MQL Quality Crisis in a Cybersecurity Lifecycle Model
Sentinel Ridge sells cybersecurity software to mid-market companies. Its lifecycle model has six stages: Subscriber, Lead, MQL, SQL, Opportunity, Customer. A lead becomes MQL at a behavior score of 100 or a demographic score of 80. Behavior points never expire: webinar attendance is +25, whitepaper download +10, pricing page +15, and email click +2. Demographic points are added for job title, company size, and industry, but the form field 'Company Size' is free text. Sales receives about 150 MQLs per month, converts 18% to SQL, and rejects 40% as 'not ready'. The average time from MQL to first sales touch is 3.2 days. Reps say they get 'students and competitors'. Marketing says sales ignores good leads. There is no recycle path; rejected leads stay in MQL. The CRO wants a 90-day plan to improve MQL quality without reducing volume.
1. Diagnose why sales rejects 40% of MQLs and why rejected leads remain stuck in MQL. Which design choices in the scoring and lifecycle model are driving the problem?
Read the worked answer
The MQL rule is an OR, so fit can be bypassed. A behavior score of 100 qualifies someone even if demographic fit is poor; a demographic score of 80 qualifies someone with no intent. That is how students and competitors enter. The behavior score never decays, so a webinar from a year ago plus routine email clicks can eventually cross 100. The point values are also misweighted: a webinar is +25, a pricing page is +15, and an email click is +2. High-volume low-intent activity can accumulate. Free-text company size means the demographic score cannot reliably add firmographic points, so the model overweights title and industry or misses size entirely. There is no negative scoring for competitor domains, student domains, or unsubscribes, and no recycle path.
Lifecycle design. Rejected leads stay in MQL because there is no closed-loop reason code and no recycle campaign. Sales rejection is not fed back into scoring, so the same profile keeps qualifying. The 3.2-day average first touch is not catastrophic, but it is slow for inbound MQLs and may contribute to rejection. The model also lacks a separate 'Sales Accepted' stage, so marketing cannot distinguish a rejected MQL from one that was never worked. The fix is not one threshold change; it is separating fit from intent, adding decay and negatives, and making rejection operational.
2. Propose a 90-day redesign of scoring and lifecycle. Keep MQL volume from collapsing, but improve quality. Include owners and a pilot.
Read the worked answer
Days 1-30: data and definitions. Marketing Ops cleans company size into a picklist, normalizes industry, and appends firmographics where possible. Sales and marketing agree on an ICP by segment: size, industry, title, and technology fit. Define fit score and intent score separately. Add negative points for competitor and student domains, free email domains for enterprise, unsubscribes, and closed-lost. Add behavior decay after 90 days. Replace the OR rule with a two-dimensional rule: fit at or above 70 and intent at or above 50, or fit at or above 40 and intent at or above 80 for a high-intent exception. Backtest against 6-12 months of closed-won and rejected MQLs before launch.
Days 31-60: pilot and feedback. Pilot in one region or segment. Add a Sales Accepted stage and mandatory rejection reason codes. Build a recycle smart campaign that moves rejected leads to nurture with a reason and a re-qualification date. Set a first-touch SLA of four business hours for inbound MQLs. Review 50 rejected and 50 accepted MQLs weekly with sales.
Days 61-90: rollout and volume protection. Roll out globally, monitor MQL volume weekly, and tune thresholds. To avoid a volume collapse, increase top-of-funnel acquisition and use the high-intent exception rather than lowering fit standards. Keep a control group of the old model for comparison if operationally possible. Owners: Marketing Ops for data and scoring, Sales Ops for stages and reason codes, Demand Gen for volume, and a joint revenue council for threshold sign-off.
3. How would you measure whether the redesign worked, without letting volume reduction be hidden as quality improvement?
Read the worked answer
Measure the chain, not one number. Track MQL volume, MQL-to-SQL conversion, sales rejection rate, Sales Accepted rate, time to first touch, pipeline per MQL, and closed-won rate by cohort. The current baseline is 150 MQLs per month, 18% MQL-to-SQL, and 40% rejection. Set targets such as rejection down to 25%, MQL-to-SQL up to 22%, and MQL volume within 10% of baseline. If volume falls more than 10%, investigate top-of-funnel and the high-intent exception before declaring success.
Use cohort and holdout logic. Compare cohorts by MQL month, not just month-over-month, because enterprise cycles are long. If possible, hold back a small control group scored under the old model to estimate incremental effect. Track rejection reasons to see whether the mix changes from 'not ready' to 'no budget' or 'bad fit'; that tells you whether scoring or targeting is the problem.
Guard against gaming. Do not let sales reject fewer leads just to hit a metric; audit a sample. Do not let marketing add manual MQLs to protect volume. Report quality-adjusted pipeline: MQLs times conversion times average opportunity value. A redesign that cuts volume 30% but raises conversion 5 points may still be worse for the business. The board should see volume, conversion, and pipeline together, with the assumptions and the control comparison.