At $500K ARR, Pipedrive's weighted pipeline forecast feels solid. Your sales team uses the standard five stages. Deal value + win probability = a forecast number you can actually trust. You hit your number three quarters in a row. Then you cross $1M ARR. Your team adds custom stages because the default ones don't fit your sales cycle. Someone creates a 'Proposal—Awaiting Finance' stage. Another rep skips directly from Lead to Negotiation. A deal appears in two pipelines because it's listed as a shared deal. By $2M ARR, your forecast is noise. You're 30% over. Then 15% under. Then you miss by $400K because three deals you counted as 'won' were actually moving between custom stages and never advancing. This isn't operator error. This is how Pipedrive calculates weighted pipeline at scale—and where it systematically breaks. How Pipedrive's weighted forecast works (and where it fails) Pipedrive multiplies deal value by a probability percentage at each stage. A $100K deal in your 'Proposal' stage at 60% probability = $60K forecast. That sounds bulletproof. Until your second sales manager adds a 'Proposal—Legal Review' stage because contracts need internal approval before they're truly 'sent.' Now the same deal exists in two stages simultaneously—once as 'Proposal' at 60% and once as 'Proposal—Legal Review' at 65%. Your forecast just counted $65K twice. Or a deal sits in 'Negotiation' for six months because the prospect won't close but also won't reject you. You set it to 40% probability to account for that limbo. But it never moves. By month four, you've counted that $50K deal 120 times across your rolling monthly forecasts. It's mathematically real in the system. It's dead in reality. The core problem: Pipedrive weights stages, not velocity. If a deal doesn't move between stages for 90 days, Pipedrive still counts it at that stage's probability. It has no concept of 'stalled.' A deal that's been in Proposal for four months and a deal that entered Proposal last week have the same forecast weight if they're both at the same stage and same value. Audit #1: The duplicate-deal trap at $1.5M ARR A B2B SaaS team (Series A, ~$1.5M ARR) was running at 35% forecast variance. Some months they'd hit their number cold. Other months they'd be $200K over, then $150K under. I pulled their CRM export and found 47 deals flagged as 'shared' between two or more team members. In Pipedrive's default reporting, shared deals are counted at full value for each pipeline view—meaning if you're running a sales forecast by rep, the same $75K deal shows up in both Sarah's forecast and Marcus's forecast, totaling $150K when only one deal exists. Their CRM showed $2.1M in active deals forecast. The actual unique deal count was $1.4M. The difference: shared deals counted twice, plus three deals that existed in two separate stages (a prospect had asked to loop in their CFO, so the rep cloned the deal into a 'Awaiting CFO Sign-off' stage and left the original in Negotiation). When we isolated to non-shared, non-duplicate deals only, their forecast accuracy jumped from 65% to 82% within a month—not because anything changed operationally, but because they stopped counting the same revenue twice. Audit #2: The custom-stage velocity collapse at $1.8M ARR An IT services firm with eight sales reps had 12 active pipeline stages by year two. The original five (Lead, Qualified, Proposal, Negotiation, Closed) multiplied into: Lead → Qualified (lead qualification) Qualified → Discovery (needs analysis) Discovery → Proposal (proposal requested) Proposal → Proposal—Client Review (waiting on client) Proposal—Client Review → Proposal—Legal Review (internal approval) Proposal—Legal Review → Negotiation (final terms) Negotiation → Negotiation—Finance Hold (budget cycle pause) Negotiation—Finance Hold → Negotiation (resuming) Negotiation → Closed Won The problem: they were setting probability by stage, but the stages weren't sequential. A deal could move from Proposal straight to Negotiation—Legal Review, skipping the 'Client Review' stage entirely. Or a deal would sit in 'Negotiation—Finance Hold' for three months, then drop back to Negotiation when the client's budget year restarted. Pipedrive's probability assignments didn't account for this non-linearity. A deal stuck in 'Negotiation—Finance Hold' at 55% probability was still being counted at 55% probability in the 30-day forecast, even though the rep hadn't touched it in 60 days. When we analyzed their forecast history over six months, deals that had been in 'Negotiation—Finance Hold' for more than 45 days had a 12% close rate. Deals in active Negotiation had a 58% close rate. But the forecast weights treated them identically if they were both at 55%. The team was systematically overforecasting deals stuck in budget-hold limbo because the stage didn't reflect actual momentum. The fix was brutal: consolidate back to six stages, and add a deal age modifier that automatically dropped probability for any deal th