Series B SaaS: How a Rebuilt Lead Scoring Model Cut Sales Cycle 34% and Grew SQL-to-Won by 2.4x
Client identity protected under NDA. Details available under mutual sign-off in a discovery call.
Composite Company Profile
A Series B B2B SaaS company operating in the sales operations vertical. Roughly $18M ARR at engagement start, 85 employees, ACV of $42,000, and a blended new-logo plus expansion motion. Ideal customer profile centered on mid-market RevOps and Sales leaders at 200 to 2,000 employee companies. Sales cycle averaged 91 days. The go-to-market team included six AEs, four SDRs, a demand gen lead, and a marketing ops manager.
The Problem
Volume was not the issue. The company generated roughly 1,400 marketing-qualified leads per month across paid, organic, webinars, and G2 intent. The problem was that AEs were burning cycles on leads that never had a real shot at closing. SDR-to-SQL conversion sat at 22%, but SQL-to-Closed Won had collapsed to 7.1% over the prior two quarters. Pipeline was inflated, forecasts were missing, and the CRO was under board pressure to explain why doubling MQL volume in the prior year had produced only a 12% lift in net new ARR.
Two prior attempts to fix scoring had failed. The first was a HubSpot-native rules-based model that scored leads on demographic fit and page views. It flagged nearly every demo request as A-tier, which meant the score carried no signal. The second attempt used an off-the-shelf predictive scoring product bolted onto the existing CRM. It produced scores but no explanation, and AEs stopped trusting the output within six weeks. Both efforts left the scoring field in Salesforce, but the sales team had learned to ignore it.
What Our Team Diagnosed
The root cause was not the scoring algorithm. It was the data model underneath it. Our analytics team pulled 26 months of closed-won and closed-lost history from Salesforce and reconciled it against product usage events, marketing touch history, and firmographic enrichment. Three findings reframed the problem.
First, the strongest predictor of Closed Won was not any inbound signal at all. It was the presence of two or more contacts from the same account engaging within a 14-day window. Single-contact opportunities closed at 3.2%. Multi-contact opportunities closed at 19.4%. The rules-based model had no account-level logic.
Second, the SDR team was routing leads based on lead score, but the score was heavily weighted toward job title strings that had drifted. Titles like “RevOps Manager” and “Sales Ops Lead” were scoring lower than “VP Sales” even though the former closed at nearly 3x the rate for this ACV band.
Third, the predictive product was training on a lookback window that included the 2023 buyer behavior, when free trials converted at very different rates than in the current market. Its highest-confidence signals were stale.
Strategy MV3 Shipped
We proposed and executed a rebuild across three fronts under the Growth AI retainer, with Vance overseeing scope and our RevOps and analytics team executing. Services engaged: RevOps automation, lead scoring model build, and demand gen attribution rework.
The rebuilt model was account-based rather than lead-based. Every scored record rolled up to an account and inherited signals from every contact on that account. The model used three signal families, each weighted from the reconciled historical data:
- Fit signals: ICP firmographic match, tech stack match via Clearbit and BuiltWith, and title-band scoring recalibrated on the last 8 quarters of closed-won data.
- Behavior signals: multi-contact engagement window, high-intent page visits (pricing, integrations, security), G2 category views, and email reply sentiment scored by an LLM classifier.
- Timing signals: role change events on target accounts (via LinkedIn Sales Nav + Cognism), funding round events, and expansion signals from existing customer accounts.
Scores were expressed as tiered bands (A1, A2, B, C, D) with a plain-English explanation string attached to each record so the AE knew why an account was flagged. We refused to ship a black-box score again.
Implementation
Total build window was 11 weeks from kickoff to go-live, followed by 90 days of tuning.
- Historical data reconciliation across Salesforce, HubSpot, Segment, and product events (weeks 1 to 3).
- Model specification, weight calibration, and back-testing against the trailing 6 quarters (weeks 3 to 6). Back-test showed the new model would have prioritized the top 15% of accounts that produced 71% of closed-won revenue.
- Salesforce and HubSpot integration, custom account score object, routing rules, and SDR playbook rewrite (weeks 6 to 9).
- SDR and AE enablement, ride-alongs, and dashboard build in Salesforce and Amplitude (weeks 9 to 11).
- Post-launch tuning cadence: weekly model review for the first 8 weeks, then monthly.
Deliverables included the scoring model documentation, a routing rules matrix, a rebuilt SDR cadence in Outreach with tier-specific messaging, an AE-facing dashboard, and a monthly attribution report ties dollars to signals.
Outcomes
Measured against the trailing 6 months prior to launch, results at the 6-month post-launch mark:
- SQL-to-Closed Won conversion: 7.1% to 17.2% (2.4x lift).
- Average sales cycle: 91 days to 60 days (34% reduction).
- Pipeline coverage ratio held at 3.4x while AE headcount stayed flat, so quota attainment rose without hiring.
- AE-reported “lead quality satisfaction” in quarterly survey: 4.1 out of 10 to 8.6 out of 10.
- CAC payback period: 19 months to 13 months.
- Net new ARR in the two quarters following launch grew 41% versus the two quarters prior, on flat marketing spend.
Timeline
Kickoff to go-live: 11 weeks. First outcome signal (SDR-to-SQL lift): week 14. Full attribution reset and reported outcomes: 6 months post-launch. Ongoing model maintenance is included in the Growth AI retainer.
What the Buyer Said
“We had two scoring rebuilds fail before this one. The difference was that MV3 spent the first three weeks in our data, not in a slide deck. When the model went live, it explained itself, and the AEs actually trusted it inside a month.” — Priya, VP Revenue Operations
NDA Framing
Client identity, exact ARR, and internal system names are protected under NDA. A full case walkthrough including the model spec, sample scoring logic, and 6-month tuning notes is available under mutual sign-off in a discovery call.
Book a Discovery Call
If you run a Series A through C SaaS company and your AEs have stopped trusting the lead score, we can rebuild the model on your closed-won history and get it live in under 12 weeks. Book a discovery call or see the ABM Agency and Demand Gen service pages for related work.
Case study drafted by the MV3 Marketing team. Vance Moore, CEO, oversees engagement scoping and QA. Our RevOps, analytics, and demand gen team executes the build.
