What is creator analytics, and how do you build it in Metabase?

Creator analytics turns course sales, digital orders, enrollments, and membership pledges into shared numbers about what a creator business actually earns. Every platform in this category reports its own slice accurately and stops there: Teachable can tell you what Teachable sold, Patreon what patrons pledged, Gumroad what shipped through its checkout. None of them can total those, subtract the other's fees, or put a course sale next to the email campaign and ad spend that produced it. That's the job — Metabase is where gross becomes net, and where the platforms' separate numbers become one P&L.

What does creator analytics measure?

• What did we actually earn this month, across every platform at once?
• How much of gross revenue is tax, platform fees, affiliate commission, and refunds?
• Which courses and products earn, and which quietly hand money back?
• Do the people who buy a course ever finish it?
• Is recurring membership revenue growing, and is churn cancellation or failed payment?
• Do buyers come back for a second product, or is every launch cold?
• Which email campaigns and ad channels produced the sales — a question no platform can answer alone?
• Does the money that reached the bank match the revenue we recognized?

Which tools feed creator analytics?

Only Thinkific has a managed warehouse connector — official Fivetran, plus a community Airbyte source. For Kajabi, Teachable, Gumroad, and Patreon there is nothing on Airbyte, Fivetran, or dlt as of July 2026, so the pull is yours to schedule. Each integration guide carries the working script and the per-platform rate limits.

What shared data models should you build?

Six tables carry the whole category. Map every platform onto these names and a card written once works regardless of which platform the rows came from.

• creator_orders — one row per sale or transaction, with gross amount, tax, platform and processing fees, refunds, and net kept as separate columns rather than a single pre-netted number
• creator_products — one row per course, offer, digital product, or pledge tier, with price, currency, and whether it's recurring
• creator_enrollments — one row per student per course, with enrolled, activated, and completed timestamps plus percentage complete, and a flag separating free or comped enrollments from paid
• creator_lesson_progress — one row per student per lesson, with lesson position and completion time: the drop-off curve lives here and nowhere else
• creator_members — one row per member per daily snapshot, with status, entitled amount, pledge cadence, and charge dates; this is an append-only history table, not a current member list
• creator_payouts — one row per payout to your bank account, with period start and end, so recognized revenue can be reconciled against money that actually arrived

Three modeling rules save most of the pain later. Divide cents once in the model, never per card — several of these APIs report money in cents. Normalize pledge cadence before summing, since an annual pledge is not twelve times a monthly one in any given month. And keep free, comped, and bundled enrollments flagged, because they almost never complete and blending them halves an otherwise healthy completion rate.

Which creator metrics matter most?

• Platform fee rate — the effective take-rate, derived from fee lines rather than the published rate card.
• Average order value — by product and tier, counting orders rather than payment-plan installments.
• Course completion rate — with a stated maturity window and free enrollments held out of the denominator.
• Repeat purchase rate — the earliest evidence that a catalog beats a single hit product.
• MRR — recurring membership and pledge revenue on one cadence-normalized basis.
• Churn rate — genuine cancellations, with failed payments counted as a separate, recoverable problem.
• LTV — lifetime support per member, read as a median because a handful of whales distort the mean.
• ARPU — average revenue per active member, which tier mix moves more than price changes do.

How do you connect tools to Metabase?

Metabase queries SQL databases and warehouses; it does not call SaaS APIs. So every route below is really a way of getting platform records into a database Metabase can read.

1. CSV upload first — export a slice from the platform API, then mb upload csv to create a table and a model in one command, and mb upload replace to refresh it. On Metabase Cloud there is nothing further to provision, and this audience is largely non-technical, so this is where to start rather than where to compromise.
2. Scheduled API pulls — the durable route for four of the five platforms: Kajabi's Public API for orders, transactions, and payouts; Teachable's Transactions API; Gumroad's /sales; Patreon's Members endpoint appended daily with a snapshot date. Small data, modest jobs, but respect each platform's paging and limits.
3. Managed connector where one exists — Fivetran for Thinkific orders, enrollments, users, products, coupons, and promotions, with incremental enrollments; the community Airbyte source covers similar ground without incremental sync.
4. MCP for exploratory questions — Thinkific's read-only server for students and enrollments, Kajabi's for content and product operations. Neither returns financial data, so neither replaces the pull.

Which dashboards should you build first?

• Course revenue — gross, tax, fees, refunds, and net by product and month. Needs one table.
• Membership growth — active members, net change, tier mix, and failed payments.
• MRR / ARR — recurring revenue with monthly and annual pledges brought onto one basis.
• Churn & retention — cancellations and cohort retention, built from snapshot history.
• Repeat purchase — second and third purchases across the catalog, keyed on buyer email.

Common mistakes

Related

Integrations

Analytics

Dashboards

FAQ