Washington Post's Cost-Per-Story Score, and Its Blind Spot

Amazon’s Playbook Shows Up in a Newsroom Meeting

On March 12, 2026, Jeff Bezos met with Washington Post editors and reporters at his home in Washington, D.C. The morning session covered the company’s finances, the afternoon covered the newsroom’s use of data, and he reportedly took questions over lunch. It was a meeting to discuss the paper’s future in the wake of major layoffs. TheWrap’s on-the-scene report

The New York Times reported that attendees opened the meeting by silently reading memos—a scene straight out of Amazon’s playbook. Bezos has long described how Amazon runs its meetings by reading and discussing six-page narrative documents instead of slide decks. Amazon’s 2017 shareholder letter

What caught my eye in this management shift was the cost of producing a single story and a composite score tallying reader response. The company calls these, respectively, the “story unit cost” and the “audience value score.” Knowing your costs is necessary, but I found myself wondering what might get left out when articles and newsrooms are judged by metrics like these.

Executives Emphasize Losses and Cost-Per-Article

The Wall Street Journal reported that The Washington Post lost $77 million in 2023, roughly $100 million in 2024, and more than $100 million in 2025. That puts total losses over the past three years at close to $300 million. WSJ report

Management pointed to falling article output and reader engagement, even as costs kept rising, as justification for the layoffs.

Acting CEO Jeff Dionofrio, a former CFO, reportedly shared the following figures with staff:

• Article output down 42% compared to 2020
• Newsroom costs up 16% over the same period
• Articles per reporter down 36%
• News and opinion pageviews down 48%
• Some categories now cost thousands of dollars per published article

Dionofrio explained that these shifts had doubled the cost per article since 2020. The company said his remarks were meant to lay out the data behind the layoffs. New York Times report, CJR article including the company’s response

The “cost per story unit” is a calculation of how much a single article costs to produce. If total costs stay the same but the number of articles falls, this figure rises. But a quick breaking-news brief and a months-long investigative piece aren’t comparable units of work. A drop in article count alone can’t capture the full value of what reporters actually do.

Scoring Reader Reactions

There’s something about Bezos’s approach that echoes how Amazon operates: put the discussion in writing, and make people back up their proposals with data. Amazon’s six-page memo¹ is itself a meeting document designed to support this kind of decision-making.

The New York Times reported that former CEO Will Lewis proposed cutting 200 newsroom jobs in November 2024, but Bezos sent the plan back, saying it lacked sufficient data backing. The report also noted that Lewis’s team subsequently assembled a small group to build a data model — and that the newsroom staff were not part of that group. New York Times report

The “Audience Value Score” is a metric that combines readers’ time spent on an article, shares, and subscription conversions into a single score from 0 to 100. According to the report, Brian Flaherty, managing editor for content strategy, described a score above 70 as a good result.

This score can help identify which articles resonated with readers. But if it’s going to be used as a basis for allocating reporting budgets or staff, then how heavily each factor is weighted and over what time window reactions are measured become crucial questions. Based on the publicly disclosed description of the metric alone, it’s hard to tell whether it can even capture an article’s long-term impact.

What Happens Beyond Immediate Costs When You Shrink the Newsroom

According to reports, editor-in-chief Matt Murray and management built their layoff plans by weighing reader demand against reporting costs. Following that, the sports and books desks were eliminated entirely, and local news and foreign bureaus were cut sharply. Nieman Lab’s layoff coverage

The areas cut hardest and the areas management says it will protect reveal where its priorities lie.

• Sports and books: desks eliminated entirely
• Local news and foreign coverage: cut substantially
• Investigative reporting and other core beats: management says it will safeguard these

When I look at the cost of a foreign bureau, I think you also have to weigh the sources and on-the-ground knowledge that bureau has built up over years. The Washington Post was a finalist for the 2025 Pulitzer Prize in International Reporting for its Gaza coverage. Pulitzer Prize record That kind of reporting capacity is hard to rebuild on short notice once you actually need it.

