A chat without a phone number or an email is a conversation, not a lead

A chat counts toward revenue only if it carries a phone number or email and the visitor's original source. Most chat setups capture neither reliably: visitors stay anonymous and the chat records the page, not the ad. Ask for one contact detail, keep the original source on the chat, export chats as leads, and report the rest as unknown.

Ask a sales team where their warmest leads come from and somebody will mention the chat. The visitor who asks about financing at 9 p.m., the one who wants to know whether you cover their zip code, the one who books a consultation from inside the widget. Then open the attribution report and look for chat. It is usually a thin line, or missing altogether.

That is not because chat does not work. It is because chat is the one intake route most businesses set up for conversation and never for measurement. A form insists on an email before it will submit. A call arrives with a phone number whether the caller likes it or not. A chat can run for twenty minutes, answer every question, send a booking link and end without anybody knowing who was on the other side.

Why chat leads disappear from attribution

Chat leads go missing in four predictable places: the visitor never leaves a contact detail, the chat tool records the page but not the source, a salesperson retypes the conversation into the CRM without its origin, or the chats are never exported into the lead file. Each one breaks the link between a conversation and a sale.

Only the first row is truly unrecoverable after the fact. The other three are configuration and habit, which means they can be fixed this month. The rest of this post takes them in order, and then deals with the question everyone argues about: whether to make visitors identify themselves before they can chat at all.

The anonymous chat: no contact detail, no match

A chat with no phone number and no email cannot be matched to a closed sale, because sales records are keyed on the customer's contact details. Most chat tools create a record the moment a visitor types, long before they know who it is, so a busy chat inbox can be full of leads nobody will ever trace.

"Blue Cloud from New York" is a perfectly good name for a support conversation and a useless one for attribution. When Blue Cloud buys a $9,000 kitchen remodel three weeks later, the invoice says Maria Alvarez, a phone number and an email. Nothing in the chat record connects to any of the three. We set out the arithmetic of this in our post on match rates: an intake route that does not capture an identifier puts a ceiling on how much revenue can ever be traced, and no amount of clever matching lifts it.

Here is the opinion part. A chatbot that answers ten questions and never asks for a phone number or an email is not a lead-generation tool, whatever the vendor's homepage says. It is a help widget with a marketing budget. If the bot is good enough to qualify somebody, it is good enough to ask how to reach them, and the moment to ask is right after it has been useful, when the visitor has a reason to say yes.

A chat that never captured a phone number or an email can be counted, but it cannot be traced. The sale it led to will be real and unattributable, and no tool can recover the link afterward.

The chat that remembers the page but not the ad

Even when a chat captures a contact detail, it often records only the page where the conversation started. The ad, campaign and click ID were in the URL of the page the visitor landed on, sometimes several pages earlier, and they are gone unless something carried them forward and handed them to the chat.

Picture the path. Somebody clicks a Google ad and lands on /water-heaters?utm_source=google&utm_campaign=tankless-spring. They read, click through to /pricing, then to /financing, and open the chat there. The chat record says the conversation started on /financing. The ad that paid for the visit was two pages ago, and the chat tool never saw it.

The fix is a small piece of plumbing, and it is worth doing properly:

1. Capture on landing. On the first page of a visit, read the UTM parameters and the click ID (gclid, fbclid, msclkid) from the URL and store them in first-party storage for the session.
2. Pass them to the chat. When a conversation starts, set them as custom attributes on the chat contact. Many chat tools accept custom data from the page; Intercom, for example, records custom data for a visitor once they start a conversation and become a lead.
3. Keep them on the export. Check that those attributes come out as columns when you export conversations. An attribute you can see in the chat sidebar but not in the export does nothing for attribution.
4. Write the source down plainly. In the lead file, the source column for a chat that came from a Google ad should say something like "Google Ads" and the campaign column should carry the campaign, exactly as you want them to appear in the report.

