Which channels produced paying, expanding accounts?

An account that starts at $90 a month can be at $18,000 a month three years later. Attribution that stops at the first invoice describes almost none of the value a channel produced.

Because the match is on the account, every later month of consumption credits the original channel — which is the only ranking that reflects how infrastructure businesses actually compound.

Documentation, open-source presence, conference talks and community programs drive a large share of adoption and are defended almost entirely on faith.

Any list with contact details — event registrations, newsletter signups, community sign-ups — can go into the lead file, tagged with its source. Ranked on consumption revenue, these channels may outperform paid search by a wide margin, which is worth being able to prove.

A developer signs up with a work email on a Tuesday, spends forty dollars a month for half a year, and then somebody in procurement signs a committed spend agreement for a substantial sum. Two systems recorded that: the product, which saw a self-serve signup, and the CRM, which saw an enterprise opportunity with a different contact on it. Neither holds the join, so the marketing report shows either a cheap signup or an expensive enterprise deal depending on which system anybody happened to open.

Exporting both and matching on the contact puts them back together. The committed contract credits the channel that produced the original signup, which is usually a documentation page, a tutorial, a community post or a conference talk rather than a sales campaign. It is common for the bottom-up motion to be funding most of the enterprise revenue while being budgeted as a content cost, and for that to be invisible in every dashboard the company currently runs on a weekly basis.

Where the enterprise contact never self-served and appears in no lead file, the deal stays in Direct / Unknown. Nothing is assumed from a shared email domain, because a domain is not a person and treating it as one would credit a channel for a colleague's click months earlier. A domain is simply not a matching key; the only weaker link the engine produces, the same national number under a different country code, is listed for somebody to accept or reject rather than counted automatically, which is what keeps the attributed figure defensible when the finance team starts checking it line by line.

Consumption businesses have an extreme spread. A hobby project, a funded startup and a media company running a live event are all on the same price list and three orders of magnitude apart in what they pay. Averaging them produces a cost-per-account figure describing nobody, and a channel ranking driven entirely by how many small accounts each source happened to produce that month rather than by what any of them went on to spend. The ranking then tracks account volume, which is close to the thing the business cares about least, and it lurches whenever a free-tier campaign happens to run.

Slicing helps and needs one extra column. Product line, region, plan tier or account size band — whichever your business genuinely differs along — mapped as Product / service, and the report shows consumption revenue for each slice with how much of it traces to a channel. It does not rank channels inside each slice, but it shows where attributed revenue is concentrated and where it is thin, which a single blended figure hides. A channel funded on a blended number may be excellent at one end of the range and weak at the other, and the slice is where that first becomes visible. One column, added once at export, is all it takes.

Region deserves particular mention, because infrastructure demand is geographically lumpy in a way that surprises teams. Data residency requirements, local competitors, latency-sensitive workloads and the language your documentation is written in all push adoption unevenly across markets. A campaign returning nothing in one region and paying for itself twice over in another is invisible inside a global total, and a global total is what almost everyone in this segment is currently reading. Splitting by region is frequently the first cut that makes a marketing plan look wrong in a way somebody can act on.

Technical content is defended on traffic, and traffic is a poor proxy here. A tutorial ranking for a broad term brings tens of thousands of readers who will never deploy anything. A comparison page, a migration guide or a documentation section answering one specific failure mode brings a few hundred readers, and a meaningful share of them become paying accounts, because the person reading it already has the problem the page describes. Traffic ranks the first page far above the second, and consumption revenue ranks them the other way round.

Ranking on consumption revenue rather than sessions reorders that list immediately. It is one of the few measurements that can justify writing fewer and deeper pieces, because it shows where the revenue came from rather than where the pageviews did. It also identifies what is worth maintaining: a guide that produced substantial revenue two years ago and has since gone stale is a concrete loss rather than an item on somebody's housekeeping backlog. It also gives a documentation team the one argument it has never had, which is a revenue figure attached to a page.

The mechanics are ordinary. A list of people who downloaded something, registered for something or arrived through an identifiable campaign becomes the lead file, each row tagged with its source; paid accounts and their consumption become the revenue file; the report joins the two on the person. Historical periods work as well as current ones, because the consumption records were kept for billing, which means a year of content decisions can be assessed without waiting another year to instrument anything first. Nothing is installed on the product, the documentation site or the console, and no permission is granted to anybody.