# How to Track Which Message Generated a Customer > Connect a message to a customer with one tracked link per message, a CRM activity log keyed to a message ID, and a documented rule over a fixed window. - **Author**: Caius Hayes - **Published**: 2026-07-28 - **Category**: Marketing - **URL**: https://heyzinc.com/blog/track-message-generated-customer --- ```tldr You can connect a single message to a customer by giving each outbound message its own tracked link, logging every touch as a CRM activity tied to a stable message ID, and applying one documented attribution rule over a fixed window. No attribution platform required. ``` You close a deal. You open the CRM. The "Lead Source" field says *organic search*, or *direct*, or the name of a newsletter you stopped sending six months ago. You know that is not what happened. You remember the reply you posted in a thread three weeks ago. You remember the DM you sent on a Tuesday. You remember the email that finally got a response. The CRM does not remember any of it, because nobody told it to. The question "which message generated this customer" is answerable. It is just not answered by the default source field, and it is not answered by buying a bigger attribution tool. It is answered by a habit: one tracked link per message, one activity log keyed to that message, and one rule you write down and keep. This is a method for founder-led sales, where the unit of work is a conversation, not a campaign. ## Why "Lead Source" lies in founder-led sales The source field on a contact or deal is not a measurement of where the customer came from. It is a record of what the system happened to see at a specific moment, usually the referral on the form fill, or the first-touch value a workflow stamped and never revisited. Configurable systems do not force one behavior here. HubSpot, for example, lets you attribute to contact creation, deal creation, or revenue, and lets you choose among first-touch, last-touch, and several multi-touch models. Salesforce has its own Campaign Influence feature for tying opportunity revenue to campaigns. The point is not which default each vendor picks. The point is that *without a deliberate rule*, the field reflects whatever the last configuration happened to capture, and in founder-led distribution that is almost never the real origin. Founder-led distribution breaks the source field on purpose, without meaning to. You post a reply in a community. You send a DM. You answer a question on social. The visitor clicks, arrives, and the referral arrives as *direct* or with a generic label, because private sharing, in-app browsers, redirects, and privacy tooling all strip the context that a campaign dashboard was built to read. We covered the mechanics of that in [why "direct traffic" is hiding your best channels](https://heyzinc.com/blog/why-direct-traffic-hides-best-marketing-channels). The short version: "direct" is an unknown-source bucket, not a behavior. So the source field is not lying out of malice. It is lying because it was built to answer "which campaign," and you are asking "which message." ## The unit of attribution is the message, not the campaign Campaign attribution asks: did Reddit work? Did the July email sequence work? Did the webinar work? Those are reasonable questions for a marketing team running a handful of broad initiatives. They are the wrong questions for a founder who had ten meaningful conversations today across replies, DMs, and emails, each one a separate bet on a separate person. Per-message attribution asks a different question: did *this specific reply* start the relationship that became this customer? The published piece on [tracked links vs. UTM parameters](https://heyzinc.com/blog/tracked-links-versus-utm-parameters) makes the same shift for links: one link per conversation, not one link per campaign. This article takes the next step: from "which conversation created the click" to "which message created the customer." That step matters because the click is the middle of the story, not the end. A click tells you a message earned attention. It does not tell you whether that attention became a meeting, a deal, or revenue. To know that, the message has to stay attached to the record all the way to close. This is a discipline, not a platform. You can do it with the CRM you already have. ## Three building blocks of per-message attribution The whole method reduces to three things. **1. A unique tracked link per message.** Every outbound message (every reply, DM, email, follow-up) gets its own link, and that link carries the context of the conversation it came from: the source page, the platform, the thread. This is the per-message layer that UTMs were never built for. (If you also run broad campaigns, a [UTM link builder](https://heyzinc.com/tools/utm-link-builder) is still the right tool for campaign-level URLs; per-message work wants per-message links.) The link is what lets you later point at one specific message and say "this one." **2. A CRM activity log keyed to a message ID.