What is lead attribution? How to know where your leads came from

Lead tracking answers "what happened to this enquiry?" — when it arrived, whether anyone opened it, whether anyone acted on it, and what it became. Source tracking adds the "where did it come from?" layer to the same record, the recorded touches; lead attribution is the reading of which of them deserve credit. The companion guide to lead tracking covers the first question. This one stays on the second.

For a local service business the question is practical, not academic. Enquiries arrive from an ad, a search result, a directory, a recommendation, or a typed-in address, and the enquiry itself does not carry the answer unless it was recorded at the time. Source tracking is the discipline of recording that context while it is still available; attribution is the discipline of reading it honestly: as evidence of what was observed, not as a verdict on why someone chose you.

Where source tracking ends and attribution begins is a matter of definition, so this guide draws its own line.

• Source tracking records the acquisition context that arrived with an enquiry: the campaign tags on the link, the referring site, the page the person landed on, the click identifiers an ad platform attached. It is a recording activity. It can be complete, partial or empty, depending on what the visit carried.
• Attribution interprets that record: which touch counts, and how credit is shared when there is more than one. It is a reading activity, and it depends on a rule. Change the rule and the same recorded touches produce a different answer.

Keeping them apart matters because they fail differently. Source tracking fails when nothing was captured. Attribution fails when a rule is applied to thin evidence and the result is read as fact. The rest of this guide follows that split: first what gets recorded, then how to read it.

A lead record can hold three recorded views of where an enquiry came from, and they answer three different questions. This is the section to slow down on, because treating them as a single answer is one of the places attribution goes wrong.

• First Touch answers "how was this person first observed arriving?" It is the earliest attributable touch that was captured for the visitor before they enquired. It does not say that touch created their interest, only that it is the first one on record.
• Last Touch answers "what was the last attributable touch captured before the enquiry?" It is the most recent recorded touch, not the reason the person filled in the form. A last touch can be a return visit made after the decision was already taken.
• Recorded source answers "what single source label was stored on this lead when it was created?" It is a classification made at that moment from the evidence that arrived with the enquiry, and it is stored as one value. That classification names the evidence that arrived; it is a recording step, not a decision about credit between touches.

The three can legitimately disagree. A person can arrive first from a link carrying campaign tags, come back later from an ad, and be recorded under a label that matches the first touch, the last touch, or neither, because the Recorded source is classified from the evidence that arrived with the enquiry, which is not necessarily the same evidence as either card. That is not a bug in the record; it is the record telling you that the visitor's path had more than one point on it. In LC Flow the First Touch and Last Touch are derived independently and shown separately: each card comes from its own captured evidence, and the Recorded source, a separate classification of what the submission carried, is displayed once, clearly named, beneath them. The product deliberately does not assert that First and Last were one path, because it has the evidence to show each touch and not the evidence to join them.

That is also where LC Flow's implementation sits on the attribution question. It records and presents First Touch, Last Touch and Recorded source as three separate fields; it does not assign fractional credit across those fields, and it does not claim that any one of them caused the enquiry. The credit-reading, deciding how much of the credit each recorded touch should get, is left to you and to the analytics tools built for it. What you get is three recorded views you can trust for what each one is.

Do not collapse the three into one "true source". Do not read First Touch as the cause of awareness or Last Touch as the cause of conversion. Each is a recorded observation with a precise meaning, and the precision is the value.

A captured touch is only as detailed as the visit that produced it. When the evidence was present, a touch can show:

• Source and medium: the campaign tags naming where the link was placed and what kind of placement it was.
• Campaign: the campaign name carried on the link, when one was set.
• Content: the variant or placement tag, when one was set.
• Search term: the term tag carried on the link, when one was set.
• Landing page: the page the touch landed on, when that page was set up to record it.
• Referrer: the site that sent them, when the browser passed one along.

Every one of those is conditional. A person who typed your address into their browser arrives with none of the campaign tags; a link shared in a message may carry no referrer; an ad link may carry a click identifier and nothing else. A record that shows only two of these fields is not broken. It is showing you what arrived. Do not expect every field on every enquiry, and do not promise yourself that you will have it.

