Who owns the feed, market by market

A market here is a list of named people. We start with the people who reacted to or commented on competitors' posts in the category, filter that pool by title, then rank by how often they engage. The tables use the most active slice of the pool.

Then we take every vendor in the category and count how much of that named market touched their posts. We divide by post count, so a vendor publishing five times a week is not rewarded for volume.

If you sell into one of these categories, follower count answers a different question. The number on these pages is how many of the people who decide in your category reacted to you in 90 days. In revenue tooling that number runs from 13.6 per post at the top to 0 at the bottom, inside the same 170 people.

If you are a founder there is a second number. Across the markets we have pulled, founders and operators out-reach their own company pages by roughly 10x. In ecommerce customer-service AI, two named founders at one vendor reached 40 of the sampled market. Their company page reached 3.

Nordics pull, 27 August 2026. 90 day window. 170 most active of 642 sampled people who engaged the category's posts.

Gong took 327 market reactions in the window, Outreach 264. Attention published 24 posts and took nothing from this market. The most-reacted post in the whole category was craft: Gong on the biggest sales call mistake, 74 market reactions from the 170. And the account this market follows most closely belongs to an account executive at Gong who writes a newsletter, ahead of every CEO in the category. 29 of the 170 engaged him.

Full table: /markets/revenue-tooling

London pull, August 2026. 90 day window. 170 most active of 264 sampled people.

Kustomer takes double the rate of anyone else in the category, and it gets there with profession content: CX Day, certification, awards, community. Gorgias, the best known ecommerce helpdesk brand in the table, sits at 0.8 per post. Siena pulls the largest raw crowd of any vendor here, 444 engagers, and converts it at half Kustomer's rate. The most engaged account in this market belongs to a professional association rather than a vendor. The CX Professionals Association was engaged by 50 of the 170.

Full table: /markets/ecommerce-customer-service-ai

UK pull, 20 August 2026. Eight anchor platforms, 90 days, 4,449 raw reactors. This table is a ratio rather than a rate.

Across all eight platforms, 79 people whose title suggests they sign, and 1,523 builders and practitioners. Every anchor skews the same way and the only difference is by how much. 27 people show up around two or more platforms, so the crowds barely overlap.

Full table: /markets/ai-agent-platforms

These are being built now. We publish no numbers until a pull is finished.

• German HR software
• ERP for the Mittelstand
• B2B payments
• Developer tools
• Cybersecurity

Each one gets the same treatment. Competitor engagers first, then a title filter, then 90 days of counting.

Four markets pulled to date. Five things repeated in all of them.

• Craft beats product news by about 2 to 1. The most-reacted post in a category is usually someone explaining how the work is done.
• Founders and operators out-reach their company pages by roughly 10x.
• The person a market follows is rarely a vendor CEO. So far it has been an account executive with a newsletter and a professional association.
• Every market has an invisible vendor. A real company, real posts, a number at or near zero from the sampled market.
• Markets split into two rooms. In ecommerce customer-service AI, 116 of 264 people were reachable in ecommerce-native rooms and 148 only in the profession's own rooms.

One finding is less comfortable. Not every market has a concentrated voice layer. In our own market, 3,472 named people, the top trusted voice overlapped 4.5 percent of a 200 person sample. That market follows nobody in particular, and we printed it the way it came out.

• A market is the people who engaged competitors' posts in the window, filtered by title.
• Market reactions per post equals reactions and comments from that sampled market on a vendor's public posts, divided by the vendor's posts in the same 90 days.
• The sample and the window are printed on every table.
• Zero is printed as zero, with the vendor's post count next to it.
• A title that looks right is not buying authority. We write the sampled market, never buyers decide.

The long version, including where the method is weak: /data/method