Nobody plans an eighteen-month data platform. It happens one bespoke integration at a time - and the escape is recognising that most of it has been built before.
Where the eighteen months go
A ground-up commodity risk data platform is not primarily a technology project; it is a thousand small negotiations with reality - schema mapping for nine source systems, quality rules per feed, lineage design, the forty-three definitions of "volume." Firms budget for the visible engineering and discover the hidden seventy percent: requirements archaeology, reconciliation logic, and rework when the first consumer of each table turns out not to be the last. Twelve to eighteen months to a first trustworthy number is the norm, not the exception.
What an accelerator actually is
Not a product. A pre-built, field-tested module for a data domain that recurs across engagements: exchange and broker feed ingestion with schema enforcement; position and trade mastering; cargo and logistics tracking; curve and reference-data management; simulation job orchestration; the Control Tower decision layer. Each module arrives with its data contracts, quality rules, lineage patterns and validation suites already written - because they were written on the last engagement, and the one before.
Why the compression is real, not marketing
The arithmetic is unforgiving: on a typical engagement, 60-70% of the build is identical to the last one, differing only in source-specific quirks. Accelerators convert that share from project work into configuration. Discovery maps the client's estate onto the module library; engineering effort concentrates where it should - the genuinely client-specific logic. That is how a seven-module platform moves from eighteen months to a twelve-week, three-stage delivery with an MVP in the leadership room at week six: not by working faster, but by not re-solving solved problems.
The caveat that keeps us honest
Accelerators work only on top of honest discovery. Retrofitting a module to a source system it does not fit produces the worst of both worlds - template rigidity with bespoke fragility. The discipline is in the gating: modules are deployed where discovery shows a genuine fit, and bespoke engineering is used where it does not. The accelerator compresses the predictable seventy percent; the engagement still earns the distinctive thirty.