The founder owns the number. The contractor who made it pretty owns nothing, unless the brief and the payment put the claim in their name and they can stand behind it.
If the person who draws the chart cannot answer how we know this is right and what would falsify it, the chart is decoration and the founder still owns the lie.
Why these three models
The decision is who is accountable for the number on the homepage, the pitch, or the monthly update. The features that fire are hidden action on a two-task job, a finished artefact misread as a checked claim, and a rare failure that should skip the design review.
Three lenses, three outcome types. Hidden action gives an equilibrium answer about what the brief bought. Feedback misperception gives a complex answer about why a finished chart is treated as evidence. The algedonic signal gives a random answer about what to do the moment the two questions fail. That span is the point.
Cairo’s service-versus-autonomy finding is organisational, not a model. Service desks wait for orders and decorate. Autonomous desks originate, report, and stand behind the piece. It is the newsroom fact this ensemble is pointed at. It is not a fourth card, and it is not a registry ID.
Behaviour and governance are folded rather than shipped. The behavioural error is the misread of the finished chart, which the second card already carries. Governance is the brief, which is the first card’s lever. Neither adds a move the three do not already give you.
1. The brief bought pretty, so pretty is what you got
Start with the contract, because the contractor is doing the job you paid for.
The drawer has two tasks. One is to produce a chart that looks finished. The other is to stand behind the claim the chart makes. You can see the first on arrival. You cannot see the second unless you ask, and most briefs never ask.
When an agent has several duties, pay on the one you can measure reallocates attention away from the ones you cannot. Holmstrom and Milgrom put it as a result: incentive pay directs attention among duties, and the case for paying on any one activity weakens as soon as a competing activity is hard to measure.5 A piece rate for output, when quality is not observed, raises volume at the expense of quality. The same arithmetic sits inside a WhatsApp brief that says make the GMV slide pop, and an invoice that bills design hours.
They are trapped. A different person in the same seat, given the same brief and the same invoice, would also ship a finished-looking chart and leave the source, the window, and the denominator for someone else. The lever is the brief, not a pep talk.
This is the service desk Cairo documented in elite newsrooms, and then watched them leave. In the past, news graphics were made mainly by visual designers and artists who took orders from reporters and editors. Not anymore, at least at elite publications: the people who produce the graphics are content creators, and they are autonomous.1 They have moved from sitting at their computers waiting to be told what to do, to content ownership. They collaborate. In many cases they pitch their own pieces and complete every stage, from the initial reporting to the final art.2
Scott Klein, at ProPublica, put the old arrangement in one line: in the old days there was the art department, you sent them a story, and you said make a chart for me.3
That is still the default brief in a small company. A founder exports a number, drops it in a chat, and asks for a slide. The person who draws it cannot reconstruct the mobile-money receipts. They were not in the room when the window was chosen. They will not be on the call when an investor asks how the figure was built. Paying for pretty reallocates attention away from the part you cannot see.
The repair is job design, which is the other half of the same paper. If verification cannot be measured in the same invoice as polish, stop paying as if it can. Either the drawer is also the reporter, paid to originate the number and to stand behind it, or verification is a named second job with a named owner. A fixed fee for a finished look, with no ownership attached, is the contract that produces decoration. It is working.
Moral hazard, the brief lens
- Assumes: the drawer has two tasks, finish a looking-good chart and verify the claim, and you observe only the first.
- Fits because: the brief and the invoice pay for pretty.
- Breaks when: the drawer is also the reporter, or you can observe verification directly.
- Evidence: grade A. Holmstrom and Milgrom 1991, multitask substitution.
- Counteracts: blaming the contractor for the job you bought.
- May reinforce: underpaying verification that should be a separate job.
2. The finished chart is being read as a process
The second lens explains why the first arrangement survives contact with smart people.
Operators infer a hidden process from a visible form. A chart that looks finished is treated as a number that was checked. The inference is cheap, it is almost automatic, and it is the same class of error as reading a local output and concluding you understand the system that produced it.
The New York Times ran the visual version in public. Data-driven graphics on the site were richly interactive, with buttons and menus. Designers inferred use from the existence of the controls. Then the desk watched. Most readers did not interact much, beyond a next button or scrolling up and down. From 2014 the work got more linear. A large share of traffic, sometimes as high as 80 percent, was arriving on phones.4 Form followed observed use, once someone looked. Until then, the buttons had been taken as proof that interaction was happening.
