Field Note 05: The Wrong Specification
Summary: Reaching the right reader is not the same as serving them well. Three decades of measuring clicks, views and dwell-time has told newsrooms what they took from readers, never what readers needed. This issue sets out three things a newsroom needs to know about a reader before it can offer something worth paying for — what that reader values, what has newly opened in their life, and where their readiness stands — and why most newsrooms can answer only one of the three.
What readers want has never really counted in digital journalism. I say counted because measurement started 30 years ago with server logs and page views. Around the same time, in the mid 1990s, the “ad impression” came out of print. Cost per thousand impressions (CPM) imported the circulation currency of newspapers and broadcast wholesale. By 2000, when Google launched AdWords, clicks had monetary value. The industry has chased views, dwell-time, and clicks ever since, leaving human needs unmeasured.
Those numbers came from selling an audience to an advertiser. They are precise about what the industry takes from a reader and silent on what it gives back. They record that a piece was opened, never why, or what in it mattered. The result is journalism that is accurate, well-sourced, and clearly written, but built to the wrong specification: a long read for someone who needed an update, or a push notification for someone who needed silence. To count what we give rather than what we take, we have to know what the reader came for. It is a question about the reader, not the output, and it has three parts. Most newsrooms can answer one, perhaps two.
Values
Values come first. Not preferences, which are surface level, but the orientations underneath. A commitment to community or concern about the environment holds across topics and years. At a regional media group in Germany, we mapped readers by values using the SINUS Milieus framework. The demographic lines came apart. A 32-year-old in a city sat closer to a 60-year-old of the same milieu in a village half an hour away than to someone the same age in the next building with the same income. Demographics describe the data of a life. Values describe what a life is being used for.
Interests
Values hold still over time, which is also their limit. They will not explain why a reader who has never opened a health story spends this week reading about oncology. The reader’s values have not moved; what has moved are the issues that have newly opened. A diagnosis does not add a topic to a list the way a hobby does, and neither does a pregnancy — each puts the reader on a road not travelled yesterday. Interests open, often overnight, and close again once the issue resolves.
Mindset and Stage
The third dimension is mindset or situational stage. A newly opened interest starts the reader often at the beginning of that road, not somewhere along it. Behaviour-change research offers, for example, a continuum for health-related concerns, which I describe here as: unaware, aware, worried without yet a name for it, decided to act, going through it, living with it, out the other side. A reader stands in different places on different subjects at the same hour. Years along on climate, not yet aware of the health question that will matter more.
Values can be measured directly because a reader can be asked. Interests and stage are harder to see. By the time a behaviour shows up (a share, a like, a return visit), it is a record of what was taken, not a signal freely given. It tells you attention moved. It does not tell you why, or where on the continuum the reader stood when it did. These proxies are the best most newsrooms have today, and they are worth using, but they inherit the same limitation as the page view: precise about the action, silent on the need behind it. Building a genuine signal for interest and readiness, something closer to what values already have in tools like SINUS, is the harder, unfinished part of this work.
Get the values right but the stage of interest wrong, and the journalism does not reach its potential. Offer a reader who has not yet named the worry a list of ten things to do about a condition that cannot yet be said out loud, and the story gets closed rather than read. What was needed at that point on the continuum was context without a demand, and information without pressure. The facts, the expertise, and the research are often already in the newsroom. It takes a few more questions in the interview, decided at the planning stage, then a different assembly: different framing, different entry point, different need. A version for the reader who has not named it yet, a version for the reader who has decided to act.
What We Can’t Measure Yet
That solves supply. It does not solve delivery. Producing both versions is a planning decision. Knowing which reader should receive which version is still the harder, unmeasured question. It is the same gap the proxies above leave open. That gap is not a reason to wait. A newsroom that has both versions ready and routes them by imperfect signal (a share, a return visit, an editor’s judgement) is still closer to the reader than one that had only ever written the single, average version in the first place. The measurement problem is real. It is not an excuse to keep producing for nobody in particular while it gets solved.
Why This Is Worth Paying For
Why would anyone want to pay for this? Under advertising, the answer could be deferred because the attention was the product and the reader was the inventory. Under subscription, there is nothing else to sell. The reader has to be given something of value, which means somebody has to know what brought that reader there, and why. A newsroom does not need to solve that measurement problem before it starts building the versions the reader will eventually need.
FROM THE FIELDS – TWO THINGS WORTH READING
Mistrust as a precondition for critical thinking and media literacy.
The verification skill everyone credits to scepticism is more closely tied to reading the news than to distrust of it. A study of 2,115 Australian adults, published in Information, Communication & Society in 2026, tested what people could actually do, not just how capable they said they felt. The insight: mistrust — pausing to question a source rather than rejecting it outright — was the strongest predictor of how confident people felt. But confidence and ability told different stories. Regular news consumers were the ones who could actually tell true from false; mistrust predicted how sure people were, not how right they were.
This is the argument of this issue arriving from the reader’s side of the screen. A dashboard that counts what a reader did tells you as little about what was gained as a reader’s own confidence tells you about what that reader can actually do. The finding worth keeping: the ability to tell true from false was closely tied to the habit of reading good journalism, not to distrust of it. That is not a civic nicety a newsroom provides at a loss, but a strong case for the product — the thing a subscription is for
Is AI Causing Journalists to “Deskill”?
Journalists worry that using AI to think for them is gradually eroding the ability to think at all. Shangyuan Wu of the National University of Singapore interviewed fourteen journalists and editors — print, broadcast, online, newswire — about what AI use is doing to their craft, and published the results in Journalism Practice in June 2026. The list of skills they fear losing is not trivial: research, writing, note-taking, analysis, interviewing, and beneath all of them the habit of deciding for themselves what the story is. One put it plainly — gather the facts, work out what matters, what to pursue, what to discard, and “if you get the machine to do all that for you, then you don’t think anymore.” What they said they must protect maps almost exactly onto the human layer: discernment, a voice on the page, storytelling, original ideation, the judgement AI cannot supply.
This is the practitioner-side mirror of what my upcoming book argues about the two-layer architecture. The machine is fluent at the distribution layer and empty at the layer above it, where a human decides what is worth knowing and why. Wu’s journalists have felt the erosion the DIKNW hierarchy (Data – Information – Knowledge – Narrative – Wisdom) predicts: offload the judgement and the capacity for judgement goes with it. One of them described good writing as having “rhythm and melody” — the reason the narrative layer has to stay human, said better by a working reporter than by any framework.
Read an audience by what it values, by what has opened for it, and by where it stands, and an undifferentiated mass of readers resolves into communities with shared needs. Those communities are content tribes, and they are the subject of the next issue, the model that turns understanding an audience into a structure for serving it, and a thing you can sell. Until then, a question for your own operation: of the numbers on your newsroom dashboard this morning, how many record what you took from a reader, and how many record what you gave her? Hit reply and tell me.
Until next time,
Dietmar
The Field Note: Twice a month, for editors, publishers, and heads of product working out what AI means for their newsrooms. One piece of original analysis. Two things worth a senior reader’s time.