Think about the last time your architecture work was formally reviewed. Chances are it went reasonably well. Nobody said the artifacts were bad. Nobody said the people were weak.
Now think about the last reorganisation.
Most architects I speak to carry both of those memories, and cannot reconcile them. They were never told they were failing. They were simply moved, merged, thinned, or told the work would sit with the teams from now on. The natural conclusion is that leadership never understood what architecture was for.
I spent several weeks going through the evidence on that, expecting to confirm it. The evidence says something else, and it is less flattering and more useful.
Disclosure, because it shapes what I am arguing: I am building an architecture maturity instrument, and I sell advisory work and products in the field it covers.
The one time anyone asked the executives
Almost every study of architecture's value asks architects. That sounds like a technicality until you look at the exception.
In 2006, George Westerman and Peter Weill at MIT CISR surveyed 153 senior non-IT executives, asking them to rate their IT organisation across ten tasks. Architecture and infrastructure planning scored 6.5 out of 10. Third of the ten, above application development, above IT strategy.
So executives do not think architecture is bad. On the available evidence they think it is one of the better things IT does.
The same study looked at which tasks separated firms getting high business value from IT from those getting low value. Four tasks did. In the authors' words, "the other 14 tasks do not statistically distinguish high from low BVIT firms."
Architecture planning was not one of the four.
Rated well. And unconnected to the outcome the researchers were measuring.
Indispensable in the abstract, invisible in the particular
That result is not a one-off, and the second version of it is stranger.
Bas van der Raadt put architecture cases in front of 29 professional process assessors, people who audit IT processes for a living and averaged over 20 years of experience. He tested whether how well the architecture function performed predicted whether organisational goals were met. It did not. His words: "in the eyes of the assessors, there is no connection whatsoever."
And in the same passage: the assessors "thought that, in general terms, EA was indispensable."
Ask them in the abstract and architecture is essential. Ask them to judge a concrete case and its performance predicts nothing.
There is a third data point, and it comes from the people who built the model everyone cites. MIT CISR's four-stage architecture maturity model is the backbone of a thousand business cases. In 2012, Jeanne Ross and Anne Quaadgras reported that in their own later data "the statistical relationship between architecture maturity and business outcomes has been lost." You can become more architecturally mature, they wrote, without getting more value from IT.
What actually ends an architecture practice
If the story were "executives could not see the value, so they cut it," you would expect that reason to show up when practices are actually cut. Two Norwegian public-sector cases are the only peer-reviewed studies I could find that follow a function all the way to its end.
In a large municipality, architecture had been introduced because a government recommendation seemed to require it. Years later people realised the recommendation covered IT architecture principles, not enterprise architecture. The mandate had never really existed. Around the same time the IT manager who had personally wanted it left, and his successor saw no reason for a separate function.
At NAV, the Norwegian labour and welfare administration, a new CIO arrived and hired more than 200 people into agile teams. The central architecture function was dissolved into those teams.
Two details from that second case are worth sitting with.
The first: the CIO and the CFO were on the architects' side. The architects themselves had helped design the new operating model. They were not defeated in an argument about their worth.
The second is the one that changed my mind. In the researchers' description of the new culture: "KPIs were no longer used to determine success."
The organisation stopped measuring, and architecture lost anyway. If the inability to show numbers were the cause, loosening the measurement regime should have helped.
Why this is the harder problem
Put those together and the picture is uncomfortable.
Architecture is not cut because executives think it is bad. It is cut because they think it is fine, and because nothing they steer by moves when it changes. When a reorganisation arrives, a budget tightens, or a sponsor leaves, a function in that position has no defence available. Not because the defence was rejected, but because it was never constructed.
That is a harder problem than being misunderstood. You cannot explain your way out of an assessment that is already positive.
It also explains something that has puzzled me for years. This field has run perception surveys for two decades, asking stakeholders whether they value architecture, and the answers keep coming back acceptable. Of course they do. They are measuring the thing that already scores well.
And it explains why this is urgent now rather than eventually. Every large organisation is currently reorganising around AI. New platforms, new teams, new operating models, new leaders brought in to run them. That is precisely the class of event that ended both Norwegian cases, and it is indifferent to how well you were rated last cycle. NAV's architects lost their function to an operating-model change they supported. Being right, and being well regarded, did not enter into it.
What to do about it this quarter
Not a maturity assessment. Something smaller and more awkward.
Find one number your executives already look at every month. Delivery predictability, rework, change failure rate, time to decision, unplanned spend. Not a number you would like them to look at. One they already do.
Then take the last quarter and try to answer a single question: did any architectural decision measurably move it?
Most people cannot answer that. That inability is the finding, and it is worth more than a maturity score, because it tells you exactly where you stand. If nothing you decided touched anything anyone tracks, then your standing rests entirely on goodwill. Goodwill survives good times.
If you can answer it, you have the beginning of something no architecture practice I know of has: a claim about your own contribution that someone else's data supports.
Next
If the evidence for architecture's value is this thin, what about the numbers people cite in its favour? I went through them one by one. Most do not survive contact with their own sources, and that is the next piece in this series.
Sources & further reading
- Getting Higher Business Value from IT: The Non-IT Executive View — Westerman & Weill, MIT CISR Research Briefing VI(3A), December 2006 (the 153-executive survey; free full text via DSpace@MIT)
- Enterprise Architecture Coming of Age — Bas van der Raadt, doctoral dissertation, Vrije Universiteit Amsterdam, 2011 (the 29-assessor study and its null result)
- How Enterprise Architecture Loses Momentum: A Case of Delegitimization — Kohansal & Haki, ICIS 2021 (the municipality case; funded by the Norwegian Research Council)
- Enterprise Architecture's Ups and Downs Over Time — Kohansal & Haki, ICIS 2021 (the NAV case, including 4 top managers among 31 participants)
- The value of and myths about enterprise architecture — Gong & Janssen, International Journal of Information Management 46, 2019 (open copy at TU Delft)
- The TOGAF® Leader's Guide to Establishing and Evolving an EA Capability (G184) — The Open Group, 2018 (asks "how do we measure the value of good advice?" and leaves it open)
The Ross & Quaadgras finding is from MIT CISR Research Briefing XII-9, September 2012. CISR briefings older than a few years are no longer served publicly; I have not been able to link a stable copy.