4 min read
Seven mistakes when computerizing a business.
The seven mistakes that come up most, why each one looks like a sensible decision in the moment, and what to do instead of each one.
None of these mistakes looks like a mistake when it's being made. Every one of them looks like the sensible call at the time, which is exactly why they keep repeating.
And they repeat at a scale that's surprising. The Standish Group has been measuring software project outcomes since 1994 in its CHAOS report (opens in a new tab), which is the field's reference. In its founding measurement, only 16.2% of projects were delivered on time and on budget. 31.1% were cancelled before completion. And the ones that shipped but ran into trouble averaged 189% over the original cost: nearly triple what was budgeted.
Worth noting that report is old and later measurements dispute its methodology. But the order of magnitude held across three decades of studies, and so did the underlying conclusion: what sinks these projects is almost never technical. It's how they were decided and how they were implemented.
Which is exactly what this list is about.
1. Starting with everything at once
The decision is made to computerize, and sales, purchasing, inventory, and invoicing all get started together, to get it "done in one go."
The problem is nothing works until everything works, and any delay in one part stalls the others. On top of that, the team has to learn four new things in the same week, and ends up going back to spreadsheets.
Instead: pick the workflow that hurts the most and take it end to end. Once that one is running and people are using it, move on to the next.
2. Copying today's mess into the new system
The vendor is asked to make the system do exactly what happens today, exceptions included.
That means paying to lock into code a tangle that could have been fixed for free. And what used to be worked around by hand now has to be programmed every time.
Instead: before ordering anything, review which exceptions are genuinely necessary and which are just habit. Usually half of them turn out to be unnecessary.
3. Not putting anyone in charge
The project belongs to "everyone," which in practice means it belongs to no one. Decisions take weeks and come back changed.
Instead: one person with the authority to decide without checking with anyone else. They don't have to be the owner or know anything about technology. They have to exist and be available.
4. Leaving training for the end
The system is built for months and shown to the team the day it goes live.
That day, every problem shows up at once: things nobody thought through, steps that don't actually work in practice, people who don't want to use it. And by then there's no budget left to fix any of it.
Instead: have two or three people from the team try it early, even incomplete. They'll find in an afternoon things nobody caught in three months.
5. Migrating every piece of old data
It seems obvious to carry ten years of history into the new system. It's one of the most expensive things you can do, because that data is almost never clean and every inconsistency has to be decided on individually.
Instead: bring over only the data you need to operate today: active customers, current products, balances. History can be looked up separately, and almost never is.
6. Measuring success by whether the system runs
The system works, nobody's complaining, and yet the company operates exactly like before.
Instead: before starting, write down what you want to change and how you'll know it changed. How many hours get saved, how many errors go down, how long a month-end close takes. If it can't be measured, you won't be able to tell whether it was worth it.
7. Not planning for what comes after
The project is thought through up to delivery and no further. Afterward come law changes, services that get updated, bugs, new people who need training.
Instead: settle from the start who handles maintenance, what it costs, and how it's requested. A system with no maintenance planned degrades on its own, and in two years you're back where you started.
The thread that ties them together
All seven are the same thing seen from seven angles: wanting to solve everything at once, instead of solving one thing and learning from it.
Computerizing a business well is boring. One workflow at a time, people using it before it's perfect, and a clear definition of what needs to improve. It isn't fast, but it's the only approach that doesn't end with an expensive system nobody uses.