Monday, weekly meeting. Two screens, the same metric, two different figures. One comes from the CRM, the other from finance, and the next twenty-five minutes go into working out which of the two is right. The meeting ends with an agreement that somebody should really consolidate this, so everyone sees the same thing. Nobody in the room asks what the company would do differently if the first figure were true, and what it would do if the second one were.
That question is where the work sits, and it is almost always looked for in the wrong place. The task that comes out of the meeting is to unify the dashboards. Every team gets the same view, the same colours, the same tiles. It feels like order because order is something you can see. The order that matters sits one floor down and stays invisible: in how the number is defined, and in the decision hanging from it.
A finding from 1967
Russell Ackoff, one of the founders of operations research, took apart five assumptions behind management information systems in Management Science in 1967. The first of them holds that managers lack relevant information. His finding was the opposite. They suffer from an over-abundance of irrelevant information, and a system that delivers still more of it makes the problem bigger.
That was written fifty years before the first self-service BI licence. Producing views has become cheap since then, close to free. Any analytics service will build you a custom dashboard in seconds, in whatever shape you ask for. Once building costs nothing, the picture becomes a commodity, and what turns scarce is the thing Ackoff was already pointing at: knowing which information changes anything.
What sits underneath the picture
Before a number can change anything, it has to be clear what it counts. Tadhg Nagle, Thomas Redman and David Sammon published a plain measurement in Harvard Business Review in 2017: 75 executives each checked a hundred freshly created records across ten to fifteen critical fields. On average, 53 per cent of records were error-free. Only three per cent of the quality scores reached a level the authors call acceptable. Almost half the records already carried a critical error before anyone drew a chart from them.
Richard Wang and Diane Strong described back in 1996, in the Journal of Management Information Systems, what data quality looks like from the side of the people using the data. They sort it into four groups: the data itself, its fit to the task, its representation and its accessibility. Representation is one of the four, and it comes behind relevance and timeliness. Working on representation alone polishes a quarter and leaves three quarters untouched.
The thermometer
A thermometer is a good instrument because there is a rule behind it. Above thirty-eight degrees we call, below it we wait until morning. The rule is what makes the measuring worth doing. Take the rule away and you have a number on a display that unsettles you and leads nowhere. Then you measure more often, because measuring feels like acting.
Most dashboards are thermometers without a rule. They show a value that gets looked at and commented on every week, and there is no threshold at which somebody does something, and no role that does it. Whether the dashboard looks identical across departments changes none of that.
Where the spread actually is
The spread does not sit in the picture. It sits in what people do with the same picture. Daniel Kahneman and colleagues described an insurance company in Harvard Business Review in 2016 that measured exactly this. Executives were asked beforehand how far apart two professionals' assessments of the same case would fall. The expectation was around ten per cent. What was measured: a 55 per cent median difference in the price different underwriters set for identical policies, and 43 per cent in the payouts different adjusters set for identical claims.
Same data, same guidelines, same firm. A unified dashboard would not have moved a single percentage point there, because the divergence appears after the number is read. Kahneman's remedy is procedures that fix the path to the decision. That is what standardising the decision means: the threshold at which something happens, the role that acts then, and the options being chosen between.
Where form does count
Form is not irrelevant for all that. Sara Hjelle and colleagues published an experiment with 524 participants in Information and Management in 2024: the format, currency and completeness of information improve decision quality indirectly, by lowering how complex the task appears. Good representation does work, and it works through what it carries.
That yields a distinction that saves a great deal of argument. What gets unified is the definition of a metric, its source and the decision path. What stays free is how a team arranges its view, because a warehouse works differently from a sales team, and some views simply do not fold into one another. What remains is visibility itself: everyone sees the same definition and can follow the same number over time. Transparency comes from shared meaning, and it survives the tiles being arranged differently.
How to start this week
Take a single metric of the ones you look at regularly. One.
- Write down the decision hanging from it. Which two or three actions come into play when this number moves? If none comes to mind, you already have your result.
- Set the threshold. At which value does something happen, and what exactly happens then? One sentence is enough.
- Name the role that acts. Which role steps in when the threshold is crossed, not which person. Without that line the threshold stays a statement of intent.
- Define the number in writing. From which source, for which period, what counts and what does not. Those four details end most Monday discussions.
- Check after four weeks whether it changed anything. If no decision hung on that number in four weeks, take it off the dashboard. It costs attention and returns nothing.
One metric a week. After a month a few numbers are left, each with a decision behind it, and the meetings get shorter.
The rest is decoration
A dashboard is worth what the decisions it triggers are worth. When the number is cleanly defined and carries an action, it works for you even sitting in a plain table. When both are missing, the prettiest view is a picture on the wall that opens the week and moves nothing.
The next part of the Clarity Playbook takes on what sits between the threshold and the action: how everyone recognises that a task is finished. Each part brings a method that builds on the one before.
One last thought, in case this reads as losing control. Taking numbers off the dashboard looks at first like giving up your overview, especially if you know the feeling of having to keep an eye on everything. In practice the opposite happens. Four numbers with a decision behind them give you more steering than forty you merely watch, and the attention you spent on watching becomes available for acting.
Care for your systems, and you care for your people.
Elisa
This article is written by Elisa Winkelmann with AI assistance and under her editorial review.
Sources
- Russell L. Ackoff, "Management Misinformation Systems", Management Science 14(4) (December 1967), pp. B147-B156. On the assumption that managers lack relevant information, and on the over-abundance of irrelevant information.
- Tadhg Nagle, Thomas C. Redman and David Sammon, "Only 3% of Companies' Data Meets Basic Quality Standards", Harvard Business Review (11 September 2017). Source of the 53% and 3% figures.
- Richard Y. Wang and Diane M. Strong, "Beyond Accuracy: What Data Quality Means to Data Consumers", Journal of Management Information Systems 12(4) (1996), pp. 5-34. On the four categories of data quality.
- Daniel Kahneman, Andrew M. Rosenfield, Linnea Gandhi and Tom Blaser, "Noise: How to Overcome the High, Hidden Cost of Inconsistent Decision Making", Harvard Business Review (October 2016). The noise-audit figures also appear in Kahneman, Sibony and Sunstein, "How noisy is your company?", strategy+business (19 May 2021).
- Sara Hjelle, Patrick Mikalef, Najwa Altwaijry and Vinit Parida, "Organizational decision making and analytics: An experimental study on dashboard visualizations", Information and Management 61(6) (2024), art. 104011. Experiment with 524 participants on the indirect effect of format, currency and completeness.
- Bart de Langhe and Stefano Puntoni, "Decision-Driven Analytics", MIT Sloan Management Review (2021). On reasoning backwards from the available courses of action to the data actually needed.