We fix one thing at a time and measure it before moving on.
Below is how an engagement runs, and when you'd be better off hiring someone else.
Conditions
- 2.1
Go to the root first.
A lot of data projects fail because they never get to the real problem. Usually it's missing keys, figures that don't reconcile, or quality issues nobody put in scope.
- 2.2
Fix one piece at a time.
We move one piece, check it works against the baseline, and only then start on the next.
- 2.3
Agree what "trusted" means.
Before building anything, we settle with you what counts as trusted data. If we think the definition is wrong, we'll say so.
- 2.4
It's fine not to know what you want yet.
That's normal at the start. Working it out together is part of the job.
- 2.5
The person who scopes the work builds it.
Alex does the diagnosis and the build, so nothing gets lost between a sales team and a delivery team.
How we prove it
No figure goes in a proposal, post or case study unless it would survive a reference call.
Every engagement starts with a baseline we agree with you and ends with the same measurement. We only promise numbers that can actually move while we're working together, which is why loss ratio is a follow-up a year later. Anything we haven't measured yet is labelled illustrative, including the examples on this site.
When not to hire us
Better you hear it now.
- 4.1
You need a large team in place next month. Karasu is one person, on purpose.
- 4.2
You want a big-bang migration delivered in one go. We release in small pieces, and each one is checked before the next.
- 4.3
You want a loss-ratio result inside the engagement. It takes 12 to 24 months to show a change. We can measure it with you a year later.
- 4.4
The problem is actuarial judgement, not data. Your actuaries are the right people for that. We can make sure the data they work from is sound.
- 4.5
You want your definition of trusted accepted without question. We'll want to talk it through with you first.
- 4.6
You want AI before your data can be trusted. We'd start with the data. If that's off the table, we're the wrong people to call.
Still the right fit?
Everything starts with the data map: two weeks, for a fixed fee.