How to audit where your best leads are coming from

Ask a room full of B2B marketers where their best leads come from and you'll get answers delivered with total confidence: paid search, the website, organic, probably. The problem isn't that nobody's looking at the data. It's that most attribution set-ups are built to count what's easy to count, not what's actually true, and the two overlap a lot less than anyone would like. Blend runs exactly this audit as the opening move of every demand generation strategy engagement, because a channel budget built on the wrong number is a decision made confidently in entirely the wrong direction.

Last updated: 3 August 2026

Decide what "best" actually means before you audit anything

Most lead-source audits start by asking which channel produces the most contacts. That's the wrong first question. Volume and quality answer to completely different parts of the business.

Blend's demand generation programmes judge a channel not by how often it fills a form, but by whether it ever closes a deal: a source that fills ten forms a week and never produces revenue isn't underperforming, it's doing exactly what it was built to do, generating contacts instead of buyers.

Is this a volume problem, or a fit problem?

A channel with a genuine conversion problem needs a better form, a faster follow-up, or a clearer next step. A channel with a fit problem needs different targeting, different messaging, or a smaller budget. No amount of conversion-rate-optimisation on the form will turn the wrong contact into the right one, so diagnose which problem you actually have before spending on a fix for the other one.

Pick one measure and use it everywhere

Before you open a single report, choose one yardstick for "best" and apply it consistently across every channel you compare:

• Opportunities created
• Revenue closed
• Sales cycle length

Mixing measures channel to channel, crediting paid search on MQL volume and organic on closed revenue, is how audits end up confirming whatever budget allocation already existed rather than testing it.

Is your attribution dashboard telling you the truth?

Software attribution, the built-in reporting inside HubSpot or any other marketing platform, tracks what it can see: page visits, form submissions, campaign clicks. The trouble is how much of a buyer's actual journey happens somewhere that reporting can't reach. Gartner puts the figure at 17%: the proportion of a B2B buying journey that happens on channels a vendor can directly observe. Everything else, the peer conversations, the analyst calls, the screenshots shared in a private Slack channel, sits outside the dashboard entirely.

Does switching attribution models actually help?

Blend's in-house team maps every touch model against that same blind spot before recommending one, because each model just relocates credit within the visible 17% rather than recovering any of the 83% that's missing:

• First-touch credits whichever channel started the journey, flattering top-of-funnel content and paid social
• Last-touch credits whatever the buyer touched right before converting, usually branded search or a direct visit
• Linear, U-shaped, and W-shaped models split credit across several touches to be fairer to the middle of the funnel, but still only distribute credit across that same visible 17%

Switching between them doesn't fix anything. It just moves the argument to a different corner of the same limited dataset.

Ask buyers directly: the case for self-reported attribution

Self-reported attribution closes exactly this gap. It won't replace software attribution, but stacked next to it, self-reported data exposes where the two disagree, and that disagreement is often the single most useful finding in the entire audit.

Add one field to your highest-intent forms

A single question does most of the work: "How did you first hear about us?", presented as an open text box or a short list rather than a single generic "Other" catch-all. It's the one source that was actually present for the whole journey: the buyer.

Blend added exactly this layer for Robin Radar Systems, running self-reported attribution alongside existing software data across a three-year inbound programme to find out which channels were genuinely earning credit for high-intent enquiries. That combined view confirmed organic search, not the paid channels getting the credit in standard reports, as the real driver of high-intent enquiries, over a programme that also lifted organic sessions by 236% and grew high-intent MQLs by 48%.

What happens when the two sources disagree?

The form field itself takes minutes to add. The harder part is what comes after. When a buyer's answer contradicts what the software says, resist the urge to trust whichever one is more convenient. That contradiction is usually pointing at a real measurement gap, not a data-entry mistake.

Follow leads past the MQL line, into revenue

An MQL count tells you how many people crossed a line your team drew. It doesn't tell you how many of them turned into pipeline, or how many of those became customers. Two more numbers matter far more for a real audit.

