Why Resilience Cannot Be Retrofitted
Short summary
Structural problems in a live service were present in the design, they were just less visible. Resilience behaves like a design discipline that has to be present from the start rather than a quality assurance phase near the end. Adding it afterwards, through patches, workarounds and post-launch fixes, costs significantly more than building it in from the beginning, and produces a structurally weaker service at the end. The cost of designing for variation during build is small and known. The cost of fixing structural fragility after launch is large, compounding, and spread across places nobody is looking.
What a retrofit actually looks like
Retrofitting resilience rarely takes the form of a redesign, it comes in the form of accumulation over time.
Let’s say that a structural gap appears in live use. There is no time or budget to redesign the journey, so something is put in place to cover it: a manual process, a staff-assisted completion route, an exception-handling script, a workaround in the support playbook. It does the job and the immediate problem stops being a problem.
Then it happens again with a different gap, and then again somewhere else. Each individual response is proportionate to the problem at hand. The accumulated result, however, is a service held together by mechanisms that exist because the design left something open.
The cost, with numbers
Consider a delivery team rebuilding a housing benefit application service. The original, launched three years earlier, has accumulated 47 documented workarounds since launch. Manual processes, staff-assisted completions, exception-handling scripts, each one addressing a structural gap in the journey that was never designed for.
Each workaround costs around £2,000 a year to maintain. That is £94,000 annually, spent entirely on holding together gaps that design left open. None of that appears in a design budget. It appears in operational cost, staff time, and support capacity, where nobody connects it to a design decision made three years earlier.
The rebuild team is given £180,000 and eight months. At the end of it, the service has 23 new workarounds. Fewer than the original, which reads like progress, and still substantial, because the team ran out of time to test variation scenarios in the final two sprints.
That last detail is the important one as it shows the same pressure that produced the original's fragility produced the fragility in the rebuild. A rebuild does not fix a structural problem if it recreates the conditions that caused it.
Why the rebuild repeated the mistake
The rebuild had every advantage. A team that knew the service's failure modes, documented evidence of 47 specific gaps, a defined budget, and eight months. Yet it still shipped requiring 23 workarounds.
The reason is that variation testing sits at the end of a delivery timeline, and the end of a delivery timeline is where compression happens. Scope pressure does not distribute evenly across a project. It concentrates in the final sprints, and whatever is cut first tends to be whatever cannot be seen in a demo.
Resilience work is almost never seen in a demo. It does not produce a screen anyone can show. It produces the absence of a future problem, which is very difficult to demonstrate and easy to defer. When those final two sprints compress, the variation scenarios are what gets cut, every time, in every organisation, for entirely understandable reasons.
This is why treating resilience as a late phase reproduces fragility regardless of intent, as it is scheduled in the position where it is most likely to be sacrificed.
Why designing it in is cheaper
The cost of designing for variation during the build is small and known. It is design time, spent early, on questions that are cheap to answer before anything is built.
The cost of fixing structural fragility after launch is large, compounding, and distributed. It appears as support volume, as workaround maintenance, as staff time absorbed by manual processes, and eventually as rebuild cycles. Each of those sits in a different budget from the design decision that caused it, which is why the total is almost never calculated.
The £94,000 a year in the housing benefit example is a real cost that the organisation pays continuously, and it existed because a set of questions were not asked during design. Those questions would have cost a fraction of that to answer once.
The compounding is what makes the comparison so unfavourable. A design cost is paid once, whereas a workaround cost is paid every year until the service is replaced, and the replacement inherits the same pressures unless something changes about how it is built.
Instead of asking whether we can fix this after launch, ask what it is going to cost us not to design for this now.
Frequently asked questions
Can you add resilience to a service after it has launched?
You can add mitigations, but not resilience in the structural sense. Post-launch responses take the form of workarounds: manual processes, staff-assisted routes, exception-handling scripts. Each covers a specific gap without closing it. The accumulated result is a service held together by mechanisms that exist because the design left something open, at a recurring maintenance cost that continues until the service is replaced.
What does structural fragility cost an organisation?
In one housing benefit example, 47 accumulated workarounds cost around £2,000 each per year to maintain, totalling £94,000 annually. That cost sits in operational budgets, staff time, and support capacity rather than in any design budget, which is why it is rarely connected to the design decisions that produced it. The cost recurs every year until the service is rebuilt.
Why do service rebuilds often recreate the same fragility?
Because variation testing sits at the end of a delivery timeline, and the end of a timeline is where scope compression concentrates. Resilience work does not produce anything demonstrable in a sprint review, so it is easy to defer when the final sprints compress. In the housing benefit rebuild, the team ran out of time to test variation scenarios in the last two sprints and shipped 23 new workarounds despite knowing the original's failure modes.
Is designing for resilience more expensive than fixing problems later?
No, and the gap in cost is large. Designing for variation costs design time, spent early, on questions that are cheap to answer before anything is built. Fixing structural fragility after launch costs support volume, workaround maintenance, staff time absorbed by manual processes, and eventually rebuild cycles. The design cost is paid once. The workaround cost recurs annually until the service is replaced.