Most executives do not suffer from too little data but from too much. Every morning there are the ten dashboards, the weekly report, the monthly statement, the CRM, the web analytics, the export from the finance system. The data is there. The decision has not got any easier.
This is about that gap: why data does not turn into a decision by itself, and what it takes for it to. The frame we will look through is old and well established, though its name is rarely spoken in a management meeting — the DIKW pyramid.
Five levels from data to decision
DIKW is a sequence of four levels, five in some versions, describing how we get from raw data to decision-ready wisdom. The acronym comes from the level names: Data, Information, Knowledge, Wisdom. The extended version inserts a fifth level between knowledge and wisdom: Insight.
The model is often attributed to the 1989 work of American organizational theorist Russell Ackoff, From Data to Wisdom, though the idea is older and the exact number and naming of the levels varies by author. That malleability is not a flaw in the model. It is the first telling sign that this is not some rigid staircase carved in stone. We will come back to it at the end.
The levels themselves are simple once stated:
- Data: a raw signal without context. A number, a timestamp, a log line. On its own it means nothing.
- Information: data given context. Structured, labelled, aggregated. For example: "March revenue is 12 percent higher than February's." This already answers a question — what happened?
- Knowledge: the relationship between pieces of information. "Revenue jumps when the campaign and the reorder wave coincide." You see not only what happened but how things connect.
- Insight: the realization that emerges from knowledge and moves a decision. Not every relationship matters equally. Insight is the one that forces action.
- Wisdom: insight turned into a decision. You not only know what matters, you decide what to do about it, weighing the goal, the risk and your values.
The power of the model shows when you run the same dataset through all five levels.
All five panels show the same few dozen data points, yet they show progressively more. Moving left to right, the data does not increase — what we can read from it does.
Not a staircase but phases of work
The image of a pyramid can mislead, because it suggests stairs you walk up. In reality every level change demands a different kind of work, and none of them follows automatically from the previous one. Look at what happens between two levels and it becomes clear why most organizations stall exactly where they do.
- Data to information: structuring. You add context: when, how much, compared to what. Machines do this step well, and a properly configured dashboard does precisely this.
- Information to knowledge: finding relationships. Interpretation is required here. A dashboard shows that two curves move together. It does not show whether that is causal or coincidental. That takes human thinking, or a model doing it on our behalf.
- Knowledge to insight: selection. Most relationships are true but immaterial. Insight is the rare point where a relationship forces a decision. That takes judgement about what to let go of and what to attend to.
- Insight to wisdom: the decision. Insight is still description. Wisdom is a choice, with responsibility and consequence. This is where the goal, the risk appetite and the value system enter, none of which can be read out of a database.
This breakdown shows the essential point: what moves between the levels is not data but work. And in the upper half of the pyramid that work does not speed up, it slows down, because it becomes progressively harder to mechanize.
The dashboard ceiling
Now it is visible why most companies stall at the same point. The dashboard, the report, the management statement all automate exactly one level change flawlessly: they turn data into information. They tell you what happened, quickly and reliably. And they stop precisely where the harder work begins.
In practice, on reaching the ceiling, organizations are often driven by the wrong instinct: gather even more data. Commission another dashboard, connect one more source, refine the report. But that is not upward movement on the pyramid, it is a sideways step that widens the bottom level. With more information, the relationship, the insight and the decision are just as absent as before. There will be more data, and the decision will be exactly as hard.
The cause of the ceiling is organizational rather than technological. The data-to-information step is performed by a system. The steps from information to knowledge, then insight, then decision have to be made someone's explicit job in the organization. In most places nobody carries that responsibility. Someone produces the report, someone looks at it, and then the question of what it means and what to do about it hangs in the air, because it is not in a calendar, not in a job description, and has no owner.
Where the model itself is wrong
DIKW is a useful thinking frame, but treating it as textbook truth would be a disservice. Three points call for caution, and they are the ones the model's critics typically raise.
First: it is not linear. In reality the levels overlap and slide into each other, and we rarely proceed in clean order from bottom to top. Often an insight flashes first, and only then do we go looking for the data to support or refute it.
Second: the levels feed back into each other. Wisdom is not a terminus at the top of the pyramid but something that flows back to the bottom: a mature decision changes what data is worth collecting in the first place. A good executive often asks not for more data but for different data.
Third: the top does not emerge from the bottom on its own. The notion that accumulating enough data will let wisdom assemble itself is false. Insight and decision require judgement, context and a value system, none of which comes from the data itself. That is why the top two levels remain a durably human responsibility even once the lower ones are handled by machines.
And as we saw, the retrofitting of the Insight level is itself telling. If a model's levels can be freely renamed and extended, we are not dealing with a law of nature but with a useful map. And a map is good because it orients you, not because it covers the terrain perfectly.
What to do with this tomorrow
The model is worth something if it changes the question you put to your own organization. Not "do we have enough data" but "at which level do we make our decisions, and whose job is it to move them up".
If the answer is that the company lives at the information level — plenty of reports, but the why and the what-should-we-do sit on nobody's desk — then the missing link is not another tool. A disciplined research or discovery process is exactly what turns that upward movement into explicit, doable work: structured information out of raw data, relationships out of that, the few genuinely decision-moving insights out of those, and finally a defensible direction.
Picture two companies in the same industry with roughly the same data assets. One sits at the dashboard ceiling and knows precisely what happened last year, yesterday and this morning. The other gets from the same data to a decision: it knows the two or three things that matter, and what it does differently because of them. The difference between them is not in the data and not in the software. It is that one organization treats moving up a level as work, and the other believes data reaches the summit by itself.