stays down, from the customer’s first contact to working again — not from ticket creation.
4. Which accounts and machines you believe are your most profitable — written down, before the data can embarrass anyone.

That fourth one matters more than it looks. The point of a baseline is not accuracy; it is having a recorded prior belief to compare against. The value of knowing which machines actually make you money is largely in how often the answer surprises you.

Why is “collect everything” the wrong goal?

Because data volume and decision quality are not related, and pretending they are is how projects lose two years.

“Collect everything” is what a team specifies when nobody has decided what the data is for. It feels safe and defers every hard question. In practice it produces three problems at once: a longer integration, a higher running cost, and a dataset nobody can interpret because no one agreed in advance which number would trigger which action.

The better test for any proposed data point is blunt: what decision changes if this number moves? If nobody can answer, it is not a requirement — it is a nice-to-have wearing a requirement’s clothes.

This gets sharper in a mixed fleet, which is nearly every fleet. EVA, the European Vending & Coffee Service Association, sizes the European vending and office coffee service market at roughly 4.5 million installed machines across 24 markets, and almost no operator of scale runs a single manufacturer. Each brand exposes different data, in a different structure, under different terms — the problem described in Why Every Coffee Machine Brand Gives You Different Numbers. Normalising a small, deliberate set of fields across brands is achievable. Normalising everything is a research project.

Who should actually own it?

Not IT. IT should be a supplier to this project, not its owner.

The owner should be whoever is accountable for the outcome the project is supposed to change — service operations, most often, sometimes commercial. The reason is simple: IT can deliver a connected fleet on time and on budget while the business value stays at zero, and by every measure IT will have succeeded.

That does not mean the technical questions are trivial. Integration, data access rights and what the system is permitted to do with a machine are real questions, and they are the reason CoffeeBrain is being built read-only and pull-only — it reads, it never writes back, and each manufacturer’s software remains the system of control. That constraint is what makes a cross-brand project acceptable to the people who own the machines.

What does telemetry become when it works?

It stops being telemetry and becomes operational intelligence — and the distinction is the whole argument.

Telemetry reports machine status: this machine is offline, this one threw an error code, this one has served four thousand cups. Operational intelligence is that same signal joined to contracts, service history and margin, so status becomes consequence: this machine is offline, in your third-largest account, on a route with no technician nearby, and it has failed twice this quarter.

One is a fact. The other is a decision. The difference between them is not more sensors — it is the join. That is what a day looks like when it lands: we described it hour by hour in A Day in an Intelligent Coffee Operation.

The five-phase implementation roadmap — and the minimum data set each phase actually requires — is Section 08 of the 9-step operating model in our office coffee service report. The principle is here. The sequence is in the report.

Key takeaways

• Coffee machine telemetry projects fail as infrastructure projects, not as technology — connecting the fleet is the easy part.
• Start from a decision made badly today and work backwards to the minimum data needed to make it well.
• Write down a baseline first — visits per machine, first-time fix rate, real downtime duration, and your believed-most-profitable accounts.
• “Collect everything” is a symptom of not having decided what the data is for; the test is what decision changes if this number moves?
• The owner should be the person accountable for the outcome, not IT — and telemetry only becomes operational intelligence once it is joined to contracts, service history and margin.

Request the white paper — From Coffee Machines to Profit Machines: a step-by-step model for a more profitable office coffee service operation.

If your fleet were fully connected tomorrow morning — which decision would you actually make differently?