• Explainer
  • E-commerce
  • 5 min read

Honest Data, Explained: Why the Commerce Engine Shows a Gap Instead of a Zero

No cost data. Not reported. Coverage stated. The four words that keep a money decision honest.

Author
ftes.ai Editorial
ftes.ai Team

The short answer

Honest data is the promise that the commerce engine never shows a zero where it means unknown: a figure that cannot be computed is labelled with the reason, coverage is stated when only part of the catalogue carries the inputs, test orders and catalogue errors are excluded and disclosed, and advice built on a degraded source goes quiet.

The role this piece is aboutData Trust FTE · Data Trust Officer

Key takeaways

  • Zero and unknown are different facts; the engine labels the second one instead of pretending it is the first.
  • Coverage is stated so a margin figure says what share of the catalogue it actually describes.
  • Test orders and catalogue errors are excluded from every calculation and the exclusion is disclosed.
  • A data-trust gate makes recommendations go quiet when their foundation is shaky, which is what the Data Trust FTE on ftes.ai is for.
  • Honest gaps are what let the other three laws mean something.

Why is a zero a claim?

A margin of zero says the product made nothing. A cost of zero says the product cost nothing. Both are strong statements, and a report that prints them when it actually has no cost record is making a claim it cannot support. Downstream, a buyer sorts by margin and the products with missing costs float to the top or the bottom for no real reason.

The commerce engine treats an unknown as an unknown. Where a cost is missing the product analyst shows No cost data; where a source does not report a metric the page shows Not reported. Nothing about the product is hidden, but nothing about it is invented either, and the sort order is not poisoned by a placeholder.

Why is coverage stated next to a figure?

When only part of the catalogue carries cost or landed-cost inputs, a brand-level margin is only as complete as that part. The engine states the coverage next to the figure, so a reader knows whether a margin describes most of the range or a slice of it. The number is still shown, because a partial truth is useful; it is just labelled as partial.

This matters most in the first weeks after connecting, before cost records and shipments have been loaded for the whole range. The Morning Brief and the Dashboard both carry the coverage so the owner does not read an early figure as a complete one. As inputs are added, the coverage rises and the label changes with it.

Test orders and catalogue errors are excluded and disclosed

Every store has test orders from a checkout trial, a duplicated variant, a product with a price entered in the wrong currency, or a placeholder SKU that never sold. Left in, they distort averages and produce alerts about products that do not exist. The engine excludes them from every calculation.

The exclusion is disclosed rather than silent. A note on the affected surface says what was excluded and why, so an owner who remembers placing a test order last week can see it was recognised, and an accountant reconciling the sales tax report knows the two lines that are missing are the ones that should be.

The data-trust gate makes advice go quiet

The engine knows when a source is fresh, stale, degraded or failed, because every sync run is logged in the run ledger. When a source is degraded, the recommendations that depend on it are withheld or labelled, and the page says which source is the problem. A reorder proposal built on an inventory feed that failed at two in the morning does not reach the Morning Brief at nine.

On ftes.ai this is the job of the Data Trust FTE: it reconciles orders, refunds, inventory and tickets across systems, publishes a trust state per data domain, and no other role acts on a domain that is not fresh. The engine's gate and the role's job description describe the same mechanism from two sides.

Honest data is what makes the other laws work

The backend calculating deterministically is only valuable if the inputs are real. The AI interpreting only figures that exist is only safe if a missing figure is marked as missing. A human approving a send is only meaningful if the evidence on the Review & Send screen is complete or labelled as incomplete. Honest gaps are the promise the other three stand on.

It also changes the tone of the product. A quiet Morning Brief on a day when nothing crossed a threshold is reported as a legitimate outcome, not filled with a manufactured alert. A surface that says Not reported is being straight with the reader, which is how trust with a store owner is actually earned.

What an owner should do with a gap

A gap is an instruction. No cost data on a best seller means loading the supplier invoice, or recording the shipment so the landed-cost engine can date the cost correctly. Not reported on a store's footfall means the people-counting integration is not connected for that location. Coverage stated on a brand margin means the rest of the range needs its inputs.

The Integrations page shows each connected system's health and the run ledger shows its last successful run, so a gap can be traced to its source in a minute. Closing gaps is the first-month work of a brand on the engine, and each one closed makes the next Morning Brief more complete.

Frequently asked questions

What does the engine show when it cannot compute a figure?

The reason: No cost data, Not reported, or the coverage the figure describes. Never a zero that means unknown.

Are test orders included in revenue?

No. Test orders and catalogue errors are excluded from every calculation and the exclusion is disclosed on the surface.

What happens when a sync fails overnight?

The run ledger records it, the source is marked degraded, and recommendations that depend on it are withheld or labelled until it recovers.

Which role owns this on ftes.ai?

The Data Trust FTE publishes a trust state per data domain, and no other role acts on a domain that is not fresh.

Does a gap mean the product is broken?

No. It means an input is missing, and the page tells you which one so you can add it.

  • E-commerce
  • Commerce engine
  • Data Trust FTE
  • Digital FTE
Further reading

Keep going.

Related guides on hiring and managing Digital FTEs.

Ready to hire your first Digital FTE?

Pick a role, connect the tools it needs, and review its first proposals this week.