Once a business starts looking, the list of things that could be improved is long. Choosing badly produces three half-finished projects and a general sense that this sort of thing does not work here.
Build the list properly first
Spend a fortnight noticing rather than planning. Specifically, look for:
Information typed more than once. Every instance is a candidate.
Work that waits. A job that sits for three days because it needs an approval, an invoice that waits for a monthly run.
Questions that get asked repeatedly. Internally or by customers. Each is either a missing piece of information or a missing self-service route.
Things people do from memory. Steps only one person knows.
Reconciliation. Any time somebody compares two systems, one of them is not trusted.
Spreadsheets doing operational work. Covered in when a spreadsheet stops being enough.
Talk to the people doing the work rather than the people managing it. They know precisely where the friction is and are rarely asked.
Score each candidate
Two axes, roughly rather than precisely.
What it costs now. Hours per week, errors per month, revenue lost, risk carried. Put a number on it even if the number is approximate — an approximate number is enormously more useful than an adjective.
What it would take to change. Cost, elapsed time, how many people's work changes, how much data has to move, and how many other systems are involved.
The high-cost, low-effort items are your starting point. There are usually two or three, and they are frequently unglamorous — an invoice that goes out automatically, a form that creates a record, a report that sends itself.
Then respect the dependencies
Scoring alone produces the wrong order, because some things must come before others.
Data before reporting. Reporting built on data nobody trusts produces arguments rather than decisions.
Process before automation. Automating an unagreed process makes the disagreement faster — see when not to automate.
Systems of record before customer-facing views. A portal displays information from somewhere; if that somewhere is unreliable, you have shown your customers your data problem.
The boring foundations first. Connectivity, backups, account hygiene. Not projects anybody is excited about, and everything else depends on them.
The biggest problem is usually the wrong place to start. Start where you will succeed, and use the credibility to attempt the difficult thing next.
Pick a first project with three properties
Quick. Delivered inside a few weeks. Long first projects lose momentum and support.
Visible. People can see the difference. Something that improves a back-office metric nobody watches builds no confidence.
Beneficial to the people affected. Not just to management. The first project sets whether staff regard this as something done for them or to them, and that impression is very hard to change later.
An invoice that goes out the day the work is done, a form that stops somebody retyping, a report that arrives without being prepared — these are unremarkable and they are how a programme starts well.
One at a time
Small businesses have limited management attention, and it is the constraint rather than money.
Three projects running simultaneously means three projects at sixty per cent, none delivering. One project finished delivers, teaches you something, and makes the next one easier.
Finish before starting. It sounds obvious and it is the discipline most often abandoned.
Measure and say so
Before each project, record what it currently costs. Afterwards, measure again and tell people.
Two reasons. It tells you whether the effort was worthwhile, which affects what you do next. And it builds the credibility that gets the next project supported.
Report honestly, including where the benefit was smaller than predicted. That is what makes the next prediction believed — see writing a technology business case.
What to leave alone
The list matters as much as the priorities.
Things that work. An old system that does its job adequately does not need replacing because it is old.
Low-volume processes. Something done four times a year does not justify a system, whatever the irritation.
Anything about to change. If the process is under review, digitising it now means doing it twice.
Things you cannot articulate. If nobody can describe how the process should work, no system will fix it. Agree the process first.
Revisit annually
The list changes as the business does, and items that were low priority become urgent when volume grows or a person leaves.
An hour once a year, rebuilding the list and rescoring, keeps the sequence honest — and frequently reveals that something near the top last year has stopped mattering.
Our consultancy team helps UK businesses build and sequence this list, including the items that should be left alone. Start a conversation.