Whether readers are currently reading and sharing something, and whether the newsroom is ready to cover the next big story, are two different questions. Cutting the cost of stationing reporters abroad lowers spending right now, but it can also shrink the newsroom’s ability to verify what’s happening on the ground when something does happen. Even when we invoke the value of reporting like Watergate, the question isn’t some engagement score that didn’t exist back then — it’s whether the newsroom had the conditions to sustain long-term investigative work.

Oswarld’s Lens

I agree that costs need to be managed. What worries me is letting cost-per-article or reader-engagement scores stand in for the question of what makes an article good in the first place.

For a business that keeps losing money, watching operating costs is necessary. Cost-per-article can also be useful for identifying which areas are absorbing the most resources. But a high cost shouldn’t be taken as evidence of waste on its own — it needs to be weighed alongside why that level of reporting was necessary.

The value of an article doesn’t always accrue only to the people who read it. When investigative reporting exposes corruption or leads to a change in safety regulations, citizens who never read the piece can still benefit. This public-good quality² isn’t fully captured by time-on-page or share counts. I think a newspaper needs to judge, separately, whether it can afford to keep funding this kind of reporting.

The same scrutiny should apply to why readers leave. After the outlet withdrew its 2024 presidential endorsement and shifted its opinion-page direction in 2025, reports surfaced of large-scale subscription cancellations. The Washington Post’s own reporting on the cancellations, NPR’s coverage

Separately, after layoffs in February 2026, the New York Times reported that over 60,000 people canceled their subscriptions. The company disputed the figure but reportedly offered no alternative number of its own. TheWrap’s follow-up reporting

If reader data was the justification for cutting articles, I think this kind of exodus should also be treated as evidence for reexamining that management decision. If only the parts of reader response that support the layoffs get emphasized, it’s hard to avoid the charge of using data selectively.

It’s also worth examining whether the owner’s other business interests and political stakes could be shaping the outlet’s decisions. That said, we can’t know Bezos’s true motives for any particular decision. What I want to question isn’t so much his motives as whether the trust he asks readers to extend actually matches the management decisions being made.

Closing

Newsrooms need cost management and performance metrics. But they also have to decide how much support to give reporting that gets little short-term response, and how to preserve the expertise reporters build together over time.

The meeting format Amazon used could help newspapers too. But what counts as “performance” has to be defined according to each newspaper’s own mission. Article volume, subscriber retention, reader trust, and public impact don’t collapse neatly into a single number.

That’s why, whenever I hear the phrase “data-driven decision,” I try to check both what was measured and what was left out. You can only judge whether a cost cut actually made a newspaper better by also looking at what happened afterward to subscriptions and reporting capacity.

Whoever sets the metrics also bears responsibility for explaining them. I think data can only serve as a tool for reviewing management decisions if the decision-makers also spell out, in advance, what results would make them change course.

References & Further Reading

• Benjamin Mullin·Katie Robertson, How Jeff Bezos Upended The Washington Post, The New York Times, 2026.03.14: reporting on the management intervention and internal metrics covered in this piece.
• Alexandra Bruell, Washington Post Losses Topped $100 Million in 2025, WSJ, 2026.02: covers year-by-year loss figures.
• Hamilton Nolan·Siddhartha Mahanta, The Washington Post Enters Its Amazon Era, CJR, 2026.03.23: analysis of the cost-per-story metric along with the company’s response.
• Michael Calderone, The Washington Post Meeting Held at Bezos’s Home, TheWrap, 2026.03.13: the meeting took place on March 12, the day before the story broke.
• Hanaa’ Tameez, Washington Post Layoffs and the Union’s Preliminary Tally, Nieman Lab, 2026.02.13.
• Amazon’s 2017 Shareholder Letter, Working Backwards (Korean title: Sunseo Pagoe): sources describing Amazon’s narrative-memo meeting documents and operating style.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. 6-Page Memo: a narrative document laying out the problem to be discussed in a meeting along with supporting evidence. Amazon has used a format in which attendees read it silently at the start of the meeting before discussing it. ↩

2. Public Good: a good where one person’s use doesn’t diminish another’s share, and it’s difficult to exclude people who haven’t paid for it. While a paywalled article itself has restricted access, the social benefit generated by the reporting can spread beyond its readers. ↩