Privacy policy

Advert Lumen helps advertisers review and compare Meta ad creatives before they spend. This policy explains what we collect, why, where it is processed, how long we keep it, and how to delete it.

What we collect

• Your account details: email address, name (if given), sign-in method, and an optional second-factor secret, stored encrypted.
• Ad reports from Meta, read-only: if you connect Meta, we request one permission, ads_read, and read the ad accounts you select: account name, currency and timezone; campaigns, ad sets and ads (objective, status, budget, schedule, targeting settings, placements); ad creatives (headline, text, link, call to action, image and video references); and daily performance reports (impressions, clicks, spend, results) with age, gender, placement and region breakdowns. We never read person-level data, custom audience membership, pixel or Conversions API events, Page posts, comments or messages.
• What you give us: drafts, creative variants, uploaded images, landing pages you point us to, and Ads Manager exports you import.
• Usage and billing records: credit and token ledgers, an audit log of connections and role changes, and (from paid plans) billing details held by Stripe.

How we use it

Only to measure and improve your own Meta campaigns: showing your account's history, reviewing and comparing your creatives, sizing your tests, and comparing our predictions with your results. We do not use one customer's ad data for another customer, we do not retarget, and we do not combine your Meta data with other ad platforms.

Language-model processing

To review creatives we send their text, image descriptions, your draft and qualitative summaries of how your audience has behaved to Meta's Model API (Muse Spark). Prompts contain no identifiers of you or of any person. We use paid usage only and never the discounted "contributor" tier, whose terms allow Meta to train on prompts. Meta's Model API terms govern how Meta handles the data it processes; we do not describe that processing as "no training" or "no retention" unless those terms say so in writing. Everything a model wrote is labelled generated; no real people were surveyed.

Where it is processed

In the United States, with these processors: Fly.io (hosting for the web app, API and workers), Neon (database), Upstash (queue, through Fly.io), Tigris Data (file storage, through Fly.io, kept in the US), Meta (Model API), Resend (email), Stripe (payments, paid plans only) and Sentry (error reports, with secrets scrubbed).

Security

Meta access tokens are encrypted with per-token keys wrapped by a key held in our host's secret manager, and never appear in logs. Every customer's data is isolated by organisation and workspace at the database level.

How long we keep it, and deletion

• Disconnecting Meta (in Settings, or by removing the app in Meta Business Manager) deletes the synced ad data and everything derived from it within 24 hours, and redacts past results that used it. Drafts and reviews you wrote are kept unless you ask us to delete them.
• Cancelling keeps read access for 30 days, then deletes your organisation's data. You can export it first.
• Billing and usage ledgers are kept for 7 years for accounting, with organisation identifiers pseudonymised.
• Backups expire within 30 days.

Meta can also ask us to delete data about a person; see data deletion for the process and a status lookup.

Your rights

You can access, export, correct or delete your data at any time from Settings, or by writing to privacy@advertlumen.com. Requests under the GDPR or the CCPA run through the same deletion process.