This is the distinction that separates an honest source record from a flattering one. "Direct" is a claim about how someone arrived: they came without an attributable referrer, campaign tag or click identifier, the way a typed-in address or a saved bookmark arrives. "Not captured" is a statement about one layer of the record: no attributable touch was captured on that side, First or Last, so that layer says nothing about how they arrived. Another layer of the same record may still hold something. The absence of evidence is not evidence of direct navigation.

The failure to watch for is a system that files the gap under Direct and never tells you that this is what its Direct means. Do that and your "Direct" figure quietly absorbs every tracking gap, every link that lost its tags, every visit the browser did not pass a referrer for, and the channel that looks strongest in the report can be the one you understand least.

LC Flow keeps the two apart by layer. A First Touch or Last Touch that was not captured is shown as exactly that, "not captured", never as Direct, and when neither side was captured the Attribution Journey is shown as not available. The Recorded source, being one label chosen at creation, does carry "Direct" when the evidence that arrived with the enquiry held nothing attributable: no click identifier, no campaign source tag, no referrer, and nothing carried over from an earlier visit that the browser still held. The report note says in plain words that Direct means no attributable source was recorded, not that the person typed your address. Read it that way. A lead whose Recorded source says Direct while no attributable touch was captured on either side is a lead for which no attributable signal was captured at any layer, and the honest reading is "no attributable source was recorded", not "they came straight to us".

A source breakdown on its own can mislead. If a captured touch exists for only part of the leads in a period, a chart of "where leads came from" is really a chart of where the leads with a captured touch came from, and the leads with no captured touch are the part of the story it cannot tell. The denominator is the whole point.

Coverage is that denominator made visible: of all the leads in the period, how many had a First Touch captured, and how many had a Last Touch captured. Read the breakdown against the coverage every time. A source that dominates a small covered set may be nowhere near dominant across all your enquiries. In LC Flow's reports, the Attribution journey section states the coverage of each side as a count of leads out of the period's total, and the breakdowns in the two cards beneath are counts among the leads that had a captured touch. The Recorded source breakdown is a different kind of figure: it accounts for the full lead population of the period, and leads without a recorded classification are kept in their own bucket, shown as "Unknown", never folded into Direct. It should not be confused with touch coverage.

Attribution is a reading of how an enquiry arrived. It says nothing about what the enquiry became. That part belongs to the rest of the lead record: whether the person was contacted, whether a quote went out, whether the job was won or lost, and, on won work, the value the business entered.

Because the attribution context and the outcome sit on the same lead record, they can be read together for that lead. The safe sentence is a plain one: this lead had this recorded source, and later this recorded outcome. That is an association you can see. It is not proof that the source caused the job, and a value typed against a won lead is what the business recorded, not something the source produced or the software observed.

Two sentences to keep out of your own reporting, however tempting: "this source made us this much money" and "this channel produced these jobs". Both convert an association into a cause. The honest version is quieter and still useful: among the leads that became work, these were the recorded sources, and of those same leads this many had a First Touch captured and this many a Last Touch.

Google Analytics measures a website: sessions, pages, events and the paths people take before a meaningful action, with attribution models that decide how credit is shared along those paths. In Google's own documentation, attribution is described as assigning credit for important actions across the ads, clicks and other touchpoints on a person's path, with the result depending on the attribution model applied. Google Analytics currently offers three models: data-driven attribution, paid and organic last click, and Google paid channels last click. The point to take from that is not which model is right. It is that attribution is model-dependent: the credit an ad or a search gets is a product of the rule or algorithm applied to the path.

A per-enquiry lead record is a different instrument. It holds one person's enquiry with the touches that were captured for them, their Recorded source, and what happened next, so the business can act on that enquiry and later read source and outcome side by side. First Touch, Last Touch and Recorded source are three recorded observations, not a replacement for any attribution model, data-driven or last click, and they answer a narrower question than a site-wide attribution report does. The two sit side by side: the analytics tool for the site and its paths, the lead record for the people who asked. Neither makes the other redundant.

