Cullen vs Sort Feed: both score outliers. Only one tells you why.

What Sort Feed gets right

Sort Feed is the established one in this category, with 80,000 users on the Chrome Web Store, and it earns that. It covers four platforms, it needs no account to start, and it does real baseline-relative work: its outlier score divides a post's views by the median views of up to 25 of that creator's recent posts, throws out anything from the last three days, and only badges a post at 2x or above. That is a genuine method, published openly in its own docs. Its export and download tooling is also better than ours, and we say so below.

Sort Feed facts from sortfeed.com (home, /pricing, /docs/sorting/outlier-score, /docs/platforms/what-works-where, /docs/account-and-help/plans-and-billing) and its Chrome Web Store listing, checked August 2026. The $108 figure is $9 × 12 at Sort Feed's own stated rate; its site prints the monthly figure only. If anything here is out of date, tell us and we will correct it.

A 25-post window measures the account as it is now, not as it was then.

This is the whole argument, and it is a design difference, not a bug on their side. Sort Feed builds one median from up to 25 of a creator's recent posts and divides every post by it. That is a fine reading of what is working lately. It is the wrong instrument for the post you actually want: the one that did 38 times the account's normal numbers back when the account was small, buried 200 places down the likes list, invisible to a recent-window median because the account has since grown past it.

Cullen scans up to 500 posts in one pass and gives every post its own baseline: the median of the posts immediately around it in that account's timeline. A post from the small era is judged against the small era. Because the neighbouring posts all shared roughly the same follower count, the follower-count problem cancels out instead of being smoothed over. Cullen then does one more thing no raw or recent-window score can: it flags the level-up, the rare breakout that did not fade but permanently raised the account's baseline afterward. That is the single most valuable post on any channel, and it has no equivalent in a divide-by-median score.

Ranking is not explaining. Sort Feed's own workflow says so.

Sort Feed's three-step how-it-works ends with the user turning the hooks and topics from their research into scripts. That last step is the actual job, and it is left to you: download the file, run the transcript, stare at it, guess. The word "AI" does not appear on sortfeed.com's homepage at all, which is honest of them and also the point.

Cullen's deep-dive does that step and shows its working. It watches the video, reads the real top comments, and returns a ledger where every claimed trigger carries a verdict and a quote. It will kill its own best-sounding theory when the comments do not back it, which is the part a confident chatbot structurally cannot do.

A finished deep-dive can also be published as a public report page on trycullen.com, showing the post, the breakout number, the baseline it was measured against and the evidence, so the case travels to a client or a collaborator without them installing anything.

The sorting job, at a ninth of the price.

Strip the analysis out and compare like for like: scanning a profile and ranking it by real performance. Sort Feed charges $9 a month billed yearly for that. Cullen charges $12 a year for the same job plus the centered-median breakout math and level-up detection.

The honest gap runs the other way too. Sort Feed exports and Cullen does not. If your workflow is handing a client a spreadsheet of engagement numbers, or bulk-downloading watermark-free video, or transcribing 500 minutes a month, Sort Feed does all three and Cullen does none of them. It also covers Facebook and long-form YouTube, which Cullen deliberately does not.