Case study · 02 - Sales Tech

Forecast & Pipeline Transformation

$68M ARR sales-tech company · 55 quota carriers · $48k mid-market ACV

A sales org that had missed forecast four of the last six quarters replaced deal-by-deal interrogation with evidence-graded commits - and got the number to hold.

Company profile

Revenue: $68M ARR
Employees: 310
Sales org: 55 quota carriers · 5 pods
ACV: $48k mid-market · $210k enterprise
Cadence: Weekly commit calls, quarterly board forecast
Forecast record: Missed 4 of last 6 quarters

Stack:

Salesforce

Clari

Gong

Outreach

Slack

Snowflake

Plus marketing attribution and the CPQ system - 9 sources connected in total.

The challenge

The company had missed its forecast in four of the last six quarters - twice high, twice low - with swings up to 19% of the quarter’s number. The board had stopped trusting the call, so every quarter ended with a scramble and every plan started with a buffer.

31%

of committed deals had no recorded customer activity in the prior 14 days. The commit was a feeling, not a fact.

6 hrs/wk

of leadership time went to commit calls - most of it interrogating reps one deal at a time to reconstruct evidence that already existed in Gong and the CRM.

different working definitions of “commit” across five pod leaders. Clari faithfully rolled up whatever each pod fed it.

44%

of next-step fields hadn’t been touched in three weeks or more. Free-text hygiene meant the system of record recorded very little.

The forecast wasn’t wrong because the math was bad. It was wrong because the inputs were vibes.

Residency, week one

A Revllama forward-deployed engineer sat through a full commit-call cycle across all five pods, reconciled two quarters of committed deals against their Gong evidence, and rebuilt the slip history deal by deal.

1
Slippage was the single biggest source of variance: reps re-committed slipped deals an average of 2.3 times before the deals closed or died, and every re-commit re-entered the forecast at full weight.
2
Gong already showed the truth - single-threaded deals, no paper process, champions gone quiet - but nobody reconciled call evidence against CRM stage. The two systems told different stories about the same deals.
3
Commits with a multithreaded buying committee and a dated paper-process step closed at roughly 3x the rate of the rest of the commit sheet. Neither signal existed as a field anywhere.
4
The historical slip pattern was rep-specific and stable - the same sellers slipped the same way every quarter - which made it predictable.

What we deployed

The engagement codified 86 pipeline and forecast rules - stage definitions, commit criteria, evidence requirements, slip predictors - and put three agents on top of them.

Forecast Integrity Agent

Audits every committed deal weekly against the evidence: activity recency, buying-committee coverage, paper-process stage, and the rep’s historical slip pattern. Flags happy-ears commits and sandbagged upside with the receipts attached, 48 hours before the call.

Deal Risk Agent

Watches every open opportunity for champion silence, stalled stages, and competitor mentions in calls - so risk surfaces when it forms, not when the forecast breaks.

Pipeline Hygiene Agent

Custom-built for this team: chases stale next-steps and close dates with reps directly in Slack, accepts the fix inline, and escalates only what stays stale. Hygiene stopped being a manager nag and became a background process.

sources connected

rules running

~450

open deals audited weekly

The CRO interrogates the forecast from Claude through the org’s MCP hub - “show me every commit with no activity in two weeks” - instead of waiting for Monday’s call. Same gateway the agents use, same audit trail.

The rollout

Week 1

Connect, codify, shadow

Sources wired, 86 rules codified from pod leaders' own (conflicting) definitions - reconciled in one working session with the CRO. Agent scoring the live commit sheet in shadow mode by day 5.

Week 2

Tune & go live

Shadow scores reviewed against week-over-week outcomes with the CRO; thresholds tuned. Live to pod leaders - production in two weeks.

Month 2

Async commit calls

Flags delivered to reps 48 hours before the call with evidence attached. Deal-by-deal interrogation moved async; the call shrank to exceptions and strategy.

Quarter end

Agent call becomes primary

The agent’s adjusted call landed within 3.2% of actuals, against an 11% average miss over the prior six quarters. Its number became the primary forecast; Clari kept for validation.

Results

27%

forecast variance cut

in the first full quarter of production

±3.2%

final-call accuracy

by quarter two, from an 11% average miss

6h → 2h

weekly commit-call time

deal interrogation moved async with evidence attached

-41%

multi-quarter slips

re-committed deals confronted with their own history

Value delivered

$2.1M-$3.4M year-one value

$1.3M-$2.2M

Recovered at-risk commits

Flagged deals re-worked before quarter end, measured against the historical outcomes of comparable unflagged slip cohorts. Discounted to 60% for attribution.

~4,900 hrs/yr

Leadership & rep time returned

Commit-call and hygiene-chasing time across 55 sellers and 8 leaders, valued at loaded cost.

Not quantified

Planning on a trusted number

Capacity and cash planning against a forecast the board trusts. Real, and deliberately left out of the total.

All estimates discounted for year-one realization. No credit taken for second-order effects.

Beyond the revenue org

The evidence-graded pipeline now feeds more than the forecast:

Finance uses the same graded pipeline for cash planning instead of applying a flat haircut to the CRM number.
Product marketing gets a weekly digest of competitor mentions pulled from flagged deals' calls.
Neither effect is included in the value figures above.

Company details anonymized. Figures are representative of the engagement pattern shown, pending customer-cleared publication.