HubSpot Lead Source Tracking: Stop Losing the Real Source

What follows is the detail: which property holds what, the five failure modes in the order they bite, and an audit that tells you in an afternoon how bad your own data is.

HubSpot ships three families of source property, and most reporting arguments come from mixing them up. They answer different questions and they are written at different moments.

• Original SourceThe channel the contact first arrived through, stamped once when the record is created and never updated again. This is your first touch, and every acquisition report should be built on it.
• Latest SourceThe channel of the most recent session, rewritten whenever HubSpot sees a new one. Useful for re-engagement, wrong for asking which campaign acquired a customer: yesterday's email click has already replaced the answer.
• The drill-downsTwo properties sit beside each source. Paid search fills them with the campaign and the term, referrals with the referring domain. The campaign name lives here; the source alone is only the bucket.

The buckets are a fixed list: Organic Search, Paid Search, Paid Social, Organic Social, Email Marketing, Referrals, Other Campaigns, Direct Traffic and Offline Sources. Two of them are confessions rather than channels. Direct Traffic means HubSpot saw a session with no referrer it could classify. Offline Sources means the contact was created without a session at all, by an import, an integration, the API or a user typing it in.

There is a chain behind every source value and every link in it can be missing. Knowing the chain turns a vague complaint about bad data into something you can check.

• The tracking code

  The HubSpot script on your pages. It reads the referrer and the query string, classifies the session and writes the cookies. No script, no session, no source.

• The hubspotutk cookie

  The visitor token that ties browsing history together in one browser. First-party to the domain the script runs on, and what a form submission needs to inherit a history.

• The form submission

  A HubSpot form sends the visitor token with the field values. HubSpot looks up the sessions behind it and stamps Original Source. Without the token there is nothing to look up.

• Everything else

  Imports, the API, a CRM integration, a rep typing in a lead after a call. These create a contact with no session behind it, which is what Offline Sources means.

The Sources report is the same session data one level up. When it disagrees with the contact properties, both are right about different populations: the report counts sessions the code saw, the property counts only those it could attach to a person.

Roughly in the order they show up in real accounts, here is what happened to the leads sitting in Direct Traffic and Offline Sources.

• The cookie was never set before the formA paid click lands on a page missing the tracking code, or on a redirect that drops the query string. By the time the visitor reaches the form, the referrer is your own site.
• The journey crossed devicesAd clicked on a phone at lunchtime, form filled at a desk two days later. Two browsers, two cookies, no link between them. In B2B that is the normal case.
• The form sits on another domain or in an embedA separate landing page tool, a booking page elsewhere, an iframe on a page that never loads your tracking code. The submission arrives with the wrong token, or none.
• The lead never used the websiteA call to the office, a reply to an outbound sequence, an event list, a rep creating the record afterwards. Offline Sources by definition, and in phone-led sales that is a lot of pipeline.
• The script did not run, or was not allowed toAn ad blocker, a browser that blocks the request, a consent banner the visitor declined. The person still becomes a contact, just without analytics attached.

A sixth pattern corrupts reporting without leaving a gap: overwriting. Latest Source is rewritten by the next session, so a nurture email quietly replaces the paid click. Build acquisition reports on Original Source, and keep any self-reported answer in its own property.

0:44

Every HubSpot thread on this subject reaches the same answer: add hidden fields for utm_source, utm_medium, utm_campaign, utm_term and utm_content. It works because a HubSpot form pre-populates a hidden field when its internal name matches a query string parameter on the page URL, so the campaign arrives in properties you own. Do it. On its own, though, it is a last-touch snapshot of one page load.

The three ways a hidden UTM field comes back empty

• It only carries what that landing page URL had. If the visitor read two more pages and came back to the form through the navigation, the parameters have left the address bar.
• It captures the last touch, not the first. A paid click in March, a newsletter conversion in June: the form fill is honestly tagged as email, and the acquisition credit belonging to the ad does not survive.
• It is blank for anyone who came back later. A lead returning through search or a bookmark submits empty fields, and a careless workflow lets that blank overwrite the good value you captured first.

There is a plumbing tax too. The fields have to exist on every form, in every embed, and again on the next form somebody builds in a hurry. Make the set part of the form template and audit it monthly. Our guide to UTM parameters covers the tagging convention behind the values.

A setup that survives has four ideas in it: capture at the first visit rather than at the form, keep first touch and last touch apart, store the click ID as well as the campaign, and never let an empty value win.