** Every touch (sent, clicked, replied, called, met) is logged as an activity on the contact record and the deal record, and every one of those activities is tagged with a stable message ID. This is where most teams silently fail. HubSpot's attribution documentation is explicit about one rule that is easy to miss: sales activities like calls, meetings, sales emails, and conversations only count toward deal and revenue attribution when they are associated to **both** the contact record *and* the deal record. Log a call on the contact but forget to associate it to the deal, and as far as the revenue report is concerned, the call never happened. The message ID is what turns a pile of activities into a path you can read later. **3. A documented attribution rule and a fixed window.** Pick a rule (first-touch, last-touch, or a multi-touch model) and a window. Write both down. Apply them to every deal the same way. The rule decides *which* message gets the credit; the window decides *when* a message stops being eligible. Without a written rule, every closed deal gets attributed by whatever mood the person closing it is in, and the numbers mean nothing. Three building blocks. The rest is execution. ## First-touch, last-touch, and the multi-touch middle ground The rule you pick decides which message "generated" the customer. Each rule answers a different question, and each has a bias you should know about. **First-touch** gives all the credit to the message that started the relationship, the first reply, the first DM, the first email that got a response. HubSpot calls this the First Interaction model. It tells you what is working at the top: which conversations are worth having at all. Its blind spot is obvious: it ignores the message that actually closed the deal. A great opening reply that never converted still looks like a winner. **Last-touch** does the opposite: all credit to the final message before the conversion. HubSpot calls this Last Interaction. It tells you what is closing. Its blind spot is just as obvious: it starves the work that made the customer know you existed in the first place. A founder who only reads last-touch will conclude that the closing email is magic and stop doing the community work that fed it. **Multi-touch** splits credit across every touch. Linear gives each touch equal weight; U-shaped and W-shaped give more weight to the first and last and to specific conversion moments; time-decay weights more recent touches more heavily. HubSpot ships all of these. Multi-touch is the honest answer when a deal genuinely touches many messages, which most founder-led deals do. The catch is that multi-touch demands more logging discipline than most small teams sustain. You need every touch logged with a message ID, associated to both contact and deal, in a system that can do the math. If your logging is patchy, a multi-touch model confidently produces wrong numbers. A pragmatic middle ground for small teams: capture **first-touch + last-touch + a raw count of touches**. First-touch tells you what started it. Last-touch tells you what closed it. The count tells you how much work it took. That is enough to decide where to spend the next hour, and it is honest about what it does not know. One wrinkle worth knowing: the classic rules-based models (first-click, linear, time-decay, position-based) were removed from Google Analytics 4 in November 2023 in favor of data-driven attribution. The CRM side still ships them; HubSpot's attribution reports offer the full set. So if you want a model you can reason about by hand, do it on the CRM side, not in GA4. HubSpot's [attribution reporting guide](https://knowledge.hubspot.com/reports/understand-attribution-reporting) lays out the models and their tradeoffs, and its [create attribution reports](https://knowledge.hubspot.com/reports/create-attribution-reports) page covers the three report types (contact create, deal create, revenue). Google's [attribution overview](https://support.google.com/analytics/answer/10596866) documents what GA4 currently offers. ## Attribution windows: pick one and stick to it A window is how far back a message can reach and still claim credit for a conversion. Without one, a reply you posted eighteen months ago can quietly take credit for a deal that closed last week. That is zombie credit, and it ruins your numbers. The convention is familiar because Google publishes it: GA4's lookback window options are **7, 30, 60, or 90 days**. The default is 30 days for acquisition events and 90 days for other conversions, and the window applies to every attribution model. CRMs let you set your own, so treat 7/30/60/90 as a vocabulary, not a law. Google's [attribution settings documentation](https://support.google.com/analytics/answer/10597962) describes the window mechanics; the choice itself is yours. How to choose: match the window to your actual sales cycle, not to a default. A 7-day window will lie to you if your cycle is 45 days: most of your real originating messages will fall outside it and get no credit. A 90-day window will lie to you if your product is an impulse buy: it will hand credit to old touches that had nothing to do with the decision. If you do not know your cycle, measure it before you set the window. Then write the window down next to the rule, and apply both to every deal. The discipline is not in picking the perfect number. It is in picking one number and keeping it. ## A workflow that connects each message to a customer Stated as a loop you can run on every message. 