How ad-comment attribution actually works

What the rest of this market shows you

We looked. As of August 2026, ReplyZen's own landing page — under "Why choose ReplyZen" — renders its headline conversion statistic as "0.00% conversions / 100 comments", with a footnote crediting a proprietary tracking pixel. Respondology's marketing site claims a standalone "-47% CPA decrease" for one TikTok campaign, with no time window, sample size, or measurement method attached. Neither is a methodology. Both are a claim, stated once, that nobody outside the vendor can check.

That's not unusual in this market — it's the norm. Every vendor selling "ad comments to revenue" reports a number computed by a system only they can see, using rules only they know. We think a media buyer deciding whether to trust a comment-moderation tool with attribution deserves better than that.

What we publish instead

Every figure on your ad ledger — comments, clicks, orders, revenue — is a live sum over rows your own workspace already owns. There is no separate attribution model running behind it. The full procedure, the exact SQL-shape evidence rules, and every named exclusion are public:

• A human click is one thing, precisely: a visit through our own link router that isn't a platform preview fetch, deduplicated inside a 10-minute window per visitor — the link's own token together with a keyed hash of the IP, never the address itself. The platform's preview crawler — the thing that hits your link the instant a DM sends, to build the link-unfurl card — is recorded too, as proof the message actually delivered, but it never counts as a click.
• An order counts as attributed on exactly two kinds of evidence, in strength order: a discount code minted for one of your links, or a UTM token matched on Shopify's own customer-journey data. Discount-code evidence always wins when both are present.
• Both evidence types are bounded to 7 days from issuance, last click. An order outside that window is not counted, no matter how likely the connection looks.
• The ledger's row is the promoted post, not the individual ad — comments land on the post across every ad that ever drove traffic to it, so a per-ad comment count would be invented, not measured.

What we refuse to count

Same-name matching between a commenter and a customer. Time proximity — "bought within an hour of the reply." Any modelled or "assisted" share of a sale. Any share of Meta's own purchase_roas, which we show verbatim, labelled as Meta's number, and never fold into ours. None of these are evidence, and none of them get added later without this page changing first.

We do not model, estimate, or share attribution — if it is not on this list, it is not in your number.

This is also why our number usually runs smaller than a tool that infers a sale from proximity or a shared name. Theirs is a claim. Ours is a receipt — and every cell on your ledger opens the exact rows that produced it.

Read the complete procedure behind every ad ledger column — the evidence upsert order, the click classifier, and every excluded case — on the full ad ledger attribution methodology page ↗.

Start your free trial to connect Shopify, drop tracked links into your replies and DMs, and see the ledger fill in with rows you can trace one by one.