Most B2B marketing teams can tell you exactly how many leads they generated last month. Far fewer can tell you how many of those leads a salesperson actually wanted to talk to, and almost nobody can say how many turned into revenue without opening three spreadsheets and arguing about definitions.
That gap is where budgets get wasted. A campaign that brings in 400 form fills looks great in a dashboard. If sales ignores 380 of them, it is a very expensive way to fill a CRM with noise.
This guide covers how to track lead quality in a way that survives contact with the sales team: which B2B lead generation metrics matter, how to build a scoring model people trust, and how to connect marketing data to closed deals.
Why lead volume misleads you
Cost per lead is easy to measure, so it becomes the default KPI. The problem is that cheap leads and good leads are rarely the same thing. Here is a simplified example of two channels over one quarter:
| Metric | Channel A: gated ebook ads | Channel B: demo-intent search |
|---|---|---|
| Spend | $10,000 | $9,000 |
| Leads | 200 | 60 |
| Cost per lead | $50 | $150 |
| Sales qualified leads | 10 | 18 |
| Closed-won deals | 2 | 5 |
| Cost per closed deal | $5,000 | $1,800 |
On a cost-per-lead report, Channel A wins by a mile. On cost per closed deal, Channel B is nearly three times more efficient. The numbers are illustrative, but the pattern shows up in real accounts constantly. If you only optimize for the top of the funnel, you will slowly shift budget toward the worst performers.
Marketing qualified leads vs sales qualified leads
Before you measure anything, marketing and sales need to agree on what the stages mean. Most disputes about lead quality are really disputes about definitions.
| Stage / Criteria | Marketing qualified lead (MQL) | Sales qualified lead (SQL) |
|---|---|---|
| Who decides | Marketing, based on score or behavior | Sales, after a conversation |
| Typical trigger | Score threshold reached, demo request, pricing page visits | Confirmed need, budget range, authority, timeline |
| What it tells you | This person looks like a fit and shows interest | This person is worth pursuing as an opportunity |
| Main risk | Threshold too low, so sales gets flooded | Reps are too strict or too slow to follow up |
Write the definitions down
Put a one-page document in front of both teams. For each stage, list the exact criteria, the owner, and the maximum time a lead can sit there before it moves or gets recycled. A rule like “MQLs must be contacted within one business day” prevents a lot of finger pointing later.
Track the handoff, not just the stages
Two numbers tell you more than almost anything else here: the MQL-to-SQL conversion rate and the share of MQLs that sales rejects. If more than a third of your MQLs are bounced back, treat that as a reason to audit, not as a verdict. The cause can be the scoring model, the targeting, slow follow-up, uneven SDR quality, or an ideal customer profile that has shifted since the model was built. One third is a rough trigger for a closer look, not an industry benchmark.
B2B lead generation metrics worth tracking
The list of possible metrics is endless. These are the ones that consistently earn their place on a dashboard.
One note on the numbers in this guide. Principles such as shared definitions, a closed loop between your CRM and ad platforms, and judging channels by revenue instead of volume apply to almost any B2B business. Specific thresholds, like a one-day response time, a score cutoff or a rejection rate that triggers an audit, are starting points. They depend on your deal size, sales cycle and team capacity, so adjust them with your own data. The same goes for the lead quality metrics below.
