Order-line aggregates only. No shopper PII imported, ever.
Per-tenant schemas with row-level security on every table.
TLS 1.3 in transit, AES-256 at rest, keys rotated quarterly.
SSO / SAML and scoped API tokens on Enterprise.
SOC 2 Type II in progress, report available under NDA.
Every applied price versioned; full catalog rollback in one action.
MAP enforcement
Minimum advertised price is treated as a constraint on the optimizer, not a suggestion to the merchandiser.
- Hard floor
- Suggested price is clamped to max(MAP, landed cost × min-margin). No override flag exists.
- Breach detection
- A competitor selling below your MAP triggers an alert with screenshot evidence and timestamp.
- No retaliation pricing
- CogniSpan never cuts your price to match a MAP breach; it escalates to you instead.
- Audit trail
- Every suggested and applied price is stored with the inputs that produced it, for 36 months.
What we commit to
| Platform uptime | 99.9% | Enterprise contractual; measured monthly |
| Scrape freshness | < 60 min | Growth plan, per tracked competitor domain |
| Model refit | Nightly | 02:00 in your store's local timezone |
| Support first response | 4 hours | Business hours on Growth, 24/7 on Enterprise |
| Incident notification | < 30 min | Status page plus email to admins |
| Data export | Self-serve | Full CSV of prices, elasticity and history |
Common diligence questions
QDoes it work if some SKUs barely sell?
Yes. Thin-data SKUs borrow strength from their category through a hierarchical model, and the confidence score tells you exactly how much to trust each recommendation. Below 0.35 confidence we recommend holding price.
QDo you need customer data?
No PII. CogniSpan reads order-line aggregates only: SKU, date, price, units, cost. Nothing about individual shoppers is imported, stored or modelled.
QHow is MAP enforced?
MAP is a hard floor inside the optimizer. A suggested price can never fall below it. When a competitor breaches your MAP, we raise an alert with evidence rather than cutting your price to match.
QWhat if the model is wrong?
Nothing moves without your approval on Starter, and on Growth every automated move respects your floor, ceiling and ±7% daily cap. Every applied price is versioned and reversible in one click.
QHow long is onboarding?
About thirty minutes of work on your side, then one overnight model run. Most brands review their first suggestion set the next morning.
QHow long is data retained?
Rolling 36 months of price and demand history, exportable as CSV and deletable on request within 30 days.
QWhat happens after I join the waitlist?
We onboard in cohorts of 25. You receive a scheduling link and a sandbox project preloaded with a catalog shaped like yours so you can judge the output before connecting anything.