First, split the rate before you trust it
A single MQL-to-SQL percentage for the whole quarter hides almost everything worth knowing. Pull every lead that crossed the qualification line in a recent period long enough to have been worked, and follow that cohort forward: accepted, returned with a reason, qualified, or still untouched. Then cut the result three ways. By original source, because a search or demo-request source and a syndicated-content source behave differently. By score band, because a model that works shows a clear gap between the leads just above the line and those well above it. And by the reason recorded when sales returned a lead, because "student", "no project" and "never reached" are three different diagnoses.
Two checks come before any cause. Confirm the cohort is being compared with its own outcomes, not this month's leads against this month's opportunities, which came from earlier leads; the lead stages guide explains why that comparison misleads. And confirm that leads with no recorded outcome are counted separately. If a large share were never touched, the problem is the hand-over rather than qualification, and the companion page on sales not following up on leads is the better starting point. Where one platform's lead forms supply most of the returned leads, the form itself may be the fault, and the diagnoses for LinkedIn lead form quality and Meta lead ads quality cover those settings.
For the side-by-side definitions of the two stages and who decides each one, see the MQL versus SQL comparison.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| High-scoring leads include students, competitors and job seekers; return code is often "not a fit" | Which signals contribute most points to leads sales returned | Score rewarding activity rather than fit | Separate fit from intent, gate disqualifiers, weight buying actions |
| Qualified volume jumped in the same month a target rose; acceptance fell | The threshold's change history against the volume target | Threshold lowered to hit a number | Restore a threshold set from past won deals; report pipeline beside volume |
| Sales and marketing describe a qualified lead differently | Whether a written definition exists that both leaders signed | Definitions never agreed | Write the stage criteria together and configure them in the CRM |
| Demo and pricing requests convert well when called the same day, poorly after | Time to first contact by lead type, measured from CRM timestamps | Follow-up too slow | Route instantly, set a response target by lead type, escalate breaches |
| Most returned leads came from report or template downloads; reason is "just researching" | Share of qualified leads whose only intent signal is a download | Researchers handed over as buyers | Treat downloads as reading signals; hand over only on buying actions |
| Early leads go straight to sales or nowhere | What happens to a lead between capture and qualification | No nurture between first touch and hand-over | Build stage-based nurture with an observable exit event |
Fix, verify and prevent each cause
A score that rewards reading rather than buying
Most scoring models are built from whatever the platform tracks easily: page views, downloads, email clicks. Those signals measure curiosity, so a lead can collect enough points to qualify without belonging to the right kind of organisation or doing anything a buyer does.
Fix: rebuild the model along the lines of the lead scoring guide: keep fit and intent as separate scores, make disqualifiers such as competitor domains or out-of-region companies hard gates, cap repeated low-value actions and give demo, pricing and trial actions most of the weight.
Verify: leads above the new line are accepted and qualified at a visibly higher rate than leads below it, and the share of returns coded "not a fit" falls.
Prevent: calibrate the model every quarter against what actually closed, with sales in the room, and log every change with its reason.
A threshold lowered to meet a volume target
When marketing is judged on the count of qualified leads, the easiest lever is the line itself. A small cut to the threshold, or a new rule that qualifies a lead on a single event, lifts the count at once and passes the cost downstream.
Fix: find when and why the threshold moved, restore a line set from where previously won deals sat, and change the target that created the pressure, as the lead stages guide recommends, so that marketing reports pipeline and cost per opportunity beside the qualified count.
Verify: qualified volume falls, acceptance and opportunity rates recover within a cohort, and opportunity count holds or rises.
Prevent: treat the threshold as part of the agreement with sales; any change needs both teams, a date and a note in the report.
Definitions nobody signed
If marketing's definition lives in the automation platform and sales' definition lives in each rep's head, every qualified lead is judged against a standard it was never built to meet. The symptom is a long-running argument about lead quality that data never settles.
Fix: write the ideal customer profile, the qualification criteria, the acceptance window and the return reasons together, as the sales and marketing alignment guide sets out, and configure them as fields and stages rather than leaving them in a slide.
Verify: the reasons sales gives for returns map onto the written criteria, and both teams quote the same qualified count in the same meeting.
Prevent: review the definitions at a fixed quarterly session rather than after a dispute.