Len De Groot, at the Los Angeles Times, named the shift: the audience may not have changed; the desk started to care about the audience, and that is the change.4
Your version is quieter and more expensive. The homepage tile, the pitch chart, the monthly update all look done. Nobody asked how the window was cut. Nobody asked what would make the number false. The artefact is being read as a process.
Hiring for graphic literacy misses it. Highly educated adults fail to infer the behaviour of simple stock-and-flow systems, and the failure is not attributable to an inability to interpret graphs, to missing context, to low motivation, or to short cognitive capacity.7 A more senior visual hire, or a course, does not touch it. In the Beer Distribution Game, people with good local information still manufacture large, costly swings, then attribute the result to external causes that did not occur.6 A contractor who delivered a clean slide will, if the number later breaks, point at the export they were given. They will be telling the truth as they lived it. The local output was the slide. The hidden process was never theirs.
Cairo is a 2017 draft dissertation at the Universitat Oberta de Catalunya, updated 5 June 2017, with corrections still marked as remaining. Treat it as that, not as a finished press book. The NYT prune is interview evidence from elite English-speaking newsrooms. He says so. The piece that travels is smaller: do not infer a checked claim from a finished look.
Feedback misperception, the form lens
- Assumes: operators infer a hidden process from a visible form.
- Fits because: a finished chart is treated as a checked claim.
- Breaks when: the number is reconstructed independently every time it ships, or use is observed rather than inferred.
- Evidence: grade A on the misread. The NYT prune is elite-newsroom interview evidence.
- Counteracts: hiring a more senior designer for a verification failure.
- May reinforce: treating every decorated number as fraud.
3. The unanswered pair is the alarm
The third lens is not a design for midnight calls. That piece already exists. The alarm in this article is narrower.
Ask the person who drew the chart two questions, in writing, before it ships.
How do we know this is right.
What would make it false.
If they cannot answer, stop the pretty-review. Do not send the chart back for another pass on colour or type. The unanswered pair is the signal. It is non-analytic. It does not diagnose the export or the window. It says only: this claim has no owner, and it should not ship.
Beer called a signal of that kind algedonic: pertaining to regulation in a non-analytical mode; raising alarm.8 Its job is to wake the person who will be blamed, not to improve the slide. Analysis is what you do after the chart is pulled, with the person who can reconstruct the number.
The unanswered pair is the alarm. It skips the design review.
Rarity is the discipline that keeps it an alarm. If every chart fails the two questions, you do not have a tripwire. You have a process that has never required ownership, and the right move is to stop shipping public numbers until one person can answer. An alarm that fires on every artefact habituates, and then it is just another comment in the review thread.
Test it once on a live chart, on purpose. Pick last month’s update. Sit with the drawer. Ask the two questions. Time how long it takes to get a source, a window, a denominator, and one thing that would falsify the claim. If the answer is a shrug, or a pointer back to you, the alarm already fired. You just have not routed it past the polish.
Algedonic signal, the unanswered-pair lens
- Assumes: the review chain attenuates by design, and one rare failure should skip it.
- Fits because: the two unanswered questions are a non-analytic wake-up, not a design critique.
- Breaks when: every chart fails the test, or the founder also cannot answer.
- Evidence: grade B-. Variety filtering is solid; Beer’s case base is thin.
- Counteracts: sending a bad chart back for another polish.
- May reinforce: founder inspection of every pixel, if the pair is asked as a style note.
Change the brief before you change the drawer
Draw the boundary first. Inside: public numbers, the brief, the two questions, and the name that signs the claim. Outside: chart type, axis zero, and how to wire a midnight interrupt for a vital operating variable. Those are other pieces. You have agreed to be surprised by a truthful chart that still fails because the story was wrong, and by a contractor who could have owned the claim if anyone had asked.
The thing you are trying to move is a stock: published claims that someone can stand behind. Decoration is an inflow. You cannot set the stock directly. You can change the brief, withhold the next one, or take a chart down.
The lag is long. A decorated lie on the homepage is often found only in diligence, months later, by someone reconstructing from receipts. A correction sized as if the lag were zero looks like another design tweak.
