How We Categorise Your Business Before We Build

Why the same request means five different systems

“We need a system to manage our orders” is a sentence we have heard from a wholesaler doing 4,000 orders a week, a design studio doing eleven projects a quarter and a clinic booking appointments. The words are identical and the correct software has almost nothing in common between them.

So before we discuss features we place the business on four axes. The answers determine the architecture, the sensible budget and, most importantly, what to build first. Getting this wrong is how businesses end up with an enterprise system for a workshop problem, or a spreadsheet replacement for something that needed a platform.

Axis 1: transaction volume

How many times a day does the core event happen — an order, a booking, a claim, a shipment? This decides almost everything about the architecture, and the thresholds are less about technology than about human attention.

The mistake we see most often is a business at the first level buying software designed for the third. It works, but the staff spend their days feeding a machine built to run without them.

Axis 2: process variability

Does every instance follow the same path, or is each one a little different? A parcel courier is low-variability: the steps are the same every time. A construction consultancy is high-variability: every project has a bespoke shape.

Low variability suits rigid, automated workflow. High variability needs software that captures and reports without dictating, because a rigid system in a variable business gets worked around within a month, and then you have both a system and a shadow spreadsheet.

The fastest way to make a business hate its new software is to encode a process that only holds 70% of the time and give people no way to record the other 30%.

Axis 3: data maturity

Where does your operational truth live today?

1. In people's heads. Nothing is written down; the estimator knows the pricing. Build a system of record first — nothing else is possible until the knowledge exists outside a person.
2. In documents and spreadsheets. Written down but not queryable. This is the most common starting point and the one where a first project pays back fastest.
3. In systems that do not talk. Good data, several sources of truth. The work is integration and reconciliation, not new features.
4. In one connected system. Now analytics, forecasting and AI become realistic, because there is finally something coherent to train on or report from.

Businesses routinely try to buy stage four while sitting at stage two. The predictions are only as good as the data underneath them, and a forecasting dashboard fed by inconsistent spreadsheets produces confident nonsense.

Axis 4: who the software actually serves

Internal staff, your customers, or both? This changes cost more than any other single factor, because customer-facing software carries obligations internal tools do not: accessibility, design polish, uptime, support load, security review, and the fact that a confused user cannot be trained.

As a rule of thumb, the same functionality costs roughly twice as much when customers touch it. That is not padding; it is the states, the edge cases and the failure handling that internal users tolerate and customers do not.

What the four answers produce

Together they generate a recommendation before we have discussed a single screen. A low-volume, high-variability, spreadsheet-stage, internal business gets a lightweight system of record and no automation in phase one. A high-volume, low-variability, multi-system, customer-facing business gets an integration layer and an event-driven core.

• The architecture — monolith, service split, or workflow engine
• The realistic budget band, which is mostly a function of axes 1 and 4
• What phase one contains, and what is deliberately deferred
• Whether AI is useful yet, which axis 3 usually settles

We write this up in a page and send it before the proposal. Clients tell us it is the most useful document they get from us, and it costs nothing.

Frequently asked questions