Follow-up that arrives after the moment has passed
A person who asked for a demo this morning is comparing vendors this afternoon. When the first call comes days later, the conversation finds a buyer who has moved on, and the lead is recorded as not qualified when it was really not reached in time.
Fix: follow the inbound lead handling playbook: route high-intent requests to a named person at once, set response targets by lead type, add a fallback queue for absent owners and escalate any lead untouched past its target.
Verify: median time to first contact for demo and pricing requests falls, and the qualification rate for those requests rises compared with earlier cohorts.
Prevent: put time to first contact on the same weekly report as lead volume, measured from timestamps rather than from what reps remember logging.
Researchers handed to sales as if they were buyers
Gated reports, templates and webinars attract people who are learning about a problem, not choosing a supplier. Kaushik's See, Think, Do, Care model describes this as audiences at different stages of intent, each needing different content and different measures, and research published through the B2B Institute stresses how few business buyers are in market at any moment. Handing a reader to sales as a buyer produces a polite conversation and a return.
Fix: stop counting downloads as qualification on their own, and follow the distinction drawn in lead generation versus demand generation: capture intent on demo, pricing, trial and contact paths, and treat content forms as reading signals.
Verify: the share of returns coded "researching" or "no project" falls, and download-sourced leads that do reach sales come with a buying action.
Prevent: label every form in the automation platform as a reading or a buying form, and let only buying forms trigger the hand-over directly.
No nurture for leads that are genuinely early
Some returned leads are the right organisation at the wrong time. Without a structured route back, they are either discarded or left in a rep's list, and the next time they are ready they may already be talking to a competitor.
Fix: build stage-based nurtureKeeping in useful contact with leads who are not yet ready to buy until they are. as the nurture sequences guide describes, with an entry rule tied to the return reason, content that answers the lead's current question and an exit event, such as a pricing visit or a demo request, that brings them back to sales; a nurture sequence plan writes those rules down before the workflow is built.
Verify: a share of recycled leads re-qualify and convert, and leads that went through nurture qualify at a higher rate than comparable leads that did not.
Prevent: make "recycle to nurture with a reason" one of the outcomes sales can choose, so that early leads are never simply closed.
A prevention routine for the qualification stage
Most qualification problems build slowly and are noticed only when a quarter's pipeline misses. A short rhythm keeps them visible. Weekly, read the leads sales returned with their reason codes and the time each waited for a first contact; a sudden rise in one reason, or in one source, is the earliest warning. Monthly, read MQL-to-SQL by source and by score band for the cohort that has had time to be worked, and check that the inputs feeding the score, such as enrichment fields and pricing-page tracking, still populate. Quarterly, recalibrate the score against won deals, revisit the threshold and the definitions with sales, and read the win rateShare of closed sales opportunities that were won. of marketing-sourced opportunities so that the downstream result is part of the same conversation. Keep one written definition of the marketing-qualified leadA lead that marketing judges ready for sales follow-up against agreed criteria., held where both teams can see it, and treat every change to it as a reporting event with a date. Before any new form, source or content programme goes live, decide whether it produces reading or buying signals and route it accordingly. When the rate does fall, start with the newest change: a new form, a new syndication partner, a changed field mapping or a new target explains most drops.
A procurement software vendor reads its rejected leads
A procurement software vendor produced 640 marketing-qualified leads in a quarter, and sales qualified 38 of them, about 6%. Splitting the cohort showed that 410 came from a gated benchmark report and a content syndication partner, and only 9 of those qualified; the 120 demo and pricing requests qualified at 21%, but their median first contact was 31 hours. The score gave 15 points per download and the threshold had been cut from 60 to 45 when the quarterly target rose. The team removed download points, restored the threshold at 60 with a fit gate for companies under 100 staff, routed demo requests to a named rep within one working hour, and sent download leads into a six-step nurture. The next quarter produced 260 qualified leads; sales qualified 52, about 20%, and opportunities rose from 17 to 26.
If the threshold itself is in question, the lead scoring threshold test checks where it should sit by sending a random half of the leads near it to sales.
Common mistakes
Blending every source into one rate. Raising the threshold without changing what the score counts. Moving the line quietly to hit a target. Treating a returned lead as wrong when it was never called. Counting a template download as a buying signal. Closing early leads instead of recycling them. Tuning platform lead forms and the scoring model in the same month, so neither change can be read.
