Why Work Management Fails

A global IT organization kept attacking its most visible problem, releases, one fix at a time. Sequencing its redesign on a single grid showed why that never worked: release work could not even begin until intake and prioritization were redesigned with it.

01 SURFACE
The condition

An IT organization had spent about two years trying to fix its release problems, and every fix faded because releases were where the trouble showed, not where it started.

The Problem Everyone Could See

Releases were where the trouble showed. In a global IT organization, three different ways of tracking a release were in use at once. Some teams used custom fields, some typed approval numbers in by hand, and some used tags. The approvals themselves lived in a separate system with no automatic link to the delivery work, so no one could reliably look back across the portfolio and confirm what had actually been released.

Inside the teams, releases had little structure. One member of the internal change team put it plainly: they pushed things out and moved on. The teams that held retrospectives did them informally, and whatever they learned stayed with the team that learned it.

Releases are the visible part of delivering value, so they are what leadership most often pushes to fix first. This organization had done exactly that, for roughly two years. The fixes did not hold.

One Fix at a Time

The organization was not short of effort or investment. It had a habit of adopting the latest framework, the newest tool, the next good idea, and layering each on top of what already existed instead of replacing anything. Each addition could work on its own. Together they did not connect.

By the end of the diagnosis, the gaps were visible and numerous, and the natural instinct was to work through them one at a time, starting with the most painful. We told the change team to resist it. A work system is a set of connected components. Close a gap at the front door without addressing what happens downstream, and intake cannot actually consolidate; the dysfunction simply moves into scoping, ranking, and decomposition. Worse, the early, visible improvement masks the unchanged reality behind it, and the organization concludes the redesign did not work when it was only too isolated for the problem it was trying to solve.

So the plan was built as five connected outcomes rather than a list of fixes: a simpler, visible way for work to enter; one forum for deciding priorities; one consistent way of using the delivery tool; a structured release practice; and reporting drawn from the work itself. How a long list of findings became those five is a story of its own.

Figure 1
Why an isolated fix looks like failure
  1. Fix the visible problem Releases, because that is where it shows
  2. Where it broke Upstream stays the same Work still arrives and shifts as before
  3. Where it broke Early gains fade Releases depend on what intake sends
  4. The trap Redesign judged a failure It was too isolated, not wrong
The visible problem gets the first fix. Because the conditions feeding it have not changed, the gain fades, and the redesign takes the blame for being too narrow.

A Row With Nothing to Start

The turning point came when the team sequenced the plan. We put the five outcomes down the left side of a grid and four columns across the top: now, next, later, and last. Every activity had to land in one column, according to what it depended on and what it would block.

Intake filled in quickly, with a baseline of how work entered today and the design of the new front door both in the first column. Prioritization opened with two activities running in parallel. Standardizing the delivery tool could start immediately too. Then the team reached releases, and the first column stayed empty. Releases could not be redesigned until intake and prioritization were stable enough to feed them the right work in the right order. Reporting looked the same: nothing could start until the system underneath it produced data worth reporting on.

The PMO leader watched the release row come together. It told her something, she said, about how they had been trying to fix release problems for the past two years. They had been fixing them in isolation, and they had kept failing because the conditions the fix needed were not in place.

That is the pattern behind most work management that keeps failing. Release performance is largely a function of what intake and prioritization send to the delivery teams. Attack it directly and you are treating the last symptom in a chain while its causes keep producing it.

Figure 2
The sequencing grid, simplified
NowNextLaterLast
IntakeBaseline; new designDemand to capacityMeasures; reportingRollout
PrioritizationParticipants; baselineForum; toolingFrameworkMeasure the forum
Delivery toolStandard setupEstimation; governanceConsistent useCollaboration
ReleasesNothing can startBaseline; governanceAdoptionMeasurement
ReportingNothing can startAssessment; needsData structuresReporting framework
Three outcomes could start at once. Releases and reporting could not, because both depended on work upstream of them. The blocked cells explained two years of failed release fixes.
A person moving a note across columns on a task board.

Five Parts That Hold Each Other Up

The design that followed treated the five outcomes as one system. Every request would enter through one of two paths, with its context, purpose, owner, and type filled in. Priorities would be decided in one shared forum against shared criteria. Every team would use the delivery tool the same way, with one structure running from yearly initiatives down to daily tasks. Reports would come straight from the work.

Releases came last, and they were built to depend on everything before them. One release path now runs from the system where changes are approved, through planning, into the delivery tool, so every approved change automatically creates its own release record. Nothing moves to production until that link is confirmed, downtime and rollback plans are written, test evidence is attached, and both the functional leader and the release manager have signed off.

The reasoning behind that order was the lesson the whole effort turned on. The five problems fed one another, so fixing any one of them alone would have left it at the mercy of the other four.

What Connected Parts Produced

In the pilot, three ways of tracking releases became one, and every deployment now traces back to the approved change behind it. When the prioritization forum commits to something, a release structure is ready to carry it, and what gets delivered feeds straight back into planning. Six ways for work to come in became two. Five reports built by hand became three dashboards that update themselves.

Not everything took hold at once. The forum was running, but it was not yet the place where every priority call got made, and the hardest remaining work was leaders using the new system visibly and consistently. Those limits are part of the point. The parts were designed to depend on one another, which means the system is only as strong as the part leaders actually use.

Work management rarely fails because an organization chose the wrong tool, framework, or fix. It fails because each fix lands on one part of a system whose parts depend on one another, and the parts nobody touched quietly undo it.

Figure 3
What the connected design produced
3 to 1ways of tracking releasesEvery deployment traced to its approval
6 to 2ways for work to come inPilot scope
5 to 3manual reports to dashboardsDashboards update themselves
Pilot-scope results. Each part depended on the others: the release path, for instance, works because intake and the forum send it work that is ready.

A fix that depends on conditions upstream will keep failing until those conditions are redesigned with it.

Share With Your Team

Choose who it is for. You send it yourself from your own email; Agileon never writes to your colleagues.

Not sure where to start?

Answer five questions about where work stalls. We send back a starting plan: where to begin and what to look at first.