A Sunday newsletter for senior and enterprise architects. One decision, one practical model, one tool you can use Monday. Written by Kevin Donovan. Join 2,500+ architects. One email every Sunday.
Share
Decision latency: find and price the weeks you wait
A warning arrives early. Permission arrives late. The recovery gets called heroic.
Mike Carroll opens his article on decision latency with a composite plant-floor scene: a machine shows signs of trouble, but the person who can approve a controlled shutdown isn’t available. The line keeps running. Hours later, the bearing fails.
Suddenly, overtime, emergency freight and a changed production schedule are all possible.
His point stayed with me: “The enterprise does not book decision latency. It books the receipts.”
In architecture, the receipt can look less dramatic. An integration ships point-to-point while the platform decision waits. A contractor stays another sprint. A renewal goes through on terms everyone wanted to change.
Each cost has somewhere to land. The wait that helped create it often doesn’t.
This week I returned to a question I wrote about in July: how much architecture work goes into improving the decision, and how little goes into reducing the time it spends waiting?
Today, let’s make that wait visible enough to do something about it.
In this edition
Where to start the clock, and which waiting earns its time.
The Decision Latency Ledger: 3 dates and 1 receipt.
A worked integration decision, including the cost calculation.
How to take the finding to a steering committee.
A reopen rule, a better use of the review meeting, and Community Corner.
Let’s get into it.👇
________
🧭
________
Start the clock carefully
In July, I described decision latency as the time between recognizing that a choice exists and committing to a course of action.
That is a useful first question. It gets us looking at the gap.
Mike uses a narrower starting point for pricing it: when enough evidence exists to act responsibly, a feasible action is available, legitimate authority can be identified, and waiting changes the outcome or its cost. He follows that interval through to authorized action.
Those clocks answer different questions. The first helps us see the whole journey. The second helps us investigate which part of it might have been avoidable.
An alert on Monday doesn’t prove that Tuesday’s decision was late. The team might still need a compatibility test, a supplier response, or evidence that the proposed action is safe.
The useful question is what changed during the interval.
Did another test rule out an option? Did a conversation reveal a dependency? Did the review establish a condition that made the choice responsible?
Or was the same recommendation sitting in an inbox because the next forum wasn’t until the 18th?
Keep useful deliberation in the record. Otherwise, the exercise becomes an argument for haste, and the people carrying the risk will rightly stop listening.
The Latency Tax sits between owners
I use 4 steps to trace the journey:
Signal → Decision → Permission → Action
Then I look for 3 kinds of delay:
Reporting: how long before the signal reaches someone who can interpret it?
Approval: how long before a recommendation gets an authorized commitment?
Coordination: how long before the people involved can act together?
These are diagnostic categories, not neat departments. Capacity discussions can delay a commitment as well as its implementation. A missing owner can affect more than 1 step.
Follow the handoffs. A recommendation can be ready while authority or capacity is still unresolved.
A dashboard may show the reporting delay quite well. The other waits can disappear between teams’ measures.
The architect records the recommendation as complete. The forum records that it reviewed the paper. The delivery team records that work hasn’t started.
All 3 records may be accurate. None explains the elapsed time.
I call the cost of those avoidable waits the Latency Tax. To reduce it, we need a trace from the choice to the consequence.
________
🧭
________
The Decision Latency Ledger
Start with 1 architecture decision from last quarter that touched more than 1 team. Choose something ordinary enough to happen again: an integration pattern, a data ownership question, or a buy-versus-build choice within a known spending boundary.
Give yourself 20 minutes for the first pass. Repeat for 2 more decisions once you know where the records are.
This is a small working record. Copy the fields into a note and fill the gaps you can support.
Copy this row
Decision and teams affected:
1. Visible / decision-ready: Date __
What evidence shows that a responsible action was possible?
2. Committed: Date __
Who authorized the choice, and where is it recorded?
3. Acted: Date __
What was the first implementation of that authorized choice?