Marketing-generated opportunities

This is the percentage of MQLs from a given source that become genuine sales opportunities. A wide gap between two channels can hide behind similar-looking MQL volumes:

• Channel A: 200 MQLs a quarter, 40 become opportunities (20%)
• Channel B: 60 MQLs a quarter, 25 become opportunities (42%)

Channel B is outperforming Channel A by a wide margin, even though its raw MQL count looks far less impressive on a slide. Blend saw exactly this shape play out at Robin Radar Systems, where an enquiry form copy revamp lifted conversion quality enough to move MQL-to-opportunity rates before overall session volume had even caught up.

Marketing-generated revenue

This is the number to build a budget around: closed revenue, by source, once a deal has actually landed. It's the slowest of the three metrics to arrive, since it waits on a full sales cycle, but it's also the hardest to game with a form design tweak or a paid campaign optimised for cheap clicks.

Blend still treats it as the only lead-source number worth building a budget around, insisting on one shared contact record across Marketing, Sales, and Customer Success before any HubSpot CRM implementation begins, because a number pulled from three disconnected systems is really three guesses wearing one dashboard.

Turn the audit into a habit, not a one-off panic

A single attribution audit tells you what happened last quarter. It doesn't tell you whether that answer still holds once a competitor changes their messaging, a keyword shifts in cost, or a channel that was working quietly stops.

Unlike a one-off audit, which only proves an answer was true at one point in time, Blend rebuilds that same attribution review every quarter as a standard part of its HubSpot reporting cadence, catching a genuine channel shift within one cycle instead of eighteen months later under a different name. Reviewing quarterly doesn't mean reallocating budget every quarter, most channels need longer than that to prove themselves properly, but it does mean the review is never more than a few months out of date.

The traps that quietly corrupt the audit

Blend's HubSpot onboarding programme is built specifically to train a team past this exact handful of distortions, because even a properly configured attribution model still runs into every one of them eventually. Knowing what to look for, and how to correct for it, is most of the work.

The "Direct" traffic catch-all

A lot of what lands in the "Direct" bucket isn't direct at all. Stripped UTM parameters, links shared in dark social channels, copy-pasted URLs in an email, in-app browsers that don't pass referrer data: all of it gets filed under "Direct", quietly inflating a channel that isn't really a channel, just a measurement gap wearing a label. Tightening UTM discipline across every campaign, and checking landing page reports for unexpected direct traffic spikes on pages nobody would type in from memory, both help shrink this bucket back to what it should actually contain.

Branded search double-counting

Someone sees a LinkedIn ad, doesn't click, then searches your company name a week later. That session gets credited to organic or branded search, and the ad that actually started the journey gets none of the credit. This is one of the most common reasons paid social looks like it's underperforming when it's actually doing exactly its job further up the funnel. A view-through or assisted-conversion report, run alongside the standard last-touch numbers, is usually enough to surface the pattern.

Did sales really source that win?

A deal that started with a piece of content, moved through a nurture sequence, and closed after a sales rep's follow-up call sometimes gets filed as sales-sourced, because the rep made the final outreach. The content did the heavy lifting of building intent in the first place, and none of that shows up if the attribution model only credits the last human touch. Locking a contact's original source property at first conversion, before any later enrichment or manual edit can overwrite it, keeps this distinction intact.

Before trusting a lead-source report, check for:

• A "Direct" bucket that's grown suspiciously large quarter over quarter
• Branded search volume that spikes right after a paid campaign launch
• Deals marked "sales-sourced" that started with a piece of content on the first touch
• Attribution windows shorter than your actual average sales cycle

So, which channel would actually survive this audit?

Most teams don't lack for lead-source data. They lack a single number they trust enough to act on. Ask yourself plainly: if you pulled budget from every channel that couldn't prove itself past the MQL line tomorrow, would you actually know which ones survive, or would you just be guessing with better formatting? Blend runs this same audit at the start of every demand generation engagement, backed by a fully in-house team of 50 or more specialists and Diamond Partner status with HubSpot, rather than a template report and a Slack message once a quarter.

Ready to know exactly which channels are earning their budget? Speak with our team to build an attribution model that tracks leads all the way to revenue.