• Reading "Direct" as proof someone typed your address. A gap in the record is a gap. A Direct figure that is not defined can absorb tracking gaps and become a channel you cannot check.
• Reading Last Touch as the cause. The last captured touch can be a return visit by someone who had already decided. It is a fact about the path, not the reason for the enquiry.
• Calling a source label a customer journey. One label, or even a first and a last touch, is not the whole path a person took. It is the part that was captured.
• Reading a source breakdown without its coverage. A percentage of the leads with a captured touch is not a percentage of your leads.
• Assuming a campaign field is the ad platform's real campaign name. The campaign on a touch is whatever the link carried. It is only as accurate as the tagging that put it there.
• Reconstructing the source later from memory. "I think they found us online" typed in a week later is a guess with a label on it. Record at arrival or record "not captured".
• Confusing platform-reported conversions with your leads. An ad platform counts what it can see through its own tags and windows; your lead record counts enquiries. They measure different things and should not be expected to match.
• Confusing association with causation. A recorded source beside a won job is a pairing you can read, not a proof of what earned the money.

This is how LC Flow's website lead tracking for small service businesses handles the "where from" layer; the principles above hold whatever you use. For website forms, capture depends on the LC Flow snippet running on the page and your domain being verified. Each lead then carries its Attribution Journey: a First Touch and a Last Touch, each shown with its captured detail when an attributable touch was captured and marked "not captured" when it was not. When neither side was captured, the journey is shown as not available. A captured touch shows the source group it was classified to and the campaign, source, medium, content, search term, landing page and referrer it actually carried. Beneath them sits the Recorded source: the single label stored when the lead was created, displayed once and named as such. Each source is classified from the evidence that arrived into one label: a touch that carried a recognised advertising click identifier or campaign tag is labelled with the advertising source it points to, and the other labels include organic search and referral. Only the Recorded source is ever classified as direct, because a touch exists only when something attributable was captured. A touch that was not captured is marked as such rather than guessed, and a Recorded source of Direct is read as the report note defines it: no attributable source was recorded.

In reports, the Attribution journey section shows the coverage of each side as a count of leads out of the period's total, the counts by source group among the leads that had a captured touch, the campaign counts, and the Recorded source breakdown across every lead. The same lead record later carries its stage and, on won work, the value you entered, so for any one lead the recorded source and the recorded outcome can be read together. If you would rather see a real Attribution Journey than read about one, you can start a 14-day free trial, no credit card required.

A source record earns trust by being clear about its edges. These are the ones to hold on to, whichever tool you use, including ours.

• A touch does not prove it caused the enquiry. It proves the visitor was observed arriving that way at that point. Why they enquired is not in the record.
• Last Touch is not the "conversion source". It is the last attributable touch that was captured before the enquiry, nothing more.
• A missing touch should not be guessed. "Not captured" is a complete and honest entry. Back-filling it from a hunch turns the record into an opinion.
• A source beside a won job does not credit the source with the money. The two facts live on the same lead. Reading them together is fair; treating the source as the reason for the revenue is not.
• Data quality depends on captured evidence. Untagged links, stripped referrers and links opened from a message can all arrive with less. The record can only be as good as what reached it.
• Source tracking does not create a complete multi-touch journey. First and Last are two recorded points, chosen because they can be recorded honestly. The path between them, and before them, is not in the record.

None of that makes attribution useless. It makes it specific: a reading of what was recorded, with its gaps labelled, set beside what the enquiry became. That is enough to answer the question a service business actually has, which is less "which channel deserves the credit?" than "where were the people who asked us for work observed coming from, and how much of that do we actually know?"

• Get started with attribution — Google Analytics Help — Google, Analytics Help, retrieved 16 September 2026. Paraphrased above for Google's description of attribution and attribution models, and for the three models it offers. It is a description of Google Analytics, not of LC Flow.

No statistics are used in this article. Every statement about LC Flow describes the product as inspected at the time of writing; every statement about attribution in general is reasoning, not a measurement.