Projects BI Smart has delivered

The situation before

Vessel maintenance run on spreadsheets and on the technician's memory. When that person left, the equipment's failure history left with them. Unplanned downtime was an event, not a number, and cost per vessel was not comparable.

What we did

• Functional hierarchy from vessel down to component, with an explicit rule for where tracking needs a serial number and where a batch is enough.
• Preventive plans by meter, not only by calendar: an operation that measures engine hours cannot have maintenance scheduled by month.
• Work orders bound to the object's site, so labor and material cost land in the right unit's accounts.
• Mean time between failures and mean time to repair calculated from the ERP's own history, with no parallel spreadsheet.

What changed

Unplanned downtime became a tracked KPI, maintenance cost became comparable across vessels, and failure history now survives a change of crew.

The situation before

A make-to-order machine manufacturer with engineering living in spreadsheets. Every design change meant manual re-entry, there was no reliable cost per machine sold, and after-sales could not tell which part had gone into which unit.

What we did

• Routing before product structure: a component line points at an operation, so the operation has to exist first.
• Engineering revisions with backdated validity, because registering everything with today's date makes order planning fail silently.
• Full engineering change cycle, parking the open order to switch revision without losing what was already reported against it.
• Production orders consuming existing stock instead of manufacturing what is already on the shelf.
• Batch and serial tracking binding the delivered part to the machine that shipped, so recall and warranty have an answer.

What changed

The company now has real cost per machine sold, engineering changes no longer require re-entry, and after-sales can start from a defective part and reach the customer who received it.

The situation before

Operations, maintenance and finance closing the month on separate bases. The board meeting opened by arguing where each number had come from, and every department arrived with a different value for the same question.

What we did

• A layered lakehouse, with incremental loading separated from the consumption layer, so reprocessing does not knock over whoever is reading.
• Star schema and DAX measures with written definitions: one source per KPI, and the calculation rule documented outside the head of whoever built it.
• Row-level security, so each area sees what belongs to it without duplicating reports per permission.
• Paginated reports and exports for the people who still close the month in a document, without forcing everyone to change routine at once.

What changed

The board now argues about the number instead of where the number came from, and closing no longer depends on manual consolidation between departments.

The situation before

Sales closed in Salesforce and someone re-typed it into the ERP. Customers registered twice under different spellings, contracts that did not match what had been sold, and billing chasing the information.

What we did

• Customer and contract mapping between both systems, with the ERP owning the official document and the CRM owning the commercial process.
• Idempotent synchronization: reprocessing the queue duplicates neither orders nor invoices.
• Concurrent edit conflict handling, because both sides write to the same record.
• An audit trail on every write, with field errors surfacing instead of being swallowed by the integrator.

What changed

Sales stopped re-typing into the ERP what they had already entered in the CRM, and customer records stopped being born duplicated.