Case study
The $82M Definition
Three orgs, three numbers, and a fight about who got credit. The fix wasn't a better dashboard. It was a definition every org could run on its own data, starting with correcting my own.
- $82M of revenue moved into the segment already managing it
- Three competing definitions replaced by one, with SVP sign-off
- Identification fixed at the source, not patched in reporting
- Standardized datasets adopted across SMGS
Context
Startup is the top of AWS's funnel. A company arrives small, and if things go well it stays for a decade. But one Salesforce customer can own many AWS accounts, and the legacy lead-routing logic had placed a lot of those accounts in other segments, ISV, DNB, Enterprise, SMB, even when the customer was a startup our teams were actively working.
So Startup teams were monitoring and developing accounts while another segment took the comp credit. WW Sales, Segment, and Startup Operations each had their own number for how much revenue this was, and the disagreement was not really about measurement. It was about credit.
The problem, framed properly
The tempting framing was that my number was right and theirs were wrong. That framing loses. What actually mattered was that the number fed coverage decisions, comp and quota attainment, segment classification, and annual planning, so while it stayed contested, every decision downstream of it was negotiable.
Underneath the politics sat an identity problem. Account-level revenue never rolled up to the right customer, because the routing logic skipped M&A events and ignored the customer-identification signals, zip code, payment card, and email domain, that would have tied a set of AWS accounts back to one Salesforce customer.
How to settle a contested definition
Escalate and let leadership pick a number
PassedFast, and it holds until the first quarter someone loses money by it. A definition chosen by authority gets relitigated by whoever it costs, and the relitigating is the expensive part.
Negotiate a compromise between the three positions
PassedSplits the difference between three numbers, none of which were anchored in how customers actually enter AWS. A compromise nobody can derive from first principles is a number nobody defends when it is inconvenient.
Anchor the definition in the funnel
ChosenStartup is where customers enter. If an AWS account can be proven to belong to a Startup Salesforce customer, it belongs to the Startup segment. That is a rule every org can run against its own data, which makes it checkable rather than arguable.
Why a rule beat a negotiation
A rule each org could apply itself changed what the meetings were about. Instead of defending my number against theirs, I could hand a segment its own accounts and let it run the test. Disagreements became specific and account-level, which is the only size at which they get resolved.
The second decision was the one that made the number survive scrutiny. Identity matching alone over-attributed. A zip code, a card, and a domain in common are suggestive, not conclusive, and a methodology that moved accounts on suggestion would have been torn apart in the first review. So an account also had to show Startup program evidence before it counted, specifically credits received or a Startup opportunity created in Salesforce. Requiring evidence on top of identity is what made $82M defensible instead of merely large.
What I built and wrote
The account-level evidence came first, because the 6-pager was only going to be as good as the data behind it. Then the argument in writing, taken to SVP sign-off, and after that the operational layer, pipelines in Redshift and Python and the Tableau dashboards that made the definition something teams could use rather than something they had read.
The part I would count as the real outcome is upstream of all that. I worked with the Lead Routing team to improve their logic, so accounts would be identified correctly as revenue arrived rather than reconciled into the right place afterwards. Reporting that corrects a bad source is a treadmill. Fixing the source ends the problem.
The part that nearly sank it
I had scoped this as a Startup versus SMB problem. When I actually ran the linkage, every segment had Startup accounts on its books, so the scope, the stakeholder list, and the politics were all several times larger than the plan I had written. I found that out mid-flight.
Worse, my own org's logic was wrong too. The legacy routing was misidentifying AWS accounts that had received Startup credits as SMB accounts, because it applied neither the M&A logic nor the customer-identification logic. I could not credibly ask four other segments to hand over accounts while Startup's own identification was broken.
So I corrected Startup's gap first, publicly, before asking anyone else to move. That single sequencing choice is what made the methodology read as a rule rather than a land grab. Then I showed each segment its own data, account by account, tracing each one to its Salesforce customer. And we phased the comp impact: credit already earned stayed with whoever earned it, and only upcoming comp moved with the account. Nobody lost a quarter they had already worked, which left the discomfort forward-looking and survivable.
Results
Worth being precise about what the $82M is. It is revenue that already existed and was being managed by nobody, until the routing was fixed and it moved to the segment working those accounts. Not savings, not recovered revenue, not incremental revenue. Accounts were re-routed and compensation was adjusted as a consequence; I did not set comp policy.