Self-Reported Attribution: Ask Buyers, Not Analytics

Every service business with a sales cycle longer than a phone call has the same hole in its reporting. The biggest deals show up in analytics as 'direct,' or as 'organic search,' or as nothing at all.

What actually happened: the buyer heard your name on a podcast, sat on it for four months, asked a peer in a private Slack group whether you were any good, and then typed your brand name into Google. Your analytics logged one thing — the last click.

Self-reported attribution closes that hole with one form field. It costs nothing, ships in an afternoon, and it is the only instrument you own that can see the parts of a buying journey a browser never touches. We put it on every lead form we build, before we touch a single tracking pixel.

Why does analytics keep telling you leads came from 'direct'?

Because 'direct' is not a channel — it is the bucket GA4 uses when a session arrives with no usable referrer. Google's own documentation defines Direct as the channel by which users arrive at your site 'via a saved link or by entering your URL,' and assigns it when source exactly matches '(direct)' and medium is '(not set)' or '(none)'.

That is a much bigger bucket than 'people who typed your URL.' Almost every high-trust, high-value referral path lands in it.

• A link pasted into Slack, Teams, WhatsApp, iMessage, or a text message — no referrer is passed
• A link inside a PDF, a deck, a desktop email client, or a mobile app
• A privacy browser or extension that strips the referrer header
• Someone who heard your name out loud and typed it into their address bar
• An AI assistant that summarized your page and the buyer navigated over afterwards

Then a second mechanism finishes the job. Google states that in GA4, 'all attribution models exclude direct visits from receiving attribution credit, unless the path to key event consists entirely of direct visits' — see Google's attribution model documentation.

Read that carefully, because it is the whole problem. Your podcast listener arrives direct, leaves, comes back a month later through a branded organic search, and converts. GA4 hands 100% of the credit to organic search. The channel that created the demand gets zero, and the channel that harvested it gets a promotion.

Why does an open text field beat a dropdown every time?

A dropdown launders every real answer into an option you wrote before you knew the answer — so it can only ever confirm what you already believed. An open text box returns the noun: the podcast name, the person's name, the community, the competitor they just left.

Think about what a buyer does when your dropdown offers Google, Social Media, Referral, and Other. She heard you on a podcast. Nothing fits. She picks Other, or Referral, or she picks whatever is first. You now have a clean, structured, useless dataset.

The free-text version of that same answer reads: 'you were on that contractor podcast in March and my ops lead sent me the clip.' That is two channels, a month, and an internal champion, from one field.

Verdict: open text, every time. The dropdown's only real advantage is that it saves you the cleanup — and the cleanup is exactly where the insight lives.

Where does the field go so it does not tank form completion?

Last field on the form, optional, never required. Zuko's form benchmarking database, covering over 93 million tracked sessions, puts the average form completion rate at 51.71% — roughly half of the people who start your form never finish it, and a required field is one more place to lose them.

Field count is the thing that hurts, not field order. Baymard Institute's 2024 research found the average e-commerce checkout contains 11.3 form fields while most sites need only 8, and that the number of form fields affects usability far more than the number of steps. That is checkout research, not lead-gen research, so apply it with care — but the direction holds.

Which means: if your quote form is already at seven or eight fields, this is not the field to bolt on. Cut something first. We cover the trade-offs in how many form fields is too many.

Run the form field and the call question together. When they conflict, the rep's answer wins — buyers under-report first touch on a form and remember it properly once someone asks them out loud. If you are not already using your confirmation page for this, read thank-you page optimization.

How do you normalize free-text answers without a data team?

You need one spreadsheet, two CRM fields, and about 20 minutes a month. Cap your bucket list at 8 to 12 categories, and never create a bucket you cannot attach to a budget line.

Store the answer twice. A source_raw field holds the buyer's exact words and is never edited. A source_clean field holds your bucket. If you only keep the bucket, you have thrown away the part that was worth having.

Build the bucket list from your own first 50 answers, not from a template you found. The buckets that matter in HVAC are not the buckets that matter in B2B software.

Two rules keep this honest. First, a bucket must map to something you can fund or defund — 'the internet' is not a bucket. Second, never guess an ambiguous answer into a bucket you like. Unclear stays unclear.

We treat 'unclear' running above roughly one in six answers as a signal that your field label is wrong, not that your buyers are vague. 'How did you hear about us?' beats 'How did you find us?' — 'find' invites people to describe their last click, which is the one thing you already have.

How do you reconcile self-reported answers against GA4's channel report?