Query-parameter targeting is a blunt but useful tool. A separate chatflow for visitors arriving with a given campaign parameter means the conversation knows where it came from without any custom code. It only matches pages whose URL still carries the parameter, usually the landing page, so it complements carrying the source forward rather than replacing it.

The handoff: when a salesperson retypes the lead

The most valuable chats are the ones that turn into a phone call or a meeting, and those are the ones most likely to lose their source. A rep takes the conversation off the widget, creates a new record in the CRM by hand, and the source becomes "Phone", "Sales" or blank.

This is not carelessness. The rep's job is to close, and the chat tool and the CRM are often different systems with no shared field for "where did this person come from". The result is that the strongest chats are credited to the phone line, or to nobody, and chat looks weakest exactly where it did the most work.

Two rules stop it:

• The chat row stays in the lead file. Whatever the rep creates in the CRM, the original chat, with its date, contact detail and source, is exported as a lead in its own right. When the same person later appears as a call or a CRM record, the files share a phone or email and the earlier chat is still there to claim first touch. Our post on duplicate leads explains why one buyer becoming three records is normal and how to credit the sale once.
• The handoff copies the source. If your CRM has a lead source field, the rep fills it with "Chat" and the campaign from the chat record, not with the route the conversation continued on. A dropdown with no default helps; a free-text field invites "called in".

If the chat and the CRM disagree about the source, the chat is usually right about where the person came from and the CRM is right about what happened next. Attribution needs the first of those, which is why the chat export matters even when the CRM looks complete. The same reasoning is behind exporting a sales file that will actually match: every system keeps the field it needs, and nobody else's.

Should you force a pre-chat form?

A required pre-chat form guarantees a contact detail on every chat and costs you the visitors who will not fill it in. Asking for one contact detail after the first useful answer usually keeps most of the conversations and most of the measurability. The right answer is a trade-off you can count, not a matter of taste.

Here is the trade worked through for one month of chat on a home-services site. The capture rates and chat volumes are illustrative assumptions, not benchmarks; the arithmetic is what matters, and you should rerun it with your own numbers. Assume 500 visitors would start a chat if nothing stood in the way, 8% of chatters eventually buy, and the average sale is $2,400.

The arithmetic. In setup A, 500 × 15% = 75 traceable chats, 75 × 8% = 6 traced sales, and 6 × $2,400 = $14,400. All 500 chats produce 500 × 8% = 40 sales worth 40 × $2,400 = $96,000, so 40 − 6 = 34 sales, $81,600, happened and cannot be traced. In setup B, 500 × 55% = 275 traceable chats, 275 × 8% = 22 sales and 22 × $2,400 = $52,800, leaving 40 − 22 = 18 sales, $43,200, untraced. In setup C the form turns 200 visitors away, and the 300 who chat produce 300 × 8% = 24 sales, $57,600, every one traceable.

Read the table the way a finance director would. Setup A is the most popular configuration and the least honest-looking: chat appears to bring in $14,400 a month when it is involved in $96,000. Setup C looks like the winner on traced revenue, but only because it measured everybody it did not turn away; whether those 200 visitors called instead or left for a competitor is exactly what the form makes impossible to know. Setup B gives up $4,800 of traced revenue against C and keeps every conversation.

Two honest caveats. Visitors who leave a contact detail are probably more serious than those who do not, so they likely close at a higher rate than 8% and setup A undercounts by less than the table shows. Not by enough to change the conclusion. And if your chat is mainly support for existing customers, a gate costs little and the whole calculation matters less; this is a decision for chats that carry new demand.

Ask for one contact detail after the chat has been useful, not before it has said anything. It keeps most of the conversations and most of the measurability, and it is the only setup in which both numbers can be counted.

Ask for the minimum: a phone number or an email, not both, and not a job title. You are collecting personal data about people who have not bought anything yet, and data minimization is a principle worth keeping even where no regulator is watching. One identifier is enough to match a sale; a form that asks for five is a form that gets abandoned.

