When Does Failure Become a Route?

Failure becomes tidier after the fact.

Once we know what happened next, a broken plan can be recast as a useful lesson. A rejected technology can become an early chapter in a success story. An unsuccessful attempt can be remembered as the thing that made the later solution possible.

Clyde Street is less obliging.

Some failures did become routes into something else. Some were followed by another test. Some were worked around. Some were simply left visible. And sometimes the response was to stop using the thing that had failed to fit the work.

That makes failure difficult to turn into a slogan.

The interesting question is not whether failure is good. A failed state, by itself, tells us only that something did not work as intended.

The harder question is what happens afterwards.

When does failure actually become a route?

A routine that did not survive its own plan

In Food for Thought 2.1, Keith was already losing the rhythm of a project he had designed for himself.

The idea had been ambitious in a very ordinary way: work through a collection of blogs alphabetically and publish a regular review. By late January 2009, events had overtaken the intended schedule. Keith was still reading and following links, but the organising routine was beginning to fail.

He did not yet turn that into a lesson.

He simply left the mismatch visible.

Three months later, in 090416 Gleanings, he returned to it more explicitly. He had aspired to write a weekly alphabet update. He had managed two editions.

The plan had not survived his reading.

Keith described his divergent thinking sending him on journeys that were difficult to stop. Then he gave those journeys a different value. He called them “wonderful vectors of discovery”.

The failed routine had not secretly succeeded. The weekly series was still a weekly series that did not happen.

What changed was Keith’s understanding of the divergence that had broken it.

The plan had assumed that the reading could be held inside a regular alphabetic structure. The reading kept producing other routes. When Keith returned to the failure, he valued those divergent journeys rather than treating them as a problem to be repaired.

The plan fails. The reading does not.

That distinction matters because it prevents a familiar retrospective trick. We do not need to say that the failure was necessary, or that every abandoned routine contains a better method waiting to be discovered.

The failure remains a failure of the routine. The later post changes what Keith makes of it.

A failed test is still a failed test

Failure looks different when the standards are external and explicit.

Goal records a 2011 stage in the long debate over goal-line technology in football. Ten systems had been tested. None met the criteria set for them.

The criteria were not Keith’s. The testing was not Keith’s. FIFA and the International Football Association Board owned the process and the judgement. Keith was recording a public state in a technological and regulatory argument he had been following.

That state was failure.

Accuracy mattered. Speed mattered. The result had to be delivered immediately. The system also had to be workable in the conditions the governing bodies had specified.

At that point the systems had not passed.

It would be easy, knowing what came later, to turn the post into the first act of a smooth technology story.

The archive resists that too.

Goal Line Technology Update July 2012 records a different state. After further testing, IFAB approved GoalRef and Hawk-Eye in principle, subject to final installation tests and the changes needed to the Laws of the Game.

The later approval does not make the earlier failure disappear.

Nor does the earlier failure prove that the technology should have been abandoned forever.

Between the two states are further tests and another institutional decision. The later route is visible because those intervening states survive.

But even the later approval needs restraint. Approval is not the same as proving that every later implementation was accurate, useful, affordable or superior to human judgement. It is another state in the history.

This is one reason failure becomes so misleading when it is told only from the endpoint. If we start with the eventual approval, the 2011 failure can look like an obstacle that was always destined to be overcome.

It was not.

At the moment of failure, the later decision did not yet exist.

Leaving unsuccessful work visible

Not every failure in Clyde Street acquires a dramatic sequel.

Exploring GitHub: Blue Skies and Stormy Seas is much smaller.

In March 2017, Keith spent a day connecting work he had encountered through Amber Thomas and Stephen Downes with his own experiments in GitHub. He had five repositories and was concentrating on a Portfolio repository. He was trying to create a place in which he could share his digital presence.

Then he recorded an unsuccessful attempt to use Dean Attali’s insights.

That is almost all the failure gives us.

Keith did not present a triumphant fix. The source does not show that the attempt later worked. It does not show that a reader benefited from seeing the unsuccessful state. It does not establish a principle of learning through public failure.

It simply leaves the unsuccessful attempt in the record.

Leaving it there changes the record.

By leaving the unsuccessful attempt there, the record does not collapse the day into working repositories and a finished result. The friction remains part of what a later reader can see.

He leaves it there.

He ends with a hope that his learning journey might become a resource for others who choose to share, contribute and co-create. Hope is the correct tense. The post gives us no basis for turning that possibility into an outcome.

Failure has not yet become a route to success.

It has become a visible state in the journey.