1. **Write the message and give it an ID.** Before you send, tag it with a stable, human-readable message ID, something like `dm-2026-07-26-acme-founders` or `reply-reddit-ops-thread-4`. The ID is the key everything else hangs off. 2. **Generate a tracked link for that message.** The link carries the conversation context (where it was sent, what it was about) so a click later resolves to this exact message, not to "social." 3. **Send the message with that link.** 4. **On click, log it.** The visit and its source context land in your tracking surface. Log a *clicked* activity on the contact, tagged with the message ID, and associate it to the deal (or create the deal if it does not exist yet). 5. **On reply, call, or meeting, log it, to both records.** Every downstream touch is an activity on the contact *and* the deal, tagged with the message ID of the message that started the thread. This is the HubSpot association rule from earlier, and it is the single most common place revenue attribution goes dark. 6. **On close, apply your rule over your window.** First-touch: the message ID on the earliest qualifying activity is your answer. Last-touch: the latest one. Multi-touch: split across all of them. Either way, you can now point at a specific message and say "this one generated this customer." The live version of this loop is where [proactive outreach](https://heyzinc.com/blog/proactive-outreach) earns its keep. The point of logging a click is not just to attribute it later; it is to notice it now, while the visitor is still on the site, and decide whether the next touch is a call, a message, or nothing. Attribution is a byproduct of a good live-activity workflow. It does not have to wait for a quarterly report. ## Where HeyZinc fits HeyZinc is not a CRM, and it does not replace HubSpot, Salesforce, or whatever you already use. What it contributes is the per-message link layer and the live-activity layer that make this whole method usable for founder-led sales. Concretely: HeyZinc [tracked links](https://heyzinc.com/blog/tracked-links-versus-utm-parameters) write a short token to the URL and keep the conversation context (source page, platform, thread) in a server-side record, so a click resolves to one specific message instead of a generic channel. The detail view surfaces visit, session, and engagement metrics, including an active-now state when a visitor is currently on the site. Companion and mobile notifications carry those alerts away from the dashboard, subject to your workspace's notification configuration. The CRM and the rule are yours. HeyZinc makes the link and the moment visible, which is the part founder-led teams consistently lose to "direct" and to delay. The payoff is a shift in when you learn which message worked. You stop finding it in next quarter's attribution export. You start finding it while the visitor is still on the page, which is also when the next message in the chain gets written. Explore [HeyZinc](https://heyzinc.com) if you want the link and live-activity layer wired up. ## Common ways per-message attribution breaks Most per-message attribution does not fail because the model is wrong. It fails because the inputs are missing. - **Shared links.** One link used across ten messages collapses the per-message signal back into a campaign signal. You are back to "Reddit worked," not "this reply worked." - **No message ID on the activity.** The click gets logged, the call gets logged, but neither is tied to the message that caused them. You have a pile of touches and no path through them. - **Activities logged to the contact but not associated to the deal.** The HubSpot rule, again: a call or meeting on the contact alone is invisible to revenue attribution. This is the silent one. - **No window.** Old messages claim new deals, and the numbers drift toward whatever happened to be oldest or newest. - **Defaulting to the source field.** Trusting whatever the CRM stamped without a written rule means trusting a value nobody chose on purpose. - **Vanity source rollups.** "Organic" and "direct" are buckets, not answers. They hide the conversation that actually started the deal. Fix the inputs and almost any reasonable model will give you something you can act on. Leave the inputs broken and no model will save you. ## The short version Per-message attribution is a discipline, not a product. Three building blocks: a tracked link per message, an activity log keyed to a message ID, a documented rule and window. One rule. One window. Written down and kept. Do that, and "which message generated this customer" stops being a quarterly guess. It becomes a field on the deal, populated by the message ID you already attached to the first reply you sent. --- - [More Marketing articles](https://heyzinc.com/blog/category/marketing) - [All articles](https://heyzinc.com/blog)