1. Fix tracking code coverage first. Every page a paid click can land on needs the code, and ads should point at the final URL rather than a redirect that strips parameters.
2. Agree a hidden field convention. One property per parameter, internal names matching the query string exactly, the whole set in the form template so new forms inherit it.
3. Create a first-touch and a last-touch pair. Source, campaign and date for each. Fill the first-touch trio only when it is empty. This one rule stops June erasing March.
4. Store the click ID, not just the UTMs. gclid with gbraid and wbraid, fbclid, li_fat_id, msclkid. A UTM is what you labelled the traffic; a click ID is the platform's own key. See the click ID glossary entry.
5. Persist it in first-party storage on your own domain. Read the parameters on the first page view, store them, read them back at the form fill wherever it happens.
6. Skip blank inputs in every workflow. Most lost source data was captured correctly once, then overwritten by a later submission with nothing in the field.
7. Copy the source onto the deal at creation and freeze it. Contact properties keep changing; a deal needs the source as it stood the day it opened.

Leads created by a rep will always be Offline Sources, so give the team a short required picklist: inbound call, outbound, referral, event, partner. If calls are a real acquisition channel, tracking numbers per campaign attach the call to a session and move those leads out of the offline bucket.

Once the source is on the deal, the question changes shape. Counting contacts by Original Source tells you which channel fills the top of the funnel, the number most likely to flatter cheap traffic. Summing deal amount by first touch tells you which channel pays for itself, and the two lists are rarely in the same order.

Build three views and stop there: closed won amount by first touch campaign, pipeline amount by first touch campaign, and contact count by source with the conversion rate to qualified beside it. When the cycle is long, add time to close by channel: a campaign that looks weak on a 30 day view often closes on a 90 day one. That is what long sales cycle attribution solves.

The source also belongs on the deal because it can travel back out: a stored click ID plus a deal amount is what an ad platform needs to take a closed deal as an offline conversion.

Before changing anything, put a number on the problem. An afternoon of work gives you the figure that justifies the fix and a baseline to measure against.

1. Export a contact list for the last 90 days with Original Source, both drill-downs, create date and lifecycle stage as columns.
2. Count the share of rows in Direct Traffic and Offline Sources. Above roughly a fifth, for a business that buys traffic, means the chain is broken upstream.
3. Set what Google, Meta, LinkedIn and Microsoft report for the same window beside the contacts HubSpot attributes to paid. Why the platforms never match your CRM explains which part of that gap is normal.
4. Sample twenty Direct Traffic contacts and work each back by hand: which form, which page, does it carry the code, do the ad accounts show a click that day.
5. Check every form and landing page for the tracking code and the hidden fields. The failures are your first sprint.
6. Check the deals too. If deal records carry no source property, no contact-level tidying will produce a revenue report.

Write down the Direct Traffic share and the platform gap before you start. In three months those two numbers are how you prove the work was worth doing.

There are three levels of ambition here and they stack: the second assumes the first works, the third assumes both. Pick the one that matches how much revenue depends on the answer.

The honest summary: the first level is free and incomplete, the second cheap and fragile, the third costs a subscription and closes most of the gap. If you spend five figures a month and cannot name the campaigns behind the deals, the arithmetic decides for you.

LeadJourney is one script on your site or in your GTM container, live in about 21 minutes without a developer. It runs server-side on your own domain at 95%+ tracking accuracy, so Safari, iOS and ad blockers take far less out of the data. At the first visit it captures gclid with gbraid and wbraid, fbclid, li_fat_id and msclkid, plus the UTMs and the landing page. That anonymous first click is joined to the person at the form fill, call or booking, so the March ad is still attached in June.

The HubSpot integration runs both ways. It reads your lifecycle and deal stages and the deal values, and writes the first touch, campaign, click ID and full journey onto the contact and the deal. You then map any stage you choose, qualified lead, proposal sent or closed won, to a conversion action: Meta over the Conversions API, Google Ads as an offline conversion import matched on the gclid, LinkedIn over its Conversions API, Microsoft Ads as offline conversions, with the deal amount and currency travelling with the event. Attribution models switch between first click, last click, linear, time decay and position based without re-tracking.

The limits are worth saying out loud. Tracking is first-party and respects consent decisions rather than working around them, so a visitor who declines is not measured, and a lead that never touched your site still needs a source typed in by a human. Call tracking closes the phone gap, but numbers are billed per number plus usage on top of the plan, which starts at €129 a month by marketing spend, hosted in Frankfurt under a GDPR data processing agreement.

Read verified reviews on Trustpilot, G2 and leadjourney.io/testimonials.

Keep going with why GA4, Meta and Google never match your CRM, the GCLID guide on capturing and storing the click ID and our UTM parameters explainer. On the product side, see fixing the lead source in your CRM, the HubSpot integration and the lead source glossary entry.