Cut that channel and watch your best customers disappear

Budget reviews follow a predictable pattern. Someone pulls the attribution data, ranks the channels by volume, and recommends cutting the ones at the bottom. The channel that generated the fewest leads gets flagged. The one with the highest cost per lead gets questioned. The one nobody can attribute pipeline to gets put on the chopping block.

The problem is that the channels at the bottom of the attribution report aren't necessarily the ones performing worst. They might be the ones the attribution model can't see properly.

The event that generated five leads also generated three conversations that influenced two enterprise deals. The attribution model counted the five leads. The two deals were attributed to the sales rep who closed them. The podcast that produced zero trackable conversions was mentioned by four closed-won customers in their post-deal interviews as the reason they first looked at the company. The attribution model gave it zero credit.

The channels that produce the highest-quality customers are often the ones that look worst in the attribution report because they operate through influence rather than direct conversion. They don't produce clicks. They produce trust. And trust doesn't register in a last-click attribution model.

The attribution model has a visibility problem

Attribution models credit what they can track. Clicks, form submissions, page visits, email opens, ad impressions. Every touchpoint that produces a trackable event gets credit in the model. Every touchpoint that doesn't produce a trackable event gets nothing.

This creates a systematic bias toward channels that are easy to track and against channels that are hard to track. Paid search is easy to track. Events are hard to track. Email campaigns produce clear touchpoints. Word of mouth produces none. Content downloads register in the MAP. A podcast episode that someone listened to while commuting and then mentioned to their colleague doesn't register anywhere.

The model isn't wrong about what it can see. It's incomplete about what it can't. And decisions made on incomplete attribution systematically defund the channels that build trust and relationships in favor of the channels that produce trackable clicks.

Over time, this creates an increasingly transactional marketing mix. The channels that survive budget reviews are the ones that produce measurable short-term conversions. The channels that build long-term brand affinity, trust, and word of mouth get cut because nobody can prove their value in a spreadsheet.

Then leadership wonders why the pipeline is full of low-quality leads that don't close. The high-quality leads were coming from the channels that got cut.

The quality signal hiding in your CRM

The data to evaluate channel quality already exists in most CRMs. It's just not being used for channel decisions.

Pull your closed-won deals from the last 12 months. Look at the customers who closed fastest, expanded soonest, and renewed without a fight. These are your best customers. Now look at how they found you. Not what the attribution model says. What they actually told you, or what you can piece together from the sales notes, the deal history, and the onboarding conversations.

In most B2B organizations, the pattern is consistent. The best customers came through referrals, events, content that built trust over months, and channels that the attribution model either undercounted or missed entirely. The fastest-closing deals were influenced by peer recommendations that no marketing system tracked. The highest-value accounts were warmed by content consumption patterns that the attribution model credited to the wrong touchpoint.

The worst customers, the ones who churned early, demanded excessive support, or never expanded, disproportionately came from the high-volume channels that look best in the attribution report. Paid search, high-volume webinars, gated content campaigns. These channels are good at generating leads. They're not always good at generating the right leads.

This isn't a universal rule. Some high-volume channels produce excellent customers. Some low-volume channels produce poor ones. The point isn't that one type of channel is better than another. It's that channel quality should be evaluated on customer outcomes, not just lead volume, and most organisations never make that connection.

Self-reported attribution reveals what the model misses

The simplest way to see what the attribution model is missing is to ask the buyer directly.

Add a free-text field to your high-intent conversion forms: "How did you hear about us?" Not a dropdown with predefined options. A free-text field where the buyer writes whatever comes to mind.

The answers will diverge from what your attribution model reports. The model says "organic search." The buyer says "a colleague recommended you." The model says "direct traffic." The buyer says "I asked ChatGPT and your company came up." The model says "paid social." The buyer says "I've been reading your blog for months and finally decided to reach out."

Each of these divergences represents a channel that's producing results the attribution model can't capture. The colleague who recommended you doesn't appear in any report. The AI conversation that surfaced your brand doesn't register as a touchpoint. The months of blog reading that built trust got credited to whatever the last trackable click happened to be.

When you collect enough self-reported attribution data, patterns emerge. Channels you thought were underperforming turn out to be driving significant interest through invisible paths. Channels you thought were your top performers turn out to be capturing demand that was created somewhere else.

How to protect the channels that matter

The fix isn't to abandon attribution modeling or stop measuring channels. It's to add a quality layer on top of the volume layer so channel decisions account for both.

Track customer quality by original channel. Not just lead-to-opportunity conversion, but longer-term metrics: time to close, deal size, retention rate, expansion revenue, customer satisfaction. If a channel produces fewer leads but those leads close faster, stay longer, and spend more, the channel is more valuable than the volume suggests.

Use self-reported attribution alongside model attribution. Run both. When they agree, you have high confidence. When they diverge, investigate. The divergence is where the invisible channels live, and those channels may be the ones producing your best customers.

Evaluate channels on a 12-month window, not a quarterly one. Some channels produce results that take months to materialise. An event in Q1 might influence a deal that closes in Q3. A content series that runs for six months might produce pipeline in month eight. Quarterly attribution windows systematically undervalue slow-burn channels that build trust over time.

Before cutting a channel, ask your sales team. Not what the data says. What they hear in conversations. "How do our best customers find us?" is a question the sales team can often answer from experience, and their answer usually includes channels that don't show up in the attribution report.

The channel that looks worst might be worth the most

Budget decisions based purely on attribution data will always favor the trackable over the influential. That bias doesn't just misallocate budget. It gradually reshapes the entire marketing mix toward short-term, transactional channels and away from the long-term, relationship-building channels that produce the customers every company actually wants.

The channel you're about to cut might be the one that brought you the three accounts you'd most hate to lose. The attribution model just didn't know how to count them.

Before the next budget review, check the customer quality data alongside the lead volume data. The story they tell might be very different. And the channel at the bottom of the attribution report might be the one that deserves more investment, not less.