Cheapest first, and reversible:
- Rewrite the next brief so it names the claim, the source, the window, and the person who will answer the two questions. Cost: the time to type four lines in the same WhatsApp thread you already use.
- Ask the two questions on the last three public numbers, with the person who drew them in the room. Cost: an hour.
- Put a one-line owner on every public chart: name, source, what would falsify it. Cost: a caption.
- Withhold the next brief from a drawer who cannot answer. That is the penalty in a repeated relationship, and it is cheaper than a legal clause you will not enforce.
Do not start by firing the contractor. Headcount is the expensive, slow, people-shaped move, and it leaves the brief intact. The next person will decorate too.
What ships before the next public number
Do now, one working session, effect the same day. Take the last three numbers you put on a homepage, a pitch, or a monthly update. Sit with whoever drew each one. Ask how we know this is right, and what would falsify it. Write down whether they could answer without looking at you. Reversible, cheap, and dominant across every story about whose fault the last slide was.
Hedge, premium is a caption, live before the next send. Add one line under every public chart: owner, source, one falsifier. If the number was already solid you have spent a line of type. If it was not, the missing line is the insurance paying out.
Defer and trigger, size fixed now. Do not rebuild the visual system this quarter. Pre-commit the trigger: the next time a drawer cannot answer the two questions, that chart comes down before the next send, and it stays down until a named person can reconstruct the number from source. Decide that rule now, while nothing is on fire.
A DO NOW whose effect arrives after the next investor note goes out is misordered. Run the two questions this week if a send is on the calendar.
What a service desk usually produces next
Run the break test before you borrow anyone else’s history. Has a rule changed. Has an actor entered or left. Has a measurement become a target. If “make it pop” is now how the slide is judged, last year’s good-looking update that happened to survive diligence describes a different process.
If nothing broke, name the shape. This is a stable service equilibrium, not a swing. The desk waits. The order arrives. The decoration ships. Cairo’s own limit is the base rate: outside top publications in the United States and in certain European countries, a majority of news graphics departments are still service departments.2 A founder-and-contractor pair is closer to that majority than to the Times. Autonomy is the exception he documented.
Credit nothing to a new designer until the brief changed. A quarter in which the numbers happened to be true, under a pretty-only brief, is the counterfactual you have to subtract. The test is whether someone other than you can reconstruct the claim. Usually, after a hire alone, they still cannot.
If the drawer is you, the reference class moves. Self-authored decoration is a different failure. The two questions still apply. The person you ask is the person you see in the morning.
What owning the claim still leaves out
These three models can tell you who the brief hired, why a finished chart gets believed, and what to do the moment the two questions fail. None of them can tell you whether the founder could answer either. If you also cannot name the source, the window, and the falsifier, the alarm is about the company, not the contractor. Pull the chart. Do not blame the drawer.
None of them separates a decorated accident from a decorated lie. Motive is outside the boundary. Diligence will not care which one you thought you were running.
The heritage reset is real. Cairo’s transition was watched inside elite US and UK newsrooms. A contractor in Kampala or Lagos, paid by the slide, is not that desk. Grade the organisational finding one step down until you have seen ownership hold in your own market. The multitask result does not need that reset. It is about what the invoice buys.
One property no member models: once you require a named owner, good decorators leave, and you ship fewer charts. That is the point, and it is also a cost. The last-minute slide that made a true-enough number readable will stop. You will notice it only when a send goes out late, or ugly, or not at all.
The one action that survives the ignorance: before the next public number leaves the building, take the last three that already did, and ask the drawer the two questions. If two of the three cannot be answered, those charts come down by T+3, and the next brief names an owner before it names a colour.
Who has to sign
The founder. The contractor cannot assign themselves the claim, and they cannot rewrite the brief that keeps them in a service seat. The cheapest first test is the hour with the last three numbers. Bring the answers, or the silences, to the meeting where the next homepage tile, the next pitch, or the next monthly update is being approved. The contractor did not own the lie. You published it.
