Ordercherry

Security and data protection

Ordercherry reads order and buyer data from Shopify to decide which offer to show and to report what each offer earned. This is how that data is protected, how it is backed up, and what happens if something goes wrong.

What we store

The settings of funnels, offers and surveys, and one row per offer shown, accepted or declined. A row carries the order it belongs to, the amount and, for the greeting and for targeting, the buyer's first name, country and Shopify customer ID. Offers on the product page and in checkout are counted per day without anything about the buyer. We do not store card numbers, full addresses, email addresses or phone numbers. The privacy policy has the whole list.

Encryption

  • All traffic runs over TLS 1.2 or higher. There is no plain HTTP endpoint.
  • The database is a managed PostgreSQL instance whose volumes are encrypted at rest by the hosting provider. Its automated backups have the same encryption.
  • Shopify access tokens are stored only in that database, never in source control.
  • The address a merchant gives their AI assistant is stored as a hash for sign-in and as an encrypted copy for showing it to the merchant again.

Access control

  • Access to production is limited to the app's maintainer. There are no shared logins, and no third party has a production credential.
  • Every account that can reach production has a unique password of at least 16 characters, kept in a password manager, and two-factor authentication.
  • Every read of a buyer's name, email, phone or address is written to an access log with the shop, the fields read, the reason and the order it was for.
  • After payment and on the Thank you page, the buyer's page talks to the app with a token Shopify signs for that checkout or order, and the app checks it on every request. Prices and discounts are set on the server and by Shopify, never taken from the browser.

Test data and production

Development and testing run against their own database. Merchant data never leaves the production database, test orders are never written into it, and production data is not copied into development.

Data loss prevention

  • The hosting provider backs up the production database every day, encrypted.
  • Backups can be restored to a point in time. A restore is exercised when the schema changes in a way that would be hard to reverse.
  • Code and configuration live in a private repository under version control, so a lost host can be rebuilt from nothing.
  • Deletions are deliberate and limited to one shop. A shop's data is deleted when Shopify sends the request to erase it, 48 hours after the uninstall, and no code deletes across shops.

Retention and deletion

Offer records are kept while the app is installed, so merchants can report on them. Uninstalling removes the shop's access tokens at once. Every other row of that shop is deleted when Shopify sends the request to erase the shop, 48 hours later. A request to erase a customer strips that buyer's name, country and customer ID from the records and deletes their survey answers, which leaves anonymous revenue figures. These run at /webhooks/compliance.

Incident response

If we suspect that personal data has been exposed, we do this.

  • Detect and triage. Platform logs and error reports are read on a report or an alert. An incident is opened when exposure is plausible, not when it is proven.
  • Contain. Rotate the affected credentials, revoke sessions and, if needed, take the endpoint offline. Containment comes before diagnosis.
  • Assess. Find which shops and which fields were affected, and over what time, from the access log above.
  • Notify. Affected merchants and Shopify are told without undue delay and within 72 hours of our becoming aware: what happened, what data it was and what they should do. We do not wait for a complete picture.
  • Recover and learn. Restore from backup if data was changed or lost, write up the cause and the fix, and ship the fix before the incident is closed.

Reporting a vulnerability

Write to silas@okayscale.dk with the details and how to reproduce it. We answer within two business days. Please give us a fair chance to fix it before you make it public.