How does a deployment start?
With a question, not a move. Your teams and ours agree on the first thing you need to know, and on the sources that hold the answer. Each of those sources is connected as it is, and nothing in it is replaced. The rest of your systems carry on as before.
Many data projects open with the same promise: move everything into one new place, clean it on the way, and the answers will follow. So the first answer waits until the move is over. Meanwhile the old systems keep running and keep changing, and the new place has to keep up with them.
We do not think that order makes sense. A ministry or a bank does not buy a data platform to own a fresh copy of its records. It buys one to answer questions it cannot answer today.
Two words are worth telling apart. A migration moves data out of the systems that hold it and into a new one, which then has to replace them. An integration connects those systems and reads them together, where they already are. Both have their uses. Only the second can give a first answer before every system has been dealt with.
Why moving everything first is a gamble
In 2022, Bent Flyvbjerg and five co-authors studied 5,392 IT projects and how far they went over budget. Many went over by a little. A smaller number went over by an extreme amount, far more often than decision makers tend to expect. The authors offer an explanation, and test it by simulation: “a problem in a single technological component can lead to chain reactions in which other interdependent components are affected, causing substantial overruns.” In plain words, when everything depends on everything else, one small fault can spread to the whole project.
A project that moves every system at once is exactly that kind of project. We would rather not start there.
One question, a few sources
So we begin the other way round. With your teams, we choose one question worth answering: a fraud network to trace, a fault that keeps coming back, two registers that should agree and do not. Then we connect only the sources that hold its answer. Each one is connected as it is, whatever its format, with a key of its own. Datasight reads it and leaves it in place, so it goes on serving the people who use it every day.
Start from one question
InteractiveWhich customers belong to the same fraud network?
- Customer filesUntouched
- Accounts and transfersUntouched
- Devices and loginsUntouched
- Company registerUntouched
- Card paymentsUntouched
- ComplaintsUntouched
- Branch networkUntouched
- TreasuryUntouched
- Human resourcesUntouched
- MarketingUntouched
0 of 10 systems read0 moved
Try itAsk the question and watch which systems it needs.
Ten systems, one question. Nothing is connected yet.
Green: read where it is, as it stands. Grey: not touched.
Illustrative. The bank, its systems and the question are examples, not a client’s.
The first results are checked by those who know
None of the first results is taken on trust. Your teams go through them with ours, and compare them with what they already know: a case they have closed, a machine with a known history, a person they know appears twice. Only after that review are the rules adjusted to the way your organisation really works.
This step matters more than any demonstration. A platform that finds what your best people already knew has earned the right to show them what they did not.
Growing from there
The next question may need a new source or bring in a new team. Each one builds on the same base as the first, and nothing has to be moved to make room for it. The platform grows at the pace of the questions you bring, not at the pace of a migration plan.
We would rather give you one sound answer early than a complete copy of your data late.
How to check it
- Ask each supplier which question its first phase will answer, and what you will have in hand when it ends.
- Ask which of your systems will be replaced, moved or switched off along the way. Compare the answers you get.
- Ask to review the first results yourselves, against cases your teams already know, before any rule is set.
Sources
- Bent Flyvbjerg, Alexander Budzier, Jong Seok Lee, Mark Keil, Daniel Lunn and Dirk W. Bester, “The Empirical Reality of IT Project Cost Overruns: Discovering A Power-Law Distribution”, Journal of Management Information Systems, 39(3), 2022. arXiv