Athena · RevOps & AI Systems Lead · Kate Khoury
One view of revenue in 90 days.
Finance, Sales and Operations read the same revenue number, from the same records, in Sigma. Every dollar traces back to a Stripe record and a HubSpot owner.
Walk the plan →See the data flow →
The story: agree what the numbers mean first, connect the systems second, automate the watching third.
How the numbers connect
Billing and CRM feed one certified view.
Start with the revenue problem
Then choose the tools
When billing and CRM disagree, everyone pays for it. Success is measured in revenue terms: a trusted forecast, a faster close, fewer revenue leaks.
Forecasts lose trust
Billing and CRM disagree, so month-end turns into spreadsheet work.
Reps start late
No view of payment status or renewal dates, so expansion and collections conversations start late.
Handoffs miss facts
Operations gets the closed deal without the billing facts it needs for onboarding.
Slide 3 · pause for corrections
What I’m assuming, so you can correct me
Tick each one as the panel confirms it. The approach holds either way; a fix now makes the rest land better.
Slide 4 · scope and sequence
Definitions before pipes
The order is the point. A sync built before definitions are agreed just moves disagreement faster. Pick a phase.
- Interviews with Finance, Sales, Operations, Tech and both specialists
- Current-state system and field inventory
- Baseline numbers: match rate, open exceptions, month-end hours
- Metric dictionary owned by Finance
- Source of truth for every field
- The customer ID crosswalk between Stripe and HubSpot
- Stripe to HubSpot sync
- Warehouse models
- Exception queue
- Parallel run against the old process
- Certified Sigma views in weekly reviews and month-end
- Old spreadsheet reports retired
- First AI monitor live
Out of scope for 90 days: pricing redesign, CPQ, revenue recognition rules, and history older than the backfill cutoff.
Slides 5 and 6 · how we work
The specialists own the build. I own the path.
One owner per decision. Executives get decisions to make, not status to read.
Four standing forums
Who owns what
They own technical execution and design proposals. I protect their capacity with a stop list, and each leaves owning a system end to end.
Slide 7
Five questions I settle before anything is built
These are where most billing and CRM projects quietly fail.
What is a customer?
One HubSpot company can hold several Stripe customers, and parent and child accounts need a rule.
What is the join key?
A Stripe customer ID on the HubSpot company, and the HubSpot company ID written to Stripe metadata.
What is ARR?
Finance’s definition, including how discounts, credits, usage, tax and currency are handled.
How does closed-won become a subscription?
And who links the deal to the subscription.
Which system wins for each field?
And who may edit it where. See the source of truth explorer below.
Slide 8 · how the data flows
Two paths, on purpose
An operational path into HubSpot for reps, and an analytical path into the warehouse for Sigma. Try breaking the sync.
Sigma reconciles from the source, so a broken sync can’t silently change a reported number.
Figures are illustrative.
Slides 5 and 7 · source of truth
Every fact has one home and one owner
Every other system reads it. Pick a fact to see where it lives.
What was billed and paid, subscription status, MRR
Slide 9 · when systems disagree
Every mismatch gets a type, an owner and a deadline
Matching runs daily in the warehouse. Nothing is a silent overwrite. Route each exception to its owner.
Rules that hold it together
- Billing facts follow Stripe; relationship facts follow HubSpot.
- The losing record is fixed at the source by its owner.
- Inside tolerance: logged and totaled, not chased.
- SLA by severity: current-month revenue fixed before close, the rest within a week.
- Finance signs off each month-end on the same model everyone sees.
Slide 10 · AI for continuous reporting
AI watches and explains. Models calculate.
The same human-in-the-loop model Athena sells to its clients. First candidate: an exception triage assistant, because the queue exists by week 8 and its value is easy to measure.
Use here
- Flag anomalies in daily reconciliation, with a plain-language summary
- Sort new exceptions, suggest a fix, route to the owner
- Suggest matches for orphan records, approved by a person
- Draft the biweekly executive update
- Answer reps’ billing questions from certified fields
Not here
- Calculating ARR, MRR or any number Finance relies on
- Writing to Stripe or changing billing records
- Merging or relinking records without approval
- Defining metrics or settling a Finance and Sales dispute
- Sending customer data anywhere InfoSec hasn’t approved
Validate before building
Slide 11 · risks
Risks I see, and what I’d do about them
Tap a risk to flip it to the mitigation. I’ll flag risk early and come with an ask.
Slide 12 · what done looks like
Something you could check at 30, 60 and 90 days
Targets in brackets are proposals to agree with Finance once baselines exist.
30days
- Metric dictionary v1 signed by Finance and the VP SalesSigned dictionary in the decision log
- Source of truth and decision rights published
- Baselines measured: match rate, open exceptions, month-end hoursShared with the data council
60days
- Crosswalk on active paying customersMatch rate [95 percent or better]
- Sync live for status, MRR, renewal date, payment statusReps see billing facts on the company record
- Exception queue running; parallel month-endReviewed with Finance
90days
- Certified Sigma views in forecast review and month-endOld spreadsheet reports retired
- Variance explained within toleranceUnexplained variance [under 1 percent of MRR]
- First AI monitor live; reconciliation time cutHours saved [about half]
What I need from you
A named Finance partner, about two hours a week
A Tech contact for integrations
Agreement on the stop list
Appendix · Q&A prep
Likely pushback, with short answers
Tap a question to reveal the answer. Practice out loud before you peek.
“Why not just turn on HubSpot’s Stripe integration?”
It may well be part of the answer, and I’d rather buy than build. In week two I’ll check what it syncs against the field inventory; anything it can’t cover, like subscription history for Sigma, goes through the warehouse path.
“Ninety days is slow. Can we have a dashboard in 30?”
Yes, a read-only interim view of Stripe revenue, labeled provisional. I won’t publish a “single view” before Finance signs the definitions; a second number nobody trusts is worse than none.
“Finance and Sales count ARR differently. Who wins?”
Finance owns the revenue definition. Sales keeps a bookings view, and Sigma shows a labeled bridge from bookings to billed to collected.
“What if the specialists disagree with your design?”
I want that early. They write short design proposals, we review together, and if we still disagree I make the call and write down why.
“What happens when Stripe and HubSpot disagree?”
Billing facts follow Stripe, relationship facts follow HubSpot. The mismatch becomes an exception with an owner, fixed at the source.
“How far back will you go with history?”
Active subscriptions in full, plus a window of invoice history agreed with Finance, probably two fiscal years. Older history stays in its current reports, labeled.
“Where exactly would you keep AI out?”
Anywhere it would calculate a reported number, write to billing, define a metric or merge records without approval.
“How would you know the AI is working?”
It has to beat a bar set before building: back-test accuracy, agreement in two weeks of shadow mode, and a measurable cut in hours or exception age. If it misses, we cut it.
“What gets deprioritized while this runs?”
I’d bring a proposed stop list in week two, built with the specialists, so it’s your decision rather than a surprise.
“How do you get reps to use this?”
A handful of fields, in the views they already use, and used in the weekly pipeline review. If billing status shows up in their forecast conversation, they’ll keep it accurate.
Questions I’ll ask in the debrief
- What does the Stripe and HubSpot connection look like today, and what has been tried before?
- Which downstream project depends on this most, and is there a date driving it?
- Where do Finance’s and Sales’ revenue numbers disagree most often today?
- What do the two specialists own now, and what would they most want from this project?
- Is Salesforce still part of the stack anywhere?