Conformance
36 checks, five categories, one score.
An automated audit of a storefront against the Open Storefront Protocol. It runs against the data we already hold about your storefront, scores five categories, and weights them into a composite from 0 to 100.
- Checks
- 36 across five categories
- Pass mark
- 60 or above, per check
- Tiers
- Four, from 50 up
- Validity
- 90 days. Automatic re-run is currently off.
Levels
Four tiers, from one composite score.
The weighted composite decides the tier. There are no per category minimums: a single number is compared against four thresholds.
- Platinum
- 90 to 100
- Gold
- 75 to 89
- Silver
- 60 to 74
- Bronze
- 50 to 59
Below 50 the run is recorded as failed. Failed is not a tier and carries no badge. There is no penalty for it and no limit on re-running.
Weights
What counts, and for how much.
- Protocol compliance
- 30%9 checks
- Data quality
- 25%10 checks
- Performance
- 20%6 checks
- Commerce readiness
- 15%6 checks
- Trust and reputation
- 10%5 checks
A single check passes at 60 or above. Category scores are the mean of their checks; the composite is the weighted mean of the categories.
The checks
All 36 of them.
Listed in full, because a check you cannot read is a check you cannot prepare for.
Protocol compliance
30% of the score- proto-001
- API key active
- At least one active storefront API key exists
- proto-002
- Storefront configured
- Storefront configuration exists and is active
- proto-003
- Products endpoint
- Products are available via the storefront API
- proto-004
- Search capability
- Knowledge graph built for semantic search
- proto-005
- Health endpoint active
- Storefront has received requests in the last 30 days
- proto-006
- Response format compliance
- Schema injection success rate above 80%
- proto-007
- Rate limiting configured
- Secret API key exists for authenticated rate-limited access
- proto-008
- CORS headers
- Cross-origin resource sharing headers configured
- proto-009
- Protocol version header
- OSP protocol version advertised on every storefront response, under both the current and the compatibility header name
Data quality
25% of the score- data-001
- Required fields present
- Products have title, price, currency and description
- data-002
- Description quality
- Product descriptions are longer than 50 characters
- data-003
- Valid prices
- Product prices are above 0 with correct format
- data-004
- Minimum product count
- At least 10 products in the storefront
- data-005
- Product images
- Products have image URLs set
- data-006
- No duplicate products
- Products have unique titles
- data-007
- Categories assigned
- Products have categories assigned
- data-008
- Knowledge graph coverage
- Knowledge graph entities exist for more than 50% of products
- data-009
- Structured data generated
- Schema.org structured data exists for products
- data-010
- Content freshness
- Products updated within the last 30 days
Performance
20% of the score- perf-001
- Average response time
- Average response time under 200ms
- perf-002
- P95 response time
- P95 response time under 500ms
- perf-003
- Uptime
- Uptime above 99% over the last 30 days
- perf-004
- Cache hit rate
- Cache hit rate above 50%
- perf-005
- No server errors
- No 5xx errors in the last 7 days
- perf-006
- Health endpoint speed
- Health endpoint responds in under 50ms
Commerce readiness
15% of the score- comm-001
- Dynamic pricing
- Dynamic pricing configured, or explicitly opted out
- comm-002
- Active offers or negotiation
- At least one active offer, or negotiation enabled
- comm-003
- Bundle system
- At least one active bundle configured
- comm-004
- Storefront template
- Storefront configuration present
- comm-005
- Webhook endpoint
- Webhook endpoint configured to receive events
- comm-006
- Completed transaction
- At least one completed transaction
Trust and reputation
10% of the score- trust-001
- Trust score
- Trust score above 50
- trust-002
- No unresolved disputes
- No unresolved disputes
- trust-003
- Feedback rating
- Average agent feedback rating above 3.0
- trust-004
- Query responsiveness
- Storefront responded to agent queries in the last 24 hours
- trust-005
- No margin violations
- No pricing margin violations in the last 30 days
Validity
90 days, and the re-run is currently off.
- Valid for
- 90 daysFrom the day it is issued.
- Re-runs within
- 7 days of expiryA daily job picks up certificates inside that window and re-certifies them. That job is currently switched off and has not run since 29 August 2026, so until it is turned back on, expiry is expiry and a re-run is something you start yourself.
Badge and API
Reading a certificate.
An SVG, at two sizes.
A certificate comes with a badge you can embed, generated as SVG at 200 by 80 and at 120 by 28.
Every badge links to a public verification page that reads the live certificate rather than a cached copy, so a revoked or expired certificate stops verifying immediately.
On the storefront surface.
A storefront's current certification sits on its own storefront surface and needs no key.
curl "https://cresva.ai/api/storefront/$BRAND_ID/certification"Questions
Common questions.
What happens if I fail?
Nothing. A run below 50 is recorded as failed, you get the per check results, and you can re-run whenever you like. There is no cost and no cooling off period.
Is there a manual review?
No. Every check is automated and runs against the data we already hold about your storefront rather than crawling you. There is no human review step at any tier.
Can I dispute a score?
There is no dispute process. Each check returns what it measured and why it scored as it did, so a result you disagree with is a specific check to look at rather than a decision to appeal. If a check is wrong, that is a bug worth reporting.
Do I need a minimum in each category?
No. The composite is a weighted mean and it is the only thing compared against a threshold. A weak category can be carried by strong ones.
Does certification affect how agents rank me?
Not by any mechanism Cresva controls. A certificate is evidence a storefront publishes about itself. What an agent does with it is the agent's decision, and we do not operate any ranking that a tier feeds into.
How long does a run take?
The checks read data we already hold about your storefront rather than crawling you, so a run is quick. We are not publishing a duration figure until enough real runs exist to quote a measured one.
Conformance runs from the dashboard.
It is configured per brand in the Cresva dashboard, alongside the storefront settings. The specification is what the protocol checks are written against.