Owning failure is not proving its value

By late 2018, Keith was writing about failure much more directly.

Thinking about options and failure brings together work on choice, innovation and failed projects. Much of the substantive framework belongs to other people. Clare Thorp, Daniel Brannon and Brandon Solwisch, Sarah Milstein, Edwin Thoen and Dewi Koning remain intellectual owners of the arguments and examples Keith follows.

Keith’s own move is to bring those prompts into his experience of high-performance sport.

He writes that for much of his professional life he had been drawn to “good enough” approaches. The more he had worked in high performance, the more he wanted to discuss fallibility in the pursuit of optimisation — and to own failure as well as success.

He also argues for transparency about innovation decisions and their outcomes.

That is a stronger statement than the GitHub post, but it still needs a boundary.

The source does not show that a team became better because it discussed failure. It does not establish that morale improved, that innovation accelerated or that a repository of failed projects produced later success. The management and data-science frameworks Keith is reading do not become validated sport interventions simply because he finds them useful.

What the post does establish is Keith’s public stance toward unsuccessful work.

His position is that failure belongs in the account alongside success, and that innovation decisions and outcomes should be transparent.

That is different from saying failure should be celebrated.

Keeping failure visible does not decide what it means.

Sometimes the route is out

Paul’s PhD supplies the most important counterweight in this trail.

The source is Keith’s post, but the research is Paul Perkins’. Keith was celebrating a near-submission PhD and reproduced Paul’s abstract. The findings, programme decisions and claims belong to Paul and the research around the modified boxing programme, not to Keith.

Within that research, automated scoring technology was used early in the programme and then repeatedly adapted.

There were reasons to keep working on it. The technology attracted public attention. Solutions to identified problems were conceived and partly implemented.

There were also persistent problems: cost, logistics, accuracy and occasional system failure. More importantly, the technology affected the kind of boxing that contestants were rewarded for. It favoured all-out attack over skilled performance.

Meanwhile, the programme itself was changing. Safety, learning, community and cooperation had become increasingly important.

Eventually, the decision was made not to use the automated scoring technology to judge contests between individuals. It no longer fitted the direction of the programme.

This is not a story in which the whole project failed.

The programme had substantial participation, underwent repeated refinement and generated findings that Paul treated cautiously when considering transfer elsewhere. The relevant failed fit concerns one technology inside that larger piece of work.

The response was not endless perseverance.

The scoring technology was removed from that use.

That makes the case particularly useful here. It shows that the next move after repeated problems need not be another attempt to make the same technology work.

In Paul’s account, the programme’s changing aims and what had been learned about the technology made continued use harder to justify.

A route can lead out of the thing that was being tried.

What failure does not decide

Across these sources, failure has no single afterlife.

The Food for Thought routine is re-read; goal-line technology moves into another testing process; the GitHub attempt remains publicly unsuccessful; Keith later argues for owning failure as well as success; Paul Perkins’ scoring technology is eventually removed from that use.

These are different histories, not stages in a method of failing well.

The failure itself does not decide what comes next.

A failed state does not tell us whether to persist, retry, work around, abandon, remove, rethink or do nothing. Across these cases, what follows depends on the circumstances around the failure: further evidence, changing aims, practical constraints or a different judgement about what the work now requires.

Clyde Street also contains ordinary disruption that should stay ordinary. In Bird Strike At Taree, a bird strike damages a plane and delays Keith’s onward journey. The birds fare much worse. Nothing in the source requires us to turn the disruption into an allegory of resilience or productive failure. Something going wrong is not enough, by itself, to make the event intellectually productive.

That keeps ordinary disruption from being recruited into an argument it cannot support.

In the cases where failure does become a route, the route is visible because the record shows what followed, who made the next decision, and what evidence, purpose or constraint accompanied it.

And sometimes the record has no uplifting ending at all.

It simply says: this did not work.

Question to carry

What has to happen after a failure before it deserves to be called a route?

Sources followed

Keith Lyons, “Food for Thought 2.1” — 09/01/26

Keith Lyons, “090416 Gleanings” — 09/04/17

Keith Lyons, “Goal” — 11/03/08

Keith Lyons, “Goal Line Technology Update July 2012” — 12/07/07

Keith Lyons, “Exploring Github Blue Skies And Stormy Seas” — 17/03/10

Keith Lyons, “Thinking About Options And Failure” — 18/11/23

Keith Lyons, “Pauls Phd” — 19/04/17

Keith Lyons, “Bird Strike At Taree” — 11/01/10