When a company decides to sort out its technology, the diagnosis is usually already in: the systems don't talk to each other, the figures don't add up, there's too much manual work, and management information can't be trusted.
What stalls things isn't knowing what's wrong. It's knowing where to start — and the order that seems most logical is almost always the one that backfires, because it starts with what's most visible instead of what holds up everything else.
Three projects you'll want to do first
And that, done first, cost money without fixing anything.
Automating. It's the most tempting one because the gain is immediate and sounds good in a meeting. The trouble is, automating a messy process just makes the error happen faster and more often — and now with nobody watching, which was the only thing keeping it in check.
Buying management dashboards. They go in within a week and look good. If the data underneath is inconsistent, the charts will show figures nobody recognises, and within two months nobody opens them any more. The budget's spent and the idea's credibility is burnt.
Building bespoke software. If the workflows haven't settled yet, what you build is a snapshot of the current mess, in code. And changing code is expensive.
What always comes first: a list
The first job isn't technical, and it can't be bought from anyone. It's writing down the list of systems the company uses — all of them, including the ones only one person uses and the ones nobody owns up to — and, next to each, what it's in charge of.
“In charge of” means: where that information is created and corrected. Does the customer record start in the sales system or the invoicing one? Does the price come from the catalogue or the quote? Is the order status set by logistics or by customer service?
This list usually fits on one sheet and takes two days to put together. It's also the most uncomfortable conversation in the whole process, because it forces decisions everyone preferred to leave open for years. That's exactly why it's useful.
Next: getting the information to flow
With ownership decided, the next step is making the systems that need a piece of data fetch it from where it originates, instead of keeping their own copy.
This is where connecting systems together comes in. It's unglamorous work, and it's what makes the biggest difference: when the numbers start adding up without anyone reconciling them by hand, it changes teams' relationship with data — as we wrote about in why nobody trusts the numbers.
Do it one piece of data at a time, in the order in which they cause arguments. Don't take on a total-integration project: those take a year and get cancelled in month seven.
Only then: automate, measure and build
With roles defined and information flowing, the three projects from the start of the list start to make sense — and become cheaper, because half the hard work is already done.
Automation now rests on workflows that already behave consistently. The dashboards show figures the team recognises. And whatever gets built bespoke is built on a clear idea of what the company does, rather than a tracing of the confusion.
It doesn't have to be perfect to move on
One caveat, because this order can also be taken too far. A level doesn't need to be flawless before you move on to the next one — it needs to be stable enough to carry what's built on top of it.
In practice, this is done area by area: sort out sales and automate sales, while logistics is still waiting its turn. It's faster, each phase pays for itself, and nobody has to wait two years to see results.
Want help making the list?
Tell us what systems you use and where the numbers don't add up. We'll go through it with you and propose the order of work.