Chats that start in an ad: Messenger, WhatsApp and Instagram

Click-to-message ads open a conversation in Messenger, WhatsApp or Instagram instead of a page on your site. The ad platform reports conversations started, and the messaging platform can tell your inbox which ad a thread came from, but neither is a sale, and the ad details only reach your lead file if somebody stores them.

Fifty appointments from 1,000 conversations is 50 ÷ 1,000 = 5%, before anybody has bought anything. A cost per conversation started is therefore a cost per expression of interest: if 5% of conversations become appointments and fewer still become sales, each sale costs at least 1 ÷ 5% = 20 times what a conversation does. Comparing cost per conversation with another channel's cost per sale flatters the messaging ad by at least that factor. We made the same argument about instant forms in our post on Meta lead ads: what the platform counts and what finance counts are different events.

So the information exists, which is more than a website chat can usually say. But it lives in an event stream. Unless a workflow writes the ad or campaign onto the contact, and the contact export includes it, the conversation lands in your lead file as "WhatsApp" with no ad behind it. If your messaging runs through a texting platform such as Podium, the same rule applies: export the conversations as leads, with the phone number and a source column, and the sale can be traced. Our post on SMS attribution covers the inbound-text side of the same problem.

How to measure chat-sourced revenue honestly

Chat-sourced revenue is the closed revenue whose earliest matching lead is a chat that carried a phone number or email and a recorded source. Report it next to the number of chats that could not be traced, and never count a chat opening, a bot session or a conversation started as revenue.

• Export chats into the lead file. One row per chat that left a contact detail: date, phone, email, and a source column that says "Chat" plus the channel that brought the visitor, such as "Chat — Google Ads". Chats without a contact detail do not belong in the file, because they cannot match anything.
• Count the untraceable chats separately. The number of chats with no contact detail is your coverage gap. Put it beside chat revenue every month, because a rising gap explains a falling chat line better than any theory about the channel.
• Keep chat openings out of ad-platform conversions. A chat-widget open counted as a conversion teaches bidding to find people who click bubbles. Our post on Performance Max describes the same pattern: chat-widget opens and page views sitting beside real inquiries in the conversion list.
• Decide first touch or last touch before you look. Chat is disproportionately the last contact before a sale, because people who have already decided use it to ask the final question. Under last touch it will collect credit for demand other channels created; the trade-offs are in our post on attribution models.
• Leave the unknown alone. The sales that came from anonymous chats are real, and they will sit in Direct / Unknown. Do not spread that bucket across channels to make chat look bigger; spreading inflates the channels you can already see, as we showed for unattributed revenue.

CloseRev reads exported files and has no chat integration, so a chat counts when it is a row in your lead file, with its date, the phone or email the visitor gave and a source such as "Chat — Google Ads" in the source column, which it reports exactly as written. It credits each sale to the earliest lead sharing its normalized phone or email; a chat with neither cannot be matched, and its sale stays in Direct / Unknown.

Start with this week's export

The quickest test of your chat measurement is to export last month's conversations and count three things: how many left a phone or email, how many carry a source beyond the page URL, and how many of those people appear in your sales file. Those three numbers tell you which of the fixes above to do first.

If fewer than half the chats left a contact detail, change when the chat asks. If the contacts are there but the sources are not, carry the UTM parameters forward. If both are there and the chats still never appear in your attribution, they were never exported into the lead file, and that is an afternoon's work. Make sure the phone numbers come out in a consistent format; our guide to phone normalization explains why that matters more than it sounds.

None of this requires a new tool or a bigger budget. It requires deciding that a chat is a lead only when it can be traced, and setting up the widget so that most of them can be.

Chat is not unmeasurable; most chat setups simply never ask who they are talking to or where that person came from. Ask for one contact detail, carry the source into the chat, export the chats as leads, and report what is left as unknown.