Ready → committed:__ calendar days
Committed → acted:__ calendar days
Useful deliberation:__
Approval wait and cause:__
Coordination wait and cause:__
Earlier reporting/evidence delay, recorded separately:__
Interim workaround and its date, if any:__
1 incremental cost, with its source:__
Owner of the proposed change:__
A few rules that keep it useful
Use dated evidence where you can: the recommendation, test result, approval message, ticket or first implementation change. Leave an unknown date unknown. An honest gap is more useful than a precise guess.
Keep earlier evidence gathering separate from the decision-ready interval. Don’t count the same week twice. If 2 activities ran in parallel, don’t add their elapsed time as though they happened one after another.
The dates locate the wait. They don’t establish its cause or prove that it was avoidable.
And keep a workaround in its own field. If a team shipped a temporary integration before the platform choice was authorized, that is important evidence. It isn’t implementation of the later choice.
3 dates locate the interval. The receipt starts a conversation about what the wait cost
Work through 1 decision
Here’s a smaller example than Tuesday’s 11-week scenario, so we can follow the arithmetic.
This is illustrative, not client data. The dates, costs and partitions below are hypothetical.
A delivery team needs an integration pattern. The shared platform is a plausible choice, but the team must establish whether it can support the required flow.
Week 1, Monday: The choice is recognized.
Week 2, Monday: Enough evidence exists to recommend a bounded implementation on the shared platform. The decision is ready.
Week 6, Monday: Written authorization arrives.
Week 7, Monday: Implementation of the authorized choice begins.
That gives us 28 calendar days from ready to committed, followed by 7 days to start implementation.
Suppose the record accounts for the first 28 days like this:
3 days of useful validation.
18 days waiting for the approval forum.
7 days waiting for confirmation from the capacity owner.
The earlier 7 days of evidence gathering remain separate. The later 7 days are mobilization. Neither interval disappears, and neither is automatically waste.
These are elapsed calendar days, not days of labor. People may have spent only a few hours on the decision within them.
Meanwhile, the delivery team shipped a point-to-point workaround in week 4 to meet a commitment. That choice belongs in the ledger too.
Mohammed Alsayed made this point in the comments: we have to design for the reality of the deadline. A temporary solution can be the responsible choice. Its replacement plan and migration cost should be part of that choice from the start.
Now follow 1 receipt
Suppose replacing the workaround required 8 contractor person-days at €800 per day. The delivery owner estimates that 3 of those days would have been needed even if the platform decision had arrived on time.
The candidate incremental expense is:
(8 − 3) × €800 = €4,000
The whole €6,400 bill doesn’t belong to the delay. The extra 5 days might.
Before presenting €4,000 as a finding, check what those days actually covered. Was the extra work caused by the temporary integration? Would it still have happened because the scope changed? Is there evidence that an earlier commitment could have prevented it?
If the explanation isn’t supported yet, call it an estimate to validate. Don’t turn it into a proven saving.
This is the discipline I take from Mike’s cost approach: compare with the feasible earlier path, trace the increment, and keep unlike benefits separate.
A one-time contractor expense is different from recurring operating cost. Freeing employee time may create capacity without reducing payroll. A forecast improvement is still a forecast until you observe it.
You don’t need a large number. You need a number whose meaning survives the next question.
________
🧭
________
Take the row to the person who can change the wait
A steering committee can do more with a specific decision class and a proposed change than with a general complaint about slow governance.
Bring the row, its evidence, and an ask:
This choice was decision-ready in week 2 and authorized in week 6. We can account for 3 days of useful validation, 18 days waiting for the forum, and 7 days waiting for capacity confirmation. Our current estimate is €4,000 in incremental replacement effort, subject to validation against the work records.
For the next decision in this class, can we name the capacity owner before review and delegate a bounded, reversible choice between meetings?
One case is enough to open a conversation. It isn’t enough to claim an org-wide average or multiply the cost across every project.
Try the change on the next few comparable decisions. Check whether the wait fell, and whether errors, rework or risk increased. A faster bad outcome is no improvement.
Put the decision rights where the work arrives
For that recurring decision class, write down:
Who recommends?
Who decides?
Who carries the consequence and must be involved?
What is already settled, and what would justify reopening it?
Name the people or accountable roles. Put the answers beside the intake path and the recommendation template, where the next team will actually look.
For a reversible choice, add the minimum evidence, scope limits, rollback method and a review date.