| Metric | What it shows | How to calculate | Watch out for |
|---|---|---|---|
| MQL-to-SQL rate | Whether marketing’s definition of good matches sales’ | SQLs ÷ MQLs in the same cohort | Mixing cohorts from different months |
| Cost per SQL | Channel quality, months before deals close | Marketing spend ÷ SQLs from that spend | Sales-sourced SQLs mixed into the count |
| SQL-to-won rate | Real lead quality at the bottom of the funnel | Closed-won ÷ SQLs | Small samples, long sales cycles |
| Pipeline generated per channel | How much potential revenue each source creates | Sum of opportunity value by original source | Opportunity values inflated at creation |
| Revenue per lead | The average worth of a lead from a given source | Closed-won revenue ÷ leads from the same cohort | Cohorts too young to have closed yet |
| Cost per acquisition (CPA) | What a paying customer costs you | Total spend ÷ new customers | Ignoring sales salaries and tools |
| Pipeline velocity | How fast leads turn into revenue | (Opportunities × deal size × win rate) ÷ sales cycle length | Averages hiding one huge deal |
| Time to first contact | Whether you waste intent | Timestamp of first touch minus lead creation | Auto-replies counted as contact |
| Customer lifetime value (LTV) to CPA | Whether the channel is profitable | LTV ÷ CPA | Using revenue instead of margin |
| Marketing ROI by channel | Where to move budget | (Attributed revenue − cost) ÷ cost | Attribution gaps |
Cost per SQL deserves a place on the dashboard next to CPA. When a sales cycle runs for months, closed-won data arrives too late to steer this quarter’s budget, while SQLs show up within weeks. Cost per SQL shows which channels produce qualified conversations before revenue confirms it. Once you have enough closed deals, add the lead-to-customer conversion rate by source, and always compare cohorts rather than calendar months.
Top of the funnel
Volume and cost still matter, but treat them as context. Look at lead-to-MQL rate by source and the share of leads that match your ideal customer profile (company size, industry, role). A source that delivers 80% on-profile leads is worth more than one that delivers 20%, even at triple the price.
Middle of the funnel
This is where MQL-to-SQL rate, speed to contact, and rejection reasons live. Rejection reasons are underrated. Add a required dropdown in the CRM (wrong company size, no budget, student, competitor, bad contact data) and review it monthly. In a few weeks it will tell you more about your targeting than any campaign report.
Bottom of the funnel
SQL-to-won rate, sales cycle length, average deal size, and CPA. These are the numbers your finance team will care about, so these are the ones that give marketing credibility.
Lead scoring for B2B marketing that sales will actually use
Lead scoring fails when it is built in a marketing meeting and handed to sales as a finished product. It works when it is built from closed deals.
Score two things separately
Mixing fit and interest into one number hides useful information. A student who downloads ten ebooks and a VP at a perfect-fit company who visits the pricing page once should not get the same treatment.
- Fit score (who they are): industry, company size, job title, region, tech stack, budget signals.
- Engagement score (what they do): pricing page visits, demo requests, webinar attendance, repeated visits, email replies.
A lead needs to clear a threshold on both before it becomes an MQL. A simple grid of A, B, and C for fit and 1, 2, and 3 for engagement is enough to start.
Use negative scoring
Points should go down as well as up. Subtract for personal email domains, competitor domains, job titles like intern or student, unsubscribes, and long inactivity. Without this, scores only ever inflate and the threshold becomes meaningless.
What a scoring model looks like in practice
The table below is an illustrative model, not a benchmark. Your own point values should come from the closed-won analysis described next.
| Group | Attribute | Points |
|---|---|---|
| Fit | Company size within your ICP | +15 |
| Fit | Decision-maker title | +20 |
| Fit | Target industry | +10 |
| Fit | Personal email address | −10 |
| Fit | Student or intern title | −20 |
| Engagement | Demo request | +30 |
| Engagement | Pricing page visit | +10 |
| Engagement | Replied to a sales or marketing email | +15 |
| Engagement | No activity for 60 days | −15 |
Set the MQL threshold on each score separately. In this example a lead becomes an MQL at a fit score of 25 or more and an engagement score of 20 or more. Two quick checks show how it behaves:
- A VP at a company of the right size (fit 35) who visits the pricing page and replies to an email (engagement 25) qualifies.
- A student who requests a demo (fit −20, engagement 30) does not, because fit is below the line. The request still deserves a polite reply, but it should not use up an SDR’s day.
The 60-day decay keeps old leads from sitting on a threshold they crossed months ago. Review the cutoffs every quarter against SQL acceptance: if sales rejects most MQLs, raise them; if good leads slip past, lower them.
Calibrate against closed-won deals
Pull your last 30 to 50 won deals and look at what they had in common before they became opportunities. Which pages did they visit? What titles were involved? How many touches did it take? Then do the same for lost or stalled deals. The differences are your scoring rules. If you have fewer than 20 wins, keep the model simple and revisit it every quarter.
