The sale that closed in Messenger never reaches your analytics

“Available pa po?” The message arrives under a product photo you posted on Facebook that morning. Your staff reply, agree on two units and a delivery address in Cainta, and send the GCash number. Ten minutes later a screenshot arrives, the parcel goes out with the courier and the sale is done. Open Google Analytics for that day and it shows a page view, maybe two. If you have tried to track sales from Messenger with a standard setup, you already know that the report and the books never agree.

No setting fixes this. The tools are looking in one place and the buying happens in another.

Where the sale happens

Chat-led selling has a shape. The first contact is a Messenger thread, a Viber message or a comment under a post. The price is confirmed in chat, sometimes negotiated. Payment is a GCash or Maya transfer, a bank transfer, or cash on delivery. Proof of payment is a screenshot. The order is written down wherever the person answering the chat writes things: a notebook, a Google Sheet, a group chat with the warehouse.

The website, if the customer visited it at all, was a brochure: a price list or a branch address checked on a phone before going back to the chat. Some customers never visited.

Web analytics was built for a different path: land, browse, add to basket, pay on the same site. Every step leaves a mark the tool can count. In the chat-led version, the only steps on the website are the ones before the conversation starts. Everything that matters happens afterwards, in apps that do not report back.

Why the report and the books disagree

The analytics report and the sales ledger measure two different things, and neither one is wrong.

Analytics counts behaviour on a property you control: pages viewed, buttons tapped, forms submitted. It stops at the edge of your site. Messenger belongs to Meta and the bank app belongs to the bank. Neither writes a conversion into your dashboard when money moves. A screenshot of a transfer carries nothing a tool could join to a visit.

The books, on the other hand, have to be right. The BIR expects an invoice or receipt for the sale however it was agreed, and the money either arrived or it did not. Nobody audits your analytics. So when the two disagree, the ledger wins by default and the report starts to look like decoration.

The damage comes from reading the wrong document for the question. A marketing lead sees zero conversions and concludes the site does nothing. An owner sees a good month in the books and cuts the Facebook budget that was feeding the chat. Both decisions feel evidence-based, and both rest on a partial view.

The view stays partial because the site, the ads and the chat are usually run by three different people, and joining their numbers is nobody’s job.

What you can attribute honestly

You cannot make analytics see the sale. You can make it see the handover, and build a small habit on the other side that records the outcome.

• Count the handover as the conversion. The last thing your site can observe is the tap on the Messenger, Viber or WhatsApp button. Track that tap as an event, with the page it happened on. “Enquiries started” is a real number, and it shows which pages send people into the chat.
• Tag the links you post. Every link you share on Facebook, TikTok, a Shopee or Lazada shop page or an email can carry a source parameter. The visit then arrives with a label before the handover, so enquiries can be split by channel.
• Ask at the start of the chat. An automated first reply with one question, “Saan niyo po kami nakita?”, and a few options to tap. Customers answer imperfectly, but a human answer is more honest than a guess.
• Prefer a payment link to a screenshot. A PayMongo or Maya payment link produces a record with a timestamp and a reference. A screenshot produces a picture. Where a link is practical, the payment can be matched to an order without retyping.
• Keep one order log. A sheet or a light CRM where the person closing the sale writes date, source, product, amount and payment method as the payment lands. This is the join between the chat and the books, and the only place the whole path exists.
• Read Meta’s own numbers for what they are. A click-to-Messenger ad reports conversations started. Useful, but not orders, and not to be added to the analytics count as if it were.

What cannot be tracked, and what not to pretend

Some gaps are permanent, and it is better to name them than to paper over them.

Your site cannot see inside a chat. Once the customer is in Messenger, the negotiation and the decision are invisible to any script on your page. Someone who read your price list and then paid from the bank app on the same phone has crossed two apps that have no reason to talk to each other.

Screenshots cannot be attributed. There is no reliable way to tie a GCash image in a chat to a website session. Typing those sales into analytics by hand afterwards corrupts the one report that was at least consistently incomplete. Keep the honest count in the log.

Ecommerce benchmarks do not apply. A site whose job is to start conversations shows a conversion rate near zero, because it is being measured on a checkout it does not have. The rates that matter are enquiries per visit, then orders per enquiry from your log.

The log itself is personal data under the Data Privacy Act, so record only the fields you reconcile on, never whole chat transcripts, and limit who can open it.

How to start counting

Start with ownership. Pick one person who owns the order log and agree the five fields it records. Ask whoever maintains the website to track the chat button as an event, which is a small change. Add the “where did you find us” question to the first automated reply. Then set one date a month when the log, the GCash and bank statements and the analytics enquiry count are read together by the people who run the site, the ads and the chat. Expect the first reading to show mostly disagreement. Keep going until the three numbers line up into one funnel: visits, enquiries, orders. That is the report worth deciding on.