A parallel run is the period where the new system and the old way both do the same work, on the same jobs, at the same time, and someone compares the two. It ends when your team says it ends, not when we say it is finished. We will not skip it, and we will not quote a build that leaves it out.
What is a parallel run?
For as long as it takes your team to trust it, every job goes through twice. Your team writes the quote or the method statement the way they always have. The system drafts its own version of the same thing. Someone puts them side by side.
Most days they match. The interesting days are the ones where they do not, because that is the system telling you something you did not know about your own process.
Nothing the system produces goes to a client during that time unless a person has read it and decided it is right. After go-live, a person still signs off anything where a mistake costs money or safety.
Why version one is always wrong somewhere
Not broken. Wrong in one specific place, and usually a place nobody could have predicted.
It happens because the rules you described are the rules you believe you follow. The rules your business actually runs on include exceptions that live in one person’s judgement and have never been written down. The client who gets a different format. The material you never quote on without ringing the supplier first. The council that wants the sections in a different order.
Nobody is hiding these. They are invisible from the inside. You only find them by producing real output on real jobs and having the person who does the work look at it.
The other reason is data. Somebody’s spreadsheet turns out to be holding something together. Two fields in the job system mean different things depending on who filled them in. The parallel run finds those while the old way is still running, which is the only safe time to find them.
What it costs you
Honestly: a bit of extra work for a few weeks, from the person whose time you are trying to save. That is the awkward part and I would rather say it before you sign than after.
It is worth it because the alternative is switching over on a Monday and discovering the exception on a live job in front of a client. The cost of that is not measured in hours.
How it ends
Not on a date I choose. It ends when the person doing the comparison stops finding differences worth acting on, and says so.
The differences thin out as each one is fixed. The last few are edge cases you decide to leave to a person on purpose. Then somebody points out they have stopped reaching for the old version. That is the signal.
After that comes go-live and training, which is mostly showing people exactly where the sign-off lives, and then the monthly running.
Why we won’t skip it
Because of what happens without it. Skip it, and version one goes live before the person who does the work has checked it against the old way. What it gets wrong turns up on a live job, and by then the trust is gone. After that the team goes back to the old way quietly and you have paid for something nobody uses.
The parallel run is not caution for its own sake. It is the cheapest insurance in the whole process, and it costs a few weeks of one person’s attention.
If someone offers to build the same thing without one, ask them what happens when version one is wrong.
What this means for the quote
The quote comes from the paid audit. It is a fixed price for one route, with three options that cut scope rather than corners. The parallel run is in all three, not an extra. Signature to live is from 4 weeks, depending on the size of the job, and that time includes it.
The nine stages are on the About page timeline if you want to see where this stage sits. On the 10-minute call, you tell me about the document that eats the most time. I say whether it fits and give you a rough estimate, a range for the first route. If it’s not worth doing, I’ll say so.