# Ghost member sources: track signup and paid conversion as separate events A reader can sign up through search and later pay through email. Preserve both events without counting the purchase twice. Published: 2026-11-01 Updated: 2026-11-02 Author: Staff Writers Publisher: Digital Publishing News Earlier version reviewed by Shaun Connell, Managing Editor Canonical: https://digitalpublishingnews.com/articles/member-attribution-is-not-causality/ ## Keep the first signup and the paid conversion separate Signup-source and paid-conversion reports can both credit the same customer. Adding their channel totals as if they describe separate buyers inflates the result. Keep the first signup and the later purchase as different events belonging to one person. Ghost, a publishing and membership platform, records member sources and paid-conversion sources. Attribution means assigning credit according to the tracking information available. It is useful for describing recorded journeys, but it does not prove that removing a channel would remove the same number of purchases. Source: https://ghost.org/help/member-sources/ ## Keep each touchpoint instead of crediting only the last one A reader discovers an article through search, joins the free newsletter and buys from an email a month later. “Search acquired the member” and “email recorded the paid conversion” can both be true. In an illustrative group of ten buyers, six started through search and later bought through email; four signed up and bought from a referral partner. Counting six search wins plus six email wins plus four referral wins would turn ten buyers into sixteen credited conversions. Invented conversion paths; count people once in the total Signup source | Paid-conversion source | Distinct buyers --- | --- | --- Search | Email | 6 Referral partner | Referral partner | 4 Total | | 10 ## Choose which acquisition source to test next The example suggests two different questions: does search bring people who later pay, and does email help those people reach an offer? Last-step attribution alone cannot answer what would happen if you removed either channel. Use the pattern to choose a test, not to declare causal revenue. To build the matrix, you need records that join signup and paid-conversion events to the same member identifier. This is an analysis worksheet, not a claim that every dashboard exposes that export. If you have only aggregate counts or cannot match the records, keep the reports separate and mark the journey unknown. When the join is available, use one row per paying member, retain both sources and event dates, and count the person once in the total. Compare acquisition groups after equal time to convert; older subscribers have had more opportunity to pay. ## Use recorded origins to choose a controlled comparison Suppose two sources attract equally many free members but the first has more paid upgrades. Before changing spend, compare signup dates and the time each member had to upgrade. Check whether one group received a different offer or a different sequence of messages. A source field can tell you where a member arrived and still leave these questions open. For the next comparison, define who is eligible for the offer and keep that population in the calculation even if some never reach the checkout page. Retain both the first signup source and the later upgrade event. That lets you explain why two reports give different totals without attributing the same purchase to multiple channels. ## Estimate what changes if a source is removed Suppose 50 signups in a month are credited to a partner, and ten pay later. That is a 20% observed upgrade rate for those credited signups. It does not tell you how many of the ten would have paid anyway after arriving through another route. To assess that decision, compare comparable readers who were and were not exposed to the partner placement. Define the window and outcome before running the test. Track new paid subscriptions across the whole eligible group even when their last visit came through email or search. If the comparison group differs in offer or timing, report that caveat. The experiment can inform incremental value while the source report continues to describe the reader path. AI production disclosure: https://digitalpublishingnews.com/how-we-use-ai/ ## Sources https://ghost.org/help/member-sources/ Editorial status: AI-assisted revision; preceding edition reviewed by Shaun Connell. Current revision awaits review.