Why files, not connectors: scheduled exports as the integration for sales analytics

Analytics vendors list their connectors like trophies. Each one is a promise to hold a customer's credentials, understand that customer's schema, and answer the phone when the other vendor's API changes. This guide makes the case for the alternative that every large organisation already runs: a file, on a schedule.

What a connector costs

Multiply by the number of systems: CRM, billing, warehouse, ERP, service desk. Each is a small product, and together they are a roadmap.

What a file costs

One format, chosen by the vendor. No credentials to anything. The customer's IT team, who already know their systems, schedules the export they already run for someone else. The security review is about one file: where it goes, who sees it, how it is deleted.

The trade is freshness for simplicity, and the freshness given up is freshness that coverage and share analytics never needed.

The four doors

1. Upload in the browser. The way to start. The file is read on the customer's machine, checked, and only stored if they choose.
2. Push to an API. The customer's system posts the same file to one endpoint on a schedule, with a key the vendor issued. The vendor holds no access to the customer.
3. Drop in storage. The customer's job writes the file to an S3 bucket, Azure container or SFTP folder the customer owns. The vendor reads that one place, on a schedule or on demand. The enterprise pattern.
4. A shared sheet. For the smallest teams, whose data is the sheet. Point at it; it refreshes.

All four deliver the same file to the same checks and produce the same report.

When an adapter is still right

A source that genuinely cannot export a file, or a customer whose IT policy forbids scheduled exports to a third party. Then an adapter is built, once, for that customer, in an enterprise engagement, and it runs on the customer's side, producing the file into the customer's own storage. The vendor still never holds the keys. If three customers need the same adapter, it becomes a product decision made with three customers' evidence rather than a roadmap made with none.

The freshness question

Nothing on the list is served better by a live connection than by a nightly file.

Where it goes wrong

Nine formats instead of one. A vendor that accepts each system's native export has nine schemas again. One format, mapped once per source.

No validation on arrival. A file that is not checked is a connector's problems with none of its benefits. The report on every file is the whole point.

Idempotency ignored. The same file sent twice doubles the roll-up. A hash per file per customer catches it.

Freshness promised, not measured. Show the timestamp of the last file on every screen.

How Covirage takes data in

Upload, API push, storage drop and Google Sheets, one format, the same validation report on every file. The upload page is the first door, and the help centre covers the others.