Revenue analyst agent for a DTC ecommerce brand on HubSpot that turns quarterly board prep into a reviewed draft

What The Revenue Analyst does

A 30-person DTC brand sells a few hundred SKUs through its own storefront, with occasional marketplace orders. Every quarter the operations lead spends about a week exporting orders from the ecommerce platform, refunds from the payment processor, returns from the returns portal, spend from three ad platforms and ticket counts from the help desk, then pasting them into a spreadsheet and a board deck. The numbers rarely tie out because gross sales, net sales and refunds are cut on different dates in each system, and the peak-season quarter always looks odd next to the others. The Revenue Analyst works for the RevOps and operations team and the founders. It owns the metric definitions (net revenue after returns, contribution margin per order, blended and channel CAC, repeat purchase rate, cohort revenue, contact rate per order), the nightly sync into a small warehouse, plain-English answers to ad hoc questions in Slack, the Monday executive digest, and the first draft of the quarterly board pack. It does not own pricing, inventory or ad decisions; it surfaces the numbers and flags anything it cannot reconcile.

Built on HubSpot

HubSpot is the right home for customer records, email and SMS revenue, and lifecycle data, and Breeze can handle quick questions about contacts and campaigns. It is not built to hold every order line, refund and shipment for a high-volume, low-ticket brand, or to reconcile processor payouts against storefront sales. A custom AI build on a small warehouse, with n8n handling the nightly syncs and read-only MCP connectors to the warehouse, HubSpot and Help Scout, gives the agent one governed model to query. Databox can sit on the same warehouse for the always-on dashboard, so the agent focuses on answers, explanations and the board draft. Because the team asked for it to be built and maintained for them, the design keeps moving parts few and logs every query so upkeep stays predictable.

What starts it

• Monday 7am: generate and send the weekly executive digest covering the prior week versus the same week last year
• A team member asks a question in the dedicated Slack channel or in HubSpot (for example, what share of Q3 revenue came from repeat customers)
• Quarter close plus the end of the return window: assemble the draft board pack
• Nightly sync detects a reconciliation break, such as storefront net sales differing from processor payouts beyond the agreed tolerance
• A tracked metric moves outside its seasonal range, such as refund rate or tickets per order spiking after a promotion

What it reads

• Ecommerce platform orders, line items, discounts and customer IDs (via API, landed nightly in a warehouse such as BigQuery)
• Payment processor charges, refunds, chargebacks and payouts
• Returns portal and 3PL or shipping data: return reasons, restock status, ship dates, carrier exceptions
• Ad platforms (paid social, search, shopping) daily spend and attributed orders
• HubSpot contacts, lifecycle stages, email and SMS campaign revenue, plus Help Scout ticket tags (order status, returns, shipping) for contact rate
• A metric dictionary and COGS and fulfillment cost table maintained in Airtable

When a person takes over

• Reconciliation breaks the agent cannot explain (for example, payouts short of net sales with no matching refunds) go to finance, because they usually mean a processor fee change, a chargeback batch or a missing data feed that needs a person to investigate
• The board pack narrative and any forward-looking statement are written or approved by the founders; the agent drafts the numbers and a factual commentary only
• Requests to change a metric definition, COGS table or attribution rule go to the operations lead, who updates the dictionary and decides whether past quarters are restated
• Questions the data cannot answer cleanly (such as why a cohort churned, or which creative drove a spike) are returned with what the data does show and routed to the marketing or CX owner rather than guessed at

Rules it follows

• Never report gross sales as revenue: board and digest figures use net revenue after discounts, refunds and returns, and any period still inside the return window is labeled provisional until it closes
• Always compare peak-season periods to the same retail weeks last year, never to the prior quarter or month, and call out promo and holiday shifts that change the comparison
• Read-only access through MCP connectors: the agent cannot edit orders, refunds, HubSpot records or ad accounts, and cannot send anything outside the leadership channel and digest list
• No customer-level personal data in digests or board materials; questions return aggregates, and order-level lookups are limited to named RevOps users
• If a source has not synced or a reconciliation check failed, the agent says so at the top of the answer instead of filling the gap with an estimate

How you know it works

• Hours the operations lead spends preparing the quarterly board pack, measured against the current week of manual work
• Share of board pack figures accepted by finance without manual correction
• Reconciliation breaks found and resolved before the digest is sent each week
• Number of ad hoc data questions answered in Slack without someone building a spreadsheet
• Time from quarter-end return window close to an approved board pack

From pilot to live

1. Pilot: last quarter, rebuiltConnect the storefront, processor, returns portal and ad platforms to the warehouse, agree the metric dictionary, and have the agent rebuild last quarter's board figures. Finance compares them line by line against the manually built deck until differences are explained or fixed.
2. Weekly digest and Q&ATurn on the Monday digest and Slack questions for the founders and operations lead, add Help Scout and HubSpot data, and tune seasonal comparisons and tolerances. Run through at least one promotion so the anomaly alerts are calibrated before peak season.
3. Full board cycle and ongoing upkeepProduce the next quarterly board pack from the agent with human sign-off, then freeze it as the official snapshot. Ongoing maintenance covers broken API connections, new sales channels or ad platforms, schema changes from the ecommerce platform, and monthly accuracy spot checks before each peak season.