Then write the trigger that would reopen it. “If circumstances change” leaves too much for the next person to interpret. A failed compatibility test, an agreed cost threshold, or a material assumption becoming false gives them something to check.
I’ve talked before about deciding with 70–80% of the information when we can revisit, revise or reverse. That is a reminder that certainty has a cost. It isn’t a substitute for the evidence standard that the particular decision needs.
Give the meeting a useful job
Pre-work helps. It lets people test the recommendation, understand the trade-offs and find missing owners before everyone is in the room.
But real decisions can still happen in the meeting.
I think of the forum as triage:
Ready and bounded: Decide, record the owner and conditions, and move.
A specific question remains: Name who will answer it and the date it returns.
A trade-off needs agreement: Put the consequence owners in the conversation and make the choice.
“Discussed” is a weak outcome. A decision, a named investigation, or a return date gives the next person something to work with.
This is where the AI conversation lands for me. Faster analysis can get more recommendations to the door. It doesn’t automatically name the person who can authorize them or make room for the work.
If the permission path stays the same, a faster answer may just join the queue sooner.
________
🧭
________
Community Corner
Three contributions sharpened this week’s discussion.
Mohammed Alsayed pushed back on treating the workaround as something architecture should merely prevent. Design for the deadline the team actually has, with a replacement plan and migration cost included. Jordan Corn is trying architecture reviews closer to real time to reduce the wait for the forum. The results are still to come. Read Tuesday’s discussion.
Simon Wright emphasized doing the discussion and agreement before the meeting, so approval can be brief. Our exchange helped me clarify the triage approach above: pre-work matters, and some choices still need the people in the room. They should leave with a decision, an owner or a return date. Read Friday’s discussion.
Thanks to everyone who added a practice, a challenge or an example. They make this more useful than a post on its own.
________
🧭
________
Where to start Monday
Pick 1 recurring decision. Open the calendar and the work records. Give the first row 20 minutes.
Find the decision-ready date, the commitment, the first authorized action, and what happened in between.
Then choose 1 receipt worth checking with the person who owns the work.
If you try it, hit reply with the longest gap you found and what it was waiting for. Approval? Capacity? Evidence? An owner nobody had named?
Clarity collapses latency. Start by making the wait visible.
Until next Sunday,
Kevin
________
________
Kevin Donovan I write Architects' Hub every Sunday for senior architects and technology leaders. You're getting this because you subscribed at architects-hub.kit.com. Know someone who’d find this useful? Forward it to them.
Kevin Donovan · Turning architecture into decisions
A Sunday newsletter for senior and enterprise architects. One decision, one practical model, one tool you can use Monday. Written by Kevin Donovan. Join 2,500+ architects. One email every Sunday.
Nov. 22 2026 | Read time: 11 mins. Kevin Donovan · Architects' HubForwarded this email? Subscribe to Architects’ Hub for the next issue. The room went quiet Hey , Two Sundays ago, our hypothetical pilot case got its commitment record. Last Sunday it got a walkthrough, and the operating owner signed with 3 tests on record. The AI assistant now drafts delayed-order replies for 3 teams, and the drafts are good. This week someone proposed letting it do more than just draft… In this edition The...
Nov 15, 2026 | Read time: 11 mins. Kevin Donovan · Architects' HubForwarded this email? Subscribe to Architects’ Hub for the next issue. Tuesday, 10:40 am. Hey ,Last Sunday we left a hypothetical customer-service assistant with a commitment record and unanswered questions about capacity, data and support. This week, move the story forward 3 weeks. The team has agreed to a bounded trial on 1 queue. Agents still review every draft. Wider rollout is waiting. I want to spend this edition with the...
November 8, 2026 | Read time: 9 mins. Kevin Donovan · Architects' HubForwarded this email? Subscribe to Architects’ Hub for the next issue. Hey ,In this edition of Architects' Hub: What the service lead heard when the sponsor said, “We should roll this out.” A commitment record filled in for one example customer-service queue, PLUS a blank copy for yours. How to take a conditional decision back to the sponsor who wants a date. Community Corner: Santiago Jesus Lasso Valenciano on the work that...