Who shaped the conversation

Dot matrix of fifteen creators across the years 2021 to 2026, dot area showing citations per year. Mark Fleming-Williams dominates 2021 and 2022 then fades. Jason DeRise appears in 2023 and peaks at fifteen citations in 2025. Only the two orange rows, Auren Hoffman and Ben Lorica, have a dot in every year. — Fifteen creators, six years, dot area proportional to citations. Eras are visible as diagonal weight: the podcast era at the top left, the newsletter era at the bottom right. Only the two orange rows run unbroken.

Eight of ten, gone

Here are the ten most-cited creators of 2021 next to the ten most-cited of the last two years. The overlap is two names.

Most cited, 2021	Most cited, 2025 to 2026
Mark Fleming-Williams	Jason DeRise
Auren Hoffman	Ben Lorica
Ben Lorica	Dan Entrup
McKinsey	Matthew Bernath
Data Transformers Podcast	Seattle Data Guy
Data Science Central	Peter Baumann
Hedgeweek	Dan Evans
Deloitte	Auren Hoffman
Tobias Macey	Sven Balnojan
Explorium	Benn Stancil

Everyone else at the top of the 2021 list has left the top of mine, and most have left the list entirely. That is not a knock on any of them. It is what a conversation looks like when the subject underneath it changes.

Three eras, three mediums

2021 and 2022 belonged to the podcast. Mark Fleming-Williams was my most cited creator both years, and it was not close: his Alternative Data Podcast was the place this industry explained itself out loud. Podcasts were 23% of everything I cited in 2021. Alongside him sat the institutions: McKinsey, Deloitte, Hedgeweek. The conversation ran through firms and shows.

2023 was the turnover year. ChatGPT rearranged the subject, and the list rearranged with it. Sven Balnojan, Seattle Data Guy and Jason DeRise took the top three slots. DeRise is the sharpest case: he does not appear in the corpus at all before 2023, then leads it in 2024, then posts 15 citations in 2025, the highest single-year count of anyone in six years.

2024 to now belongs to the newsletter. Individual writers publishing on their own mastheads are 31% of everything I cite this year, up from 9% in 2022. Podcasts are down to 3%. The unit of influence in this industry is currently one person, a publishing cadence, and a point of view.

The institutions left the room

Look at the 2021 list again: four of the ten are firms or platforms, not people. McKinsey alone earned 16 citations across five years. In the current top ten there are no institutions at all, and McKinsey has not appeared in the corpus this year.

I do not think the firms got worse. I think the format did. When the questions change every quarter, a byline with a Tuesday deadline beats a research department with a publication calendar. The people writing weekly earned the authority the decks used to hold.

The list refuses to settle

The churn is not history. It is happening now. The most-cited voice of 2026 so far is Dan Entrup, who entered the corpus in 2024. Behind him: Asymmetrix, Joe Reis, Jody Hesch, Matt Robinson, Dan Evans. Meanwhile Jason DeRise, the runaway leader of 2025, sits at four citations this year. One more turn of the wheel, in real time.

If the pattern holds, some of the names that will dominate the 2027 edition of this piece have not published their first issue yet.

The two constants

Ben Lorica and Auren Hoffman appear in every single year of the corpus, 2021 through 2026. Nobody else does.

What they share is instructive. Both changed their subject matter substantially across the period, Lorica from data engineering toward AI systems, Hoffman across the whole data economy, and neither changed their cadence. The lesson from five years of watching this list is not that consistency wins, and not that reinvention wins. It is that consistency of rhythm plus reinvention of subject is the only combination that survives every era.

Fixed voice, moving target. Nothing else survived all six years.

Method, and what this is not

This measures who I cited, one curator, weekly, for five years. It is a proxy for the industry's conversation, not a census of it. About 83% of citations carry a nameable creator; the rest are unattributed or institutional pages. Creators are resolved by merging name variants and by publication domain, and that resolution is slightly tighter than the one behind the August piece, so a few counts differ by small amounts from the leaderboard printed there. Same corpus, sharper merge, two more issues. 2026 figures are year-to-date through August 14.