The most widely deployed enterprise data pipeline in the world is a spreadsheet attached to an email. It has no schema enforcement, no audit trail, no access control, and a merge process consisting of a coordinator copy-pasting rows on Friday afternoons. We know because we keep being hired to replace it.
This post walks through one such engagement - a client's supplier onboarding workflow, shared with permission and anonymized - and the design principles behind FormOK, our form platform, that the project validated.
The starting point
The client onboards several hundred suppliers a year. Each onboarding began with a 40-column Excel template emailed to the supplier: company details, banking information, tax classification, compliance declarations. Suppliers filled it in - or filled in most of it, or renamed columns, or pasted formatted text into date fields - and emailed it back.
A three-person operations team then spent their week on validation archaeology: chasing missing tax IDs, normalizing "USA / U.S. / United States", and retyping everything into the ERP. Median onboarding time was 18 business days, and the team's own estimate was that a third of submissions required at least one correction round-trip by email.
The failure was not the spreadsheet. It was where validation lived: at the end of the pipeline, in human heads, after the data had already traveled.
Move validation to the point of entry
The rebuild replaced the template with a FormOK form. The structural change is simple to state: every rule the operations team applied by eye became a rule the form enforces at submission time.
- Field-level validation - tax ID formats by jurisdiction, IBAN checksums, registry number patterns - rejects malformed input while the supplier is still looking at the field
- Conditional sections mean a domestic supplier never sees, and can never half-fill, the cross-border compliance block
- File uploads for certificates are constrained by type and size, and land in object storage with the submission, not in a mailbox
A validation error at entry costs the supplier thirty seconds. The same error discovered downstream cost the client a multi-day email round-trip. That asymmetry is the entire business case, and it is why "make the form stricter" was the single highest-leverage change in the project.
Structured output is the product
A submitted FormOK form is not a document to be read; it is a JSON record with a stable schema, delivered by signed webhook to whatever sits downstream - in this case, middleware feeding the ERP. Field keys are versioned, so the form can evolve without silently breaking the consumer. The operations team's copy-paste step did not get faster; it stopped existing.
Approval did not disappear - banking-detail changes still require a human sign-off, for good reason. But the reviewer now approves a validated, diffable record inside a workflow with an audit trail, rather than eyeballing column AN of an attachment. Review became judgment work instead of data entry.
Results
Median onboarding time went from 18 business days to 4. Correction round-trips dropped from roughly one in three submissions to under one in twenty, and the remaining ones are substantive questions, not formatting repairs. The operations team did not shrink; it absorbed a supplier-base expansion that would previously have required two more hires, and the compliance office gained something it never had: a queryable record of who submitted, changed, and approved what, and when.
The general lesson
The pattern generalizes far beyond procurement. Wherever an organization moves structured data by unstructured means - onboarding, compliance attestations, internal requests, partner reporting - the same three moves apply: validate at entry, emit structured records, and make the approval trail a property of the system rather than a habit of the team.
Spreadsheets are a fine analysis tool. As a data transport, they are deferred validation with a friendly icon - and the deferral is always paid for by someone, at the most expensive point in the pipeline.