To retire a process, you stop new work entering it, finish what is in flight, and preserve what it produced. Deleting it loses the history; leaving it running creates confusion. Why obsolete processes linger Nobody wants to be the one who deleted something. * Forms still accepting submissions nobody handles * Automations
Finding a process nobody owns is common in any company past a certain size. It runs, nobody maintains it, and it degrades quietly until it fails. How to spot one The signs are visible in the data before anyone raises them. * Nobody able to say who is responsible * No changes
To design a process for occasional users, assume they remember nothing. Someone submitting a request twice a year is a first-time user every time. What frequent-user design gets wrong The builder uses it daily and designs for themselves. * Field labels meaningful only to the team * No explanation of
To handle rework in a process, records that go backwards have to be visible as such. Rework hidden inside normal flow makes a process look healthier than it is. Why rework goes unmeasured A card moving backwards looks the same as one moving forwards. * Returned work indistinguishable from new work
To build a triage step worth having, it must remove more work than it adds. A triage phase that simply queues everything before the real queue makes the process slower. When triage adds nothing It is often added by default rather than by need. * One person reading everything to sort
To run a process with statutory deadlines, two things matter equally: meeting the date, and being able to evidence that you did. What is different about these deadlines Missing one has consequences beyond inconvenience. * No flexibility to negotiate an extension * Evidence required, not just completion * Consequences falling on the company
To design an escalation ladder, define what triggers each level. Escalation that depends on someone deciding to escalate happens too late and inconsistently. Why escalation fails It relies on judgement at the moment judgement is hardest. * Escalation happening only after a complaint * Different thresholds applied by different people * Senior people
To structure a process that spans two teams, the handoff needs an owner on both sides. Work sitting between departments is work nobody has picked up. Where handoffs fail The gap between two processes is where records wait longest. * Each team assuming the other has it * No notification when work
Structuring work that waits on an external party matters because the timing data is meaningless without it. A process that spends half its life waiting on customers looks slow when it is not. Why the distinction matters Improvement efforts aim at whichever step looks slowest. * Internal delay and external waiting
To design a process with a hard deadline, work backwards. A process designed forwards from the request discovers the deadline problem only when it is too late to act. Why fixed deadlines are missed The urgency arrives at the end rather than at the start. * Each step taking as long
To handle work that arrives in batches, the process has to absorb a spike without the team working through it at the same rate as a normal week. What batch arrival does to a process Average capacity is irrelevant when arrival is concentrated. * A month's volume arriving in
To design a fair work queue, remove the choice about what to pick up. A shared queue where everyone selects their own work produces predictable distortions. What selective picking produces Nobody intends it and everyone does it. * Easy items handled first, difficult ones ageing * Some people carrying disproportionate load * The