Connect your CRM, ad platforms, and offline conversions
You cannot judge lead quality from data that stops at the form submission. Quality only becomes visible later, in the CRM, sometimes months later.
Set up CRM integration properly
The integration between your forms, marketing automation, and CRM should pass the original source, campaign, and first-touch and last-touch data into every lead record. Check that UTM parameters are captured in hidden fields and that they survive when someone fills out a second form a week later.
Send offline conversion tracking back to ad platforms
Google Ads and LinkedIn can import conversion events from your CRM. Instead of optimizing for a form fill, you tell the platform which leads became SQLs or paying customers. Over time the algorithm learns what a good lead looks like and stops chasing cheap ones. The steps are roughly the same on every platform:
- Store the click ID (gclid, li_fat_id, or equivalent) on the lead record when the form is submitted.
- Define the CRM stages that count as conversions, usually SQL and closed-won.
- Set up a scheduled upload or native integration that sends those stages with the click ID and, ideally, deal value.
- Switch campaign optimization to the new conversion event once the platform has enough stable, good-quality signals. The volume it needs depends on the platform, the conversion action and your sales cycle, so check each platform’s current documentation instead of assuming a fixed number. When closed-won deals are too rare to learn from, SQL is usually the better event to send.
Keep the data clean
Duplicates, missing sources, and manual stage changes with no timestamps will quietly ruin every report. Assign one person to own CRM hygiene and run a monthly audit of the fields you rely on.
Lead routing and sales funnel alignment
A well-scored lead that sits in a queue for three days is a wasted lead. Lead routing rules should be simple enough that everyone can explain them.
- Route by territory, segment, or account owner, whichever your sales team already uses.
- Send high-intent actions such as demo requests straight to a rep and alert them in real time.
- Set a fallback owner and an escalation rule if nobody responds within a fixed window.
- Review routing monthly, because team changes quietly break old rules.
On the alignment side, a short weekly or biweekly meeting between marketing and sales is worth more than any dashboard. Bring ten recent leads, look at what happened to each, and ask what the scoring model got wrong. Do this consistently and conversion rate optimization starts to happen almost on its own.
Attribution models: pick one and know its limits
No attribution model is correct. Each one answers a slightly different question.
- First touch: useful for seeing which channels start relationships, weak for long cycles.
- Last touch: easy to implement, tends to over-credit branded search and direct traffic.
- Linear or position-based: spreads credit across touches, better suited to multi-stakeholder deals.
- Self-reported attribution: a free-text field on the form asking how people heard about you. It is messy but often catches word of mouth, podcasts, and communities that tracking pixels miss.
A practical approach is to report on one model consistently and compare it against self-reported data every quarter. When the two disagree sharply, that is a signal to investigate, not to pick the more flattering one.
A 30-day plan to get started
- Week 1: Document MQL and SQL definitions with sales. Agree on a maximum response time.
- Week 2: Audit CRM fields, source tracking, and UTM capture. Add required rejection reasons.
- Week 3: Build a first version of fit and engagement scoring from your closed-won deals.
- Week 4: Build one report showing cost per SQL, cost per closed deal, and MQL-to-SQL rate by channel. Start offline conversion imports if you have the click IDs.
Do not wait for perfect data. A rough model that people use beats a sophisticated one that nobody trusts.
Common mistakes to avoid
- Reporting leads by month created instead of by cohort, which makes conversion rates look worse than they are.
- Changing scoring rules constantly, so trends become impossible to compare.
- Treating every form as equal. A contact request and an ebook download are different intents.
- Judging channels after 30 days when your sales cycle is 90.
- Ignoring sales feedback because it is anecdotal. Anecdotes from the people who talk to leads every day are still data.
Final thoughts
The point of all this is revenue from every dollar spent on acquisition. Lead count only matters when it leads there. If you do one thing this week, sit down with sales, agree on what an MQL and an SQL are, and write it down, because every other metric in this guide depends on it. Your first version of the model will be imperfect. After a couple of quarters of adjustments you will know which leads deserve budget and which ones only look good in a report.
