B2B Marketing Funnels Are a Reporting Fiction — Here Is the Signal Model That Replaces Them

Pull up your funnel report right now. Look at the stage counts. Then answer one question: which specific human on that report deserves a phone call in the next 48 hours?

You cannot answer it. Not from that report. B2B marketing funnels tell you how many. They never tell you who. And “how many” has never once booked a meeting.

The funnel is not a picture of how companies buy. It is a picture of how your marketing team files people after the fact — a tidy set of buckets built for a monthly slide, not for a salesperson holding a phone at 9:15 on a Tuesday morning.

The Funnel Is a Filing System Wearing a Forecast Costume

Here is the uncomfortable part. The linear funnel does not just fail to help — it actively misrepresents what your buyers are doing.

Gartner’s research on the B2B buying journey found that buying is not a sequence of stages at all. It is a set of six “buying jobs” that buyers loop through in whatever order the situation demands. In Gartner’s 2024 buying research, buyers revisited each job an average of 1.7 times inside a single purchase, and 90% of buyers looped back to at least one job before deciding.

Read that again with your CRM open. Ninety percent of your prospects go backward at some point. Your b2b marketing funnels only allow forward motion. So every time a real buyer behaves like a real buyer, your system records it as noise — or worse, as stage progression that never happened.

Meanwhile Gartner also found buyers spend roughly 17% of their total buying time meeting with all potential suppliers combined. Split that across three or four vendors and your rep gets somewhere around 5% of the decision. The other 95% happens on your website, in your emails, in Slack threads you will never see, and in a spreadsheet somebody’s finance lead built without telling you.

The funnel report covers the 5%. The signals cover the 95%.

The Lead Was Never the Unit of Measure

The second structural problem: a funnel counts individuals, and individuals do not buy anything.

Forrester’s research on buying groups puts the average B2B purchase at roughly 13 people. Their position is blunt — decades of measuring marketing on leads and MQLs is now hurting performance, because one engaged person tells you someone is curious, not that an account is moving.

The number that should make you rebuild something: in Forrester’s account of Palo Alto Networks shifting from MQLs to buying groups, opportunities with multiple people attached were eight times more likely to advance than single-contact opportunities, and the shift produced a 17% higher closed-won rate.

Eight times. Not eight percent. That is not a reporting nuance — that is the difference between a pipeline that closes and a pipeline that decorates a dashboard. And nothing in a standard funnel stage tells your rep whether one person or four are engaged inside an account.

The Hypothesis: Stop Reporting Stages, Start Detecting Motion

Our thesis is this: the funnel is a measurement layer pretending to be a targeting layer. Companies do not need to fix their funnel stages. They need to build a second system underneath the funnel whose only job is to answer one question — who moved, how much, and how recently?

We call that engagement compression. Thousands of behavioral events across email, site, CRM, ads and forms get compressed into a short, ranked, human action list. Not a stage. Not a count. A name, a reason, and a time.

The Architecture We Would Build Underneath Your Funnel

Here is the architecture we would build — not a claim about what it did for anyone, but the wiring itself:

ICP DEFINITION
  ↓
TAM LIST BUILD (Prospect Pump™)
  ↓
WARM OUTBOUND EMAIL + CONTENT DISTRIBUTION
  ↓
BEHAVIOR CAPTURE (opens, clicks, pages, returns, forms, ads)
  ↓
ENGAGEMENT SCORING — weighted, decaying, account-rolled
  ↓
THRESHOLD BREACH → HOT LEAD ALERT (Easy Button™)
  ↓
HUMAN CALL — with the reason attached

Three design rules make this work, and most implementations break on all three.

1. Weight by cost, not by convenience. An email open costs a prospect two seconds. A pricing page visit costs them a decision to look. A return visit three weeks later costs them remembering you existed. Score them in that order. Most scoring models weight what is easiest to capture, which is why they surface openers instead of buyers.

2. Decay everything. A score that only goes up eventually ranks your oldest contacts highest. Behavior from 60 days ago is history; behavior from 6 days ago is intent. If your lead scoring system has no time decay, it is a loyalty program, not a buying signal.

3. Roll scores to the account. Given Forrester’s 8x finding, the single highest-leverage change most teams can make is counting distinct engaged humans per account as its own score input. One person at 80 is interesting. Three people at 45 is a buying group forming.

