Measure before and after: get a baseline before the six month project starts

On-Off Group

October 2026

Pick one customer journey, measure it before you change anything, then measure the same thing again after the pilot. Turnaround time and completion rate, counted the same way twice, give you an answer when your boss asks if it worked.

If you have to show your boss something real from the customer journey work this month, start with the before number.

Measure before and after. One journey, one number, counted the same way twice. Once before you change anything, and again after the pilot has been live long enough to have real cases in it.

Below: how to pick the number, how to get it when the data is a mess, and what to say if the result is small.

Your boss asks if it worked, and you have no before number

The pilot ships. Branch staff are using the new onboarding flow. Someone upstairs asks whether it worked.

The honest answer is a shrug. Something may well have improved, but nobody wrote down what the old flow was like. It was never timed. Drop-off was never counted. All you have is what people remember, and people remember kindly.

This is the gap we find most often in bank and insurer teams. There is no baseline. So the result turns into an argument, and the person with the better slides usually wins it.

The cost is not just this project. It is the next ask. When you go back for budget with a story instead of a number, you get the same hearing as any vendor pitching a deck. Second time round, people stop making time for the work.

Pick one number, on one journey

Narrow it down until it feels almost too small. One journey. Account opening, or claims submission, or a card replacement.

Then pick turnaround time, measured end to end. Not the bit your team owns. From the moment the customer starts to the moment the thing is actually done for them, including the wait while a document sits in someone's queue.

If you can count how many people start and how many finish, take completion rate too. That is enough. Two numbers on one journey are worth more than a page of numbers across a whole program.

Write the definition down in one sentence before you collect anything, and be specific about the edge cases: does a rejected application count, does a resubmission start the clock again. If the definition drifts between the before and the after, the comparison is worthless and somebody in the room will notice.

We have written about picking a single journey rather than a platform-wide effort in Instead of a platform project: ship one customer journey, measure it, then widen.

Getting the before number when the data is a mess

Most teams stop here because the reporting cannot answer the question. Count by hand instead.

Pull twenty recent cases and time them from the records you do have: timestamps, email trails, the log a branch keeps in a spreadsheet. Twenty is enough to see whether you are dealing with two days or nine. Note the spread as well as the average, because the slow cases are usually where the complaints come from.

Then watch it happen. Five customers using the current journey, on video, and two branch staff doing their side of it. You will see the pauses the data never records: the re-keying, the call to verify something, the form people abandon and start again on a different device.

That hand-counted number is defensible in a way a dashboard often is not, because you can explain exactly how it was made. Keep the raw sheet. When the after number looks good, someone will ask how the before number was calculated.

What a real before and after looks like

A large insurer had customer-facing web journeys that took six months or more to get from idea to live. That six months was their own number, measured by their team, not by us.

We ran a series of five-day design sprints. Each journey was designed and tested with customers in two weeks and live in four. Twenty-six weeks down to four, measured by the client. Completion rates on the new journeys rose 80 per cent and the old drop-off points disappeared.

The size of the improvement matters less than this: they had a number before the sprints and a number after, counted the same way. Somebody knew what the six months was made of before any sprint started, so four weeks could be checked.

Same with the drop-off points. They could say those were gone because someone had counted them before the change.

Small honest numbers, and when to say so

Sometimes the after number barely moves. Turnaround time goes from nine days to eight and a half.

Say it plainly. A half-day improvement tells you the bottleneck is somewhere you did not touch, which is useful information you can act on next month. Reporting a small number honestly also buys you something: the next time you bring a big one, people believe it.

The temptation is to switch metrics. The turnaround time did not shift much, so the slides quietly become about satisfaction scores and workshop attendance instead. People spot that switch, and it does you more damage than the flat number would have.

If you are adding an AI step to a journey that is already broken, the before number will show you that. AI on a broken journey speeds up the failure. Fix the steps in the journey first, then automate. Related reading on getting something working before the paperwork: Build a rough working version of an AI feature instead of writing the business case first, and on moving from a prototype to something live, How to take an AI prototype to production without a six month project.

What to do this week

Choose one journey and write its turnaround time definition in a single sentence, including what counts as done.
Pull twenty recent cases and time them by hand. Record the average and the slowest case.
Record five customers and two branch staff going through the current journey, so you have the before version on video.
Send the before numbers to your boss now, before the pilot starts, so the comparison is already agreed.
Put a date in the diary to re-count the same twenty-case sample the same way after the pilot has been live a few weeks.

Questions people ask

What should we measure first?
Turnaround time on one journey, end to end, from the customer's first action to the moment the thing is done. Add completion rate if you can count starts and finishes.

We have no clean data. Can we still get a baseline?
Yes. Time twenty recent cases by hand from whatever records exist, or watch five customers and staff do it. A hand-counted number you can repeat beats a dashboard you cannot explain.

How long should the before and after measurement take?
The before number usually takes a few days of counting. The after number uses the same method on the same journey once the pilot has been live long enough to have real cases in it.

What if the improvement turns out to be small?
Say so. A small honest number tells you where to look next and it protects your credibility for the bigger ask.

On-Off Group trains teams, tests products with real customers, finds where a transformation has stalled and builds what gets it moving, for banks, insurers and enterprises in the Philippines, since 2015. How we help with baselines and measurement.