Your Amazon data remains yours.

Data we process

When a seller connects their Amazon account, Impact Finder reads, with the seller’s permission: the seller or vendor identifier and the marketplaces they sell in; listings and catalogue details, including the seller’s own prices; fee estimates; sales and traffic by product and day; search, basket and repeat-purchase analytics for brand owners; review topics for top products; and fulfilment inventory, storage, fee and reimbursement reports. It also stores the Amazon refresh token that keeps the connection working.

It keeps what MrPrime staff add or do in the service: cost prices entered for a client’s products, the business name and Amazon stores for each connection, the client’s own Slack channel if one is connected (its posting address is stored encrypted), a record of which staff member connected or removed an account, and, for failed staff sign-ins, the email address and network address used.

Impact Finder does not request buyer personal data. It makes no calls to Amazon’s order, buyer or messaging services, so it does not receive buyer names, addresses, payment details or messages.

Why we use it

We use this data to run the service our clients contract for: keeping the connection working, collecting read-only data from Amazon, checking that each source is complete and current, finding and valuing opportunities, producing client reports, giving support and protecting the service. We do not sell Amazon account data, and we do not put one client’s identifiable data in another client’s workspace or in the public demo.

When the export between the two products is configured, the same data also gives HaloProof, another MrPrime product, a daily baseline for that client. The list of providers below says exactly what it receives.

Who can see client data

Impact Finder is used by MrPrime’s team to run a managed service; clients do not have their own logins. Today every MrPrime staff member who signs in can see every client’s data, because there are no per-client permissions yet. Limiting each person to the clients they work on is planned. Account managers share client reports with the client they belong to.

MrPrime staff also use AI assistants, currently OpenAI Codex and Anthropic Claude Code, to release, run and check the service. While doing that they may read production data, including client account data. What they read is sent to OpenAI or Anthropic to be processed and is kept in the assistant’s session records on the staff member’s computer.

Amazon access is read-only

Impact Finder only reads from Amazon. Its code allows a fixed list of read operations and report types and refuses any other request before it leaves our server, so it cannot change listings, prices, stock, advertising or account settings. Apart from exchanging its stored token for short-lived access, the only requests that are not plain reads ask Amazon to prepare a report or to estimate fees, and neither changes the seller’s account.

Storage and protection

Impact Finder runs on OpenAI’s ChatGPT Sites hosting. Its database and file store are Cloudflare D1 and R2, provided through ChatGPT Sites, and Cloudflare states that it encrypts both at rest. ChatGPT Sites does not offer a choice of the region where data is stored.

Amazon refresh tokens are encrypted with AES-256-GCM before they are stored, are bound to the connection and seller they belong to, stay on the server and are never sent to a browser. Raw Amazon report files are also encrypted by the app with AES-256-GCM before they are stored. The summaries built from them are compressed but not separately encrypted by the app; they rely on the storage provider’s encryption at rest.

Every stored Amazon record and file is labelled with the client, connection, seller and marketplace it came from. Impact Finder uses HTTPS for every connection, restrictive browser-security headers, origin checks and server-side authorisation.

The public demo contains fictional data only. Requests from the public demo domain to live Impact Finder data services are blocked before authentication, even if a browser also carries a staff session.

Service providers and who receives data

These are the organisations and services that handle Impact Finder data today:

• Amazon supplies the data, through the authorisation the seller grants.
• OpenAI ChatGPT Sites hosts the service, its server logs and its stored data. Storage is Cloudflare D1 (the database) and Cloudflare R2 (the file store), provided through ChatGPT Sites.
• Vercel runs a scheduler that wakes the service every minute so that data refreshes on time. It receives no client data; the service replies with counts only.
• GitHub holds the source code and runs automated checks. The running service sends it no client data. Some older client working files remain in the private code history, and removing them is planned.
• Slack carries alert messages. Once a day, if any account has missed a refresh or has a broken Amazon connection, a message to MrPrime’s own Slack channel lists those accounts by business name and Amazon store, with refresh dates and a short status; it contains no sales figures, product codes or seller IDs. If a client’s own Slack channel is connected, messages about that client’s account are posted there, and no other client’s information goes to it.
• AI assistants used by MrPrime staff, currently OpenAI Codex and Anthropic Claude Code, may read production data, including client data, while operating and checking the service, as described above.
• HaloProof, another MrPrime product, can receive a retained daily baseline for a connected client when the export between the two is configured: the seller ID and marketplace; seller-wide sales, units, sessions and page views by day; top products by ASIN with their sales, units, sessions, conversion and Buy Box share; and search performance by ASIN. It never receives Amazon tokens, raw report files or buyer data, and only accounts that are still connected are served.
• OpenAI’s API is used only by an optional AI review feature that works on fictional demonstration cases built into the code, with response storage turned off. It is never sent client Amazon data.

We do not sell Amazon account data. We disclose data where the law requires it or to protect the service and its users.

Retention, revocation and deletion

What we keep: the encrypted Amazon token, until the account is disconnected in Impact Finder; connection details; raw report files, encrypted; the summaries built from them; listing history and recommendation states; cost prices entered for the client; refresh records; and the record of who connected or removed each account. There is no fixed retention period yet. Setting one is planned.

When an account is disconnected in Impact Finder, its stored Amazon token is deleted straight away, so no more data can be collected. A seller can also revoke Impact Finder in Amazon at any time; collection then stops because Amazon no longer accepts the token, which stays encrypted in storage until the account is disconnected in Impact Finder.

Data already collected is not yet deleted automatically, including when a seller disconnects, revokes access or stops using the service. Automatic deletion of all of a seller’s data within 30 days of them revoking access is being implemented; until it is in place, data already collected is kept.

To disconnect an account, or to ask about access to, correction, export or deletion of your data, email info@mrprime.com. We check that the request comes from someone entitled to act for the business before acting on it, and we confirm the outcome.

Privacy and security contact

For privacy or security questions about Impact Finder, or to report a security concern, email info@mrprime.com.

Controller and further privacy information

Simon and Prime Ltd is the controller for its service administration and the processor of client-authorised Amazon data where the client determines the purpose of the work. For company details, individual rights, complaints and broader website processing, read the MrPrime privacy policy.