How We Work

A client knows something is not working and suspects the reason is not technical. A process still runs, but it takes thirty days when it should take ten, or it bottlenecks on two people, or it produces a result nobody can defend. The documentation does not explain it, and the people who can either lack the standing to do so or have already left.

A client knows exactly what is needed and cannot find anyone able to build it. The requirement is genuinely difficult, the niche is too narrow for a product to exist, and the firms large enough to attempt it would be both prohibitive and wrong for the problem. This is more common than it sounds. One client required what amounted to an enterprise resource planning platform for a global research program, and no product served that niche; it has been in production for a decade. Another required a real-time exchange carrying on the order of a thousand posts a day, each generating thousands of notifications, approaching sixty million a month and delivered as they occur. That client knew precisely what was wanted. Finding anyone who could build it was the difficulty.

Where the conversation does not lead anywhere is when the requirement is clear and the work is ordinary: a standard directory, a routine community, a content system. That work is well served by platforms and by capable implementation partners, and we will point a client toward one.

Some of our best work does not appear on this website and never will.

A problem arriving from an executive office often has two versions. The stated version appears in the strategic plan. The actual version is the one keeping someone awake, and it generally involves a person, a relationship, or a risk that nobody has written down because it can and will be addressed.

We are careful about what we repeat, where we say it, and how it will appear to someone who was not in the room. That is less a matter of policy than a precondition for the first conversation.

Analysis and design of this kind require sustained and uninterrupted attention, a day at a time rather than an hour. Holding a dozen interacting constraints in mind is not something that can be set down and picked up again; a single interruption collapses the structure, which then has to be rebuilt from the bottom.

We therefore schedule this work in days and limit how many run consecutively. A firm can produce only so many days of genuine judgment in a year, and we prefer to establish that limit deliberately rather than discover it late in an engagement.

Less than the industry claims, and more than we expected.

It did not change where judgment sits. That question is the same as it was in 1998. What it changed is when the question can be answered. We can build the exception path in the second week and watch someone use it, rather than describing it in a workshop and finding the flaw in month five. We can put three interface directions in front of the people who will live with them instead of arguing about which to build.

The judgment now happens earlier, and against real evidence. That is a change in method rather than in speed.

The engagements that go well are those in which someone tells us early that we are wrong.

We ask direct questions, we disagree with premises, and we expect the same in return. A firm that agrees with everything a client says is offering capacity rather than judgment, and capacity is widely available.

Some problems are best solved by putting everyone in one room for several days and not leaving.

The form is something like a charrette: the client's people, our people, and whoever else the problem requires, working continuously on one question until the shape of an answer appears. It compresses months of correspondence into a few days and surfaces disagreements early, which is where they belong.

These sessions work best away from the office, somewhere without easy communication and without the option of slipping out to attend something else. What matters is the concentration. Not every engagement needs one; those that do tend to be the ones in which the real problem is not yet agreed.