The plumbing is unglamorous and it is the whole job: CRM automation to move the data, content and distribution to generate something worth reacting to, and TAM mining and outbound to make sure the list is big enough that signal actually shows up.

The Economics: Why This Beats Buying More Leads

Illustrative example — run your own numbers. These are not results. They are arithmetic, and the point is the shape of the curve, not the digits.

Say a rep makes 40 calls a week from a funnel-stage list. Assume a 4% conversation-to-meeting rate on effectively random ordering — 1.6 meetings a week.

Now assume scoring correctly identifies the top 15% of that same database by recent engagement, and that calling into recent, specific behavior triples the connect-to-meeting rate to 12%. Same 40 calls, same list, same rep: 4.8 meetings a week. Three times the output with zero additional lead spend.

Compare that to the alternative most teams choose. Tripling meetings by tripling volume means buying three times the leads and hiring roughly two more reps — call it a five-figure monthly increase. The scoring layer is a build cost and a software line item. That is the entire argument for prioritization over acquisition: you are not short on prospects, you are short on order.

There is a supporting reason this works beyond call efficiency. Gartner found that customers who got information from suppliers they perceived as genuinely helpful were 2.8x more likely to experience high purchase ease and 3x more likely to close a bigger deal with less regret. You cannot be helpful at the right moment if you do not know what moment they are in. The signal tells you.

Where This Gets Uncomfortable

Two honest failure modes, and we would rather you hear them from us.

First: if you build this and your database produces almost no signal, the scoring layer did not fail — it diagnosed you. Engagement compression is a prioritization technology, not a demand-creation technology. If 6,000 contacts generate eleven meaningful behavioral events a month, you do not have a routing problem. You have a top-of-funnel and content problem, and you should fix that first with consistent content and distribution before you spend a dollar on scoring logic. We would rather tell you that in week one than sell you a Watchtower™ build that watches an empty street.

Second: the buying-group model is genuinely harder to operate than the funnel. The critics are right about this. Identifying who is actually in a buying group requires contact data that decays fast, and a group profile that surfaces three of twelve stakeholders is just a contact list with a fancier name. Most marketing automation platforms are still architected around the individual record, so account-level rollups take real engineering. Anyone telling you this is a checkbox has not built one.

The practical compromise: start with account-level engagement counting, which you can build today from data you already own, and treat full buying-group mapping as a later phase. Partial beats theoretical.

The Playbook — Build the First Version Yourself

You do not need us for version one. Here is the whole thing.

Step 1. Pick five events and price them. Not thirty. Five. Example weights: email open 1, link click 3, pricing page view 15, second session within 30 days 20, form submit 30. Argue with the numbers — that is the point of having them written down.

Step 2. Add decay. Multiply each event by 0.5 once it passes 30 days and drop it entirely at 90. One formula. Do it in the scoring field, not in a report.

Step 3. Roll to the account. Account score equals highest contact score, plus 10 for every additional contact scoring above 20 in the last 30 days. That single line is your buying-group proxy.

Step 4. Set one threshold and one alert. Pick a number that produces 5 to 15 alerts per rep per week — no more. Alerts must contain the name, the company, the three most recent behaviors, and the timestamp. An alert without the reason attached gets ignored by week two.

Step 5. Track one metric for 60 days: meetings booked per 100 dials from scored lists versus unscored lists. If the scored list does not beat the unscored list, your weights are wrong — not the concept. Change the weights.

Keep your b2b marketing funnels for board reporting. Nobody is asking you to delete the dashboard. Just stop asking it to do a job it was never built for. The funnel reports the past. The signal layer directs the next hour.

Your Prospects Are Already Talking

They are opening, clicking, returning, reading, and quietly assembling a buying group you cannot see — and 90% of them will loop backward at least once while your funnel insists they are marching forward.

Build something that listens. Connect the systems. Track the behavior. Score meaningful engagement with weight and decay. Roll it to the account. Compress thousands of activities into a list a human can act on before lunch. Then put your salespeople where they are actually valuable: talking to the people most likely to care today.

Want to build it yourself? Steal the framework above — it is the real one. Want us to bolt it onto the stack you are already paying for? See what this costs, or work with Jeremy directly for 90 days and find out whether your database is under-scored or under-fed.