Sources and notes
- Alberto Cairo Touriño, Nerd Journalism: How Data and Digital Technology Transformed News Graphics, doctoral thesis, Universitat Oberta de Catalunya, draft dated 5 June 2017. Introduction, printed pp. 8 to 9. News graphics used to be made mainly by visual designers and artists who took “orders” from reporters and editors; at elite publications the people who produce them are now content creators and autonomous. Landing page (author, programme, 9 June 2017 record date, abstract): UOC Doctoral School showcase. Quotes were read from the 5 June 2017 draft PDF. Cairo labels that file the latest draft, with corrections still remaining. Treat it as a 2017 draft dissertation, not a finished press book.
- Cairo, ibid., Chapter 6, “From service to autonomy,” printed pp. 238 to 239. Move from a service model (desks waiting to be told what to do) to content ownership: professionals do not take orders, they collaborate, and in many cases they pitch their own pieces and complete every stage from initial reporting to final art. Same chapter: outside top publications in the United States and in certain European countries, a majority of news graphics departments are still service departments. Used in section 1 and in the reference-class paragraph.
- Cairo, ibid., Chapter 4, “Service vs autonomy,” printed p. 172, quoting Scott Klein of ProPublica. “We very much see the work we do as journalism itself.” In the old days there was the art department; you sent them a story and said “make a chart for me.” Used in section 1 for the old brief.
- Cairo, ibid., Chapter 4, “Attention to audience,” printed pp. 192 to 193, and Chapter 6, printed p. 250. Len De Groot: “I think we’ve started to care about the audience, and that’s a change.” Chapter 6 records the New York Times change: richly interactive graphics gave way, from 2014, to more linear work after the desk observed that most readers do not interact much beyond a next button or scrolling, on traffic that is often mobile and sometimes as high as 80 percent. Used in section 2 as form-follows-use, a fold-in to feedback misperception, not as a new model. Cairo’s own external-validity limit, Chapter 6, printed p. 251: the results apply to certain organisations in the English-speaking world, particularly the United States.
- Bengt Holmstrom and Paul Milgrom, Multitask Principal-Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design, Journal of Law, Economics, and Organization 7 (Special Issue), 1991, pages 24 to 52. Open copy: https://www.sfu.ca/~allen/HolmstromMilgrom.pdf. Verified in the PDF: when volume is easy to measure and quality is not, piece rates raise volume at the expense of quality; incentive pay directs the allocation of attention among duties; an increase in compensation on one task reallocates attention away from others; the case for incentives on any one activity decreases with the difficulty of measuring competing activities. Used in section 1 for what a pretty-only brief buys.
- John D. Sterman, The Beer Distribution Game, MIT, https://web.mit.edu/jsterman/www/SDG/beergame.html. Each player has good local information and severely limited global information. After play, most sketch customer demand as a surge and a crash; Sterman writes that blaming the customer is plausible, psychologically safe, and dead wrong. Customer demand steps from four to eight cases in week five and then stays flat. Used in section 2 for the misread of a local output (the finished slide) as evidence about a hidden process. The alarm-that-skips-the-meeting piece used this page for the chain’s information limits, not for the blame-the-customer finding.
- Matthew A. Cronin, Cleotilde Gonzalez and John D. Sterman, Why don’t well-educated adults understand accumulation? A challenge to researchers, educators, and citizens, Organizational Behavior and Human Decision Processes 108(1), 2009, pages 116 to 130. Author copy: https://www.mit.edu/~jsterman/CroninGonzalezSterman061210.pdf. Persistent poor performance is not attributable to an inability to interpret graphs, contextual knowledge, motivation, or cognitive capacity. Used in section 2 to close the “hire a better visual person” exit.
- Stafford Beer, Diagnosing the System for Organizations, Wiley, 1985, page 134. The glossary defines algedonic as “pertaining to regulation in a non-analytical mode; raising alarm”. Used in section 3 for the design point that the unanswered pair carries no diagnosis. This article does not reuse the channel-architecture argument of the alarm-that-skips-the-meeting piece; the alarm here is the failed pair of questions, not a wired interrupt on a vital operating variable.
A note on what is deliberately absent. This piece does not tell you which chart type to pick, whether to truncate a line-graph axis, or which operating numbers may ring your phone at midnight. Those are other decisions. It also does not mint service-versus-autonomy as a model. That is Cairo’s organisational finding. The three live IDs are the cards above.
Joshua Agonya Pi’Rwot, Founder.