Process Mining on the Salesforce Sales Funnel with Fluxicon Disco

Business problem

Sales leadership was convinced the opportunity flow followed the textbook: Qualification → Needs Analysis → Proposal → Negotiation → Closed. In reality, only 1 in 6 deals moved through that happy path. Deals jumped backward, skipped validation, stalled between approvers — and the forecast debate every quarter was really a debate about a process nobody had ever actually looked at.

Data available

• Salesforce OpportunityHistory: every stage change with timestamp + user
• OpportunityFieldHistory on Amount and CloseDate
• ~640k events across 18 months of opportunities
• User object: role, region, line manager — for resource-level analysis
• Account segmentation (strategic / key / standard)

Solution implemented

1. 1Pulled a clean event log from Salesforce via SOQL + Python (pm4py): case id = OpportunityId, activity = new stage, resource = user. Reproducible as a monthly refresh.
2. 2Loaded the log into Fluxicon Disco — 180+ unique paths collapsed into 6 dominant patterns covering 84% of deals.
3. 3Used Disco's bottleneck view to locate the real waiting points. The slowest step was not 'Proposal → Negotiation' as everyone assumed — it was 'Qualification → Proposal' (deals waiting on a central pricing team).
4. 4Quantified rework loops: deals bouncing between Proposal and Qualification more than once = 22% of cycle-time overrun. Built a Power BI monitor on the same KPIs so findings survive the presentation.

Impact

What I learned

You cannot improve a sales process you have never actually looked at. Disco on top of OpportunityHistory takes days, not quarters — and changes the conversation from opinion to evidence.