Put two columns side by side over the same 90 days, for the same leads, and read the deltas — not the totals. The totals will never match, and they are not supposed to.

Here is the sequence we use. It takes under an hour.

• In GA4, open Traffic acquisition, set the date range to the last 90 days, and pull the default channel group split for your lead event only — not for sessions
• In your CRM, pull the same 90 days of leads, de-duplicate them, and strip spam and existing clients
• Count source_clean across those leads and convert both sides to a percentage of the same denominator: leads, never sessions
• Map the buckets that can be mapped (Organic Search to your 'search' bucket, Paid Search to 'paid')
• Leave everything that cannot be mapped — podcast, named referral, private community — in its own column. That column is the point of the exercise

Do not try to force the two into agreement. GA4 answers 'what was the last click before the form.' The field answers 'what does the buyer believe made them contact you.' Two different questions, two different instruments, both partly wrong. If you want a primer on which analytics tool to run underneath this, see GA4 vs Plausible vs Fathom.

What does the gap between the two actually mean?

The gap is your dark funnel — demand created in places that never fire a pageview: podcasts, private Slack and WhatsApp groups, LinkedIn comments, an association meeting, one operator telling another over lunch.

That is not a fringe slice of the journey any more. Gartner's research on the B2B buying journey reports that 75% of B2B buyers prefer a rep-free sales experience, and that buyers are 1.8 times more likely to complete a high-quality deal when they use supplier-provided digital tools in partnership with a sales rep rather than on their own.

The same Gartner research states that 99% of B2B purchases are driven by organizational changes. The trigger for your next deal is a reorg, a resignation, or a compliance deadline inside a company you cannot see — not an ad impression. No pixel will ever catch that. A buyer typing 'our new COO wanted a second opinion' into your form will.

Be honest about the limits: the gap is an estimate, not a measurement. GA4 over-credits the last click. Buyers over-credit whatever they remember most vividly. The truth sits between them. But a channel that keeps appearing in free text and never in GA4 is real — and a channel that dominates GA4 while nobody ever names it deserves a hard look.

How do you turn the answers into a budget decision?

One rule: a channel gets funded when a buyer who paid you names it, and gets cut when nobody has named it in 90 days. That is the same 90-day kill switch we apply to every channel we run — a line item with no qualified leads behind it does not get another quarter.

Rank buckets by closed revenue, not by lead count. A source producing four leads that close at 50% beats one producing forty that close at 2%, and the lead-count view will tell you the exact opposite.

That means source_clean has to travel from the lead record onto the deal record. If the field only lives on the enquiry and dies at the opportunity stage, you have built a vanity metric with extra steps.

• Wait for roughly 30-50 named leads before you move money — below that you are reading noise
• Pull out the named referrers by hand. Three buyers naming the same accountant is not a data point, it is a partnership you have not signed yet
• Where buyers keep naming search, that is your case for funding a B2B SEO program — the demand is real and the channel is already converting it
• Where buyers keep naming a channel you spend nothing on, that is the cheapest growth available to you this quarter
• Feed the buckets into lead scoring so your reps know which sources actually close — see intent data and lead scoring for small teams

When is self-reported attribution the wrong tool?

It is the wrong tool for any decision you have to make weekly, and it is close to worthless below about 30 answers. It is a quarterly budget instrument, not a daily optimization one.

Short, urgent, low-consideration buying cycles are the clearest miss. Someone with a burst pipe did not 'hear about you' — they tapped the first result. For emergency trades, call tracking and click data beat the form field, and we would start with call tracking software instead.

• You cannot bid-manage a Google Ads account on free text. Keep conversion tracking, offline conversion imports, and server-side tagging — see why most service businesses can't tell which ad dollar made them money
• Recall bias is real: buyers name the last vivid touch, not the first one. Someone who found you via search six months ago and saw a LinkedIn post last week will say LinkedIn
• In a buying group, the person filling in the form is often not the person who heard about you. If more than two people are involved, ask 'how did your team first hear about us?'
• It does not replace analytics or conversion tracking. It sits on top of them and answers the question they structurally cannot
• It will not fix a form nobody submits. Traffic and offer problems come first; attribution is a reporting problem, not a demand problem

The field works precisely because click tracking fails. That is not an argument for switching one off — it is an argument for running both and reading the difference.

If you sell to businesses and your reporting is mostly 'direct,' the fix is not a better analytics tool. It is one honest question on your form, and a channel plan that funds what buyers actually name. That is the work inside our B2B SEO program. Get my free audit and we will show you where your leads are really coming from.