Milestone obsession: when local progress undermines project performance

An infrastructure programme can report a milestone achieved while becoming less likely to deliver a usable asset on time. Conversely, a work package can remain deliberately incomplete at an intermediate date without delaying the activities that depend on it. These possibilities are not contradictory: they arise because a project is a network of interdependent activities, not a collection of independent completion percentages.

The distinction matters when deciding whether to accelerate, resequence or defer work. The relevant outcome is not simply the earliest date attached to one event. It is the performance of the remaining delivery system, including safe execution, verified quality, resource feasibility and final acceptance.

This article examines that distinction through goal-setting research, construction production theory and schedule-assessment guidance. The scheduling example and decision tests below are original analytical illustrations, not findings from a measured Konstraff project.

Milestones are useful; incomplete performance measures are not

A scientific treatment should begin by rejecting the claim that ambitious targets are inherently harmful. Locke and Latham’s review of 35 years of goal-setting research describes how goals influence performance, and why their effects depend on mechanisms and moderating conditions. Clear objectives have a legitimate management function. 1

A competing concern is what the objective leaves out. Ordóñez and colleagues argue that intensive goal pursuit can narrow attention toward specified outcomes while neglecting other important dimensions. Their analysis also identifies potential changes in risk-taking and learning. This is a warning about the design and supervision of goals, not proof that every deadline produces poor decisions. 2

Applied to construction, the hypothesis is straightforward: when a reporting system gives an intermediate date disproportionate importance, a team may improve that date by transferring cost or difficulty to later work. The dashboard improves; the project does not necessarily improve with it. Neither study establishes a universal financial penalty for infrastructure milestones, so claims that milestone pressure always destroys margins would go beyond the evidence.

We use milestone obsession here to describe that specific management failure: treating success against an intermediate measure as sufficient evidence of success for the project as a whole.

The schedule must represent the work that remains

The GAO Schedule Assessment Guide distinguishes a logically calculated critical path from a resource-feasible programme. Conventional critical-path calculations assume unlimited resources; a schedule can therefore appear achievable while assigning the same scarce crew or equipment to incompatible activities. Resource constraints must be represented in the analysis. 3

Float also requires careful interpretation. Total float concerns delay relative to the relevant project or constrained finish; free float concerns delay before an immediate successor moves. An activity can have total float but no free float. Float is shared along a path, and GAO explicitly distinguishes it from schedule contingency. Missing logic or unrealistic constraints can make apparent flexibility misleading. 3

The resulting decision rule is not “finish everything early” or “defer anything with float.” It is to compare feasible sequences using the same scope, acceptance conditions and resource assumptions. A small activity may control access for a much larger one. Its cost or percentage of the contract does not establish its scheduling importance.

For the same reason, we recommend treating a proposed deferral as a schedule change to be tested, rather than a note to be placed beside the programme. Its remaining effort, dependencies and return-to-work conditions should be visible in the model.

Deliberate sequencing is different from unfinished production

Ballard and Howell place the quality of work assignments at the centre of production control. Their research emphasises assignments that are sound, appropriately sequenced and suitably sized, rather than simply releasing more work to keep production units occupied. 4

Koskela’s concept of making-do identifies a different condition: starting or continuing a task without the necessary inputs. Those inputs include information, personnel and equipment, not only physical materials. Apparent activity can therefore coexist with incomplete readiness. 5

There is empirical reason to take this distinction seriously. Amaral and colleagues’ observational study of nine construction projects in seven companies in Goiás, Brazil, identified 207 making-do occurrences. Rework was the leading reported impact, followed by deterioration in safety. However, the sample was not a randomised experiment, and the study did not test milestone policies. Its findings support concern about incomplete production conditions, not a causal estimate of the cost of deadline pressure. 6

For practical control, distinguish an intact package that is deliberately scheduled later from a partially executed package whose missing prerequisites have been ignored. In the first case, the later sequence can be planned and assessed. In the second, the team may have created additional work without knowing its extent.

This distinction also prevents loose labels such as “minor works” from replacing engineering judgement. Cleaning needed for inspection, grouting required for performance, or a record required before an irreversible operation cannot be treated as optional merely because it is small. The applicable technical and acceptance requirements must determine whether it can move.

A small scheduling model explains the trade-off

Consider three activities, measured in elapsed working days from Day 0. All values and conditions in this example are hypothetical. Durations are fixed, the necessary information is available, and the two work areas can be used independently.

Activity A takes four days and releases an installation area. Activity B takes three days in a separate area and uses the same crew as A. B is not a prerequisite for that release and is authorised to finish by Day 12. Activity C takes five days, uses a different crew, and can start only after A has been completed and its release accepted. Required verification is included in the stated durations.

In the second sequence, B and C proceed concurrently with separate crews in separate areas. The crew shared by A and B still performs seven crew-days of work. No task is deleted, no duration is shortened, and no acceptance requirement is relaxed. The three-day improvement comes entirely from the ordering of activities around a dependency.

The conclusion is conditional. If B is necessary for A’s acceptance, if B obstructs C’s access, or if both require an unmodelled crane or supervisor at the same time, the proposed overlap may not be feasible. A remobilisation cost or damage risk could also offset the benefit. And if the contractual milestone requires both A and B, A alone must not be reported as achieving that milestone.

The example demonstrates a scheduling mechanism. It does not demonstrate that deferral generally saves three days, reduces labour demand, or increases profit. Those outcomes require project-specific analysis.

Test the recovery window, not just the size of the backlog

A list of remaining items says little about whether the work can be completed. Ten short tasks for an available finishing crew may be easier to recover than one specialist activity requiring an unavailable possession, inspector or piece of equipment.

We propose a simple first screening measure for each resource group and recovery window:

Recovery load = required remaining resource-hours / realistically available resource-hours.

Use consistent units and include the relevant return visits, preparation, inspection and close-out effort. Available hours should exclude resources already committed elsewhere. Where distinct skills are required, calculate separate loads rather than treating all labour as interchangeable.

A value above one indicates insufficient capacity under the stated assumptions. A value of one or less passes this aggregate capacity check, but does not prove feasibility: enough hours over a month do not help when the required access exists for only one shift. Precedence, location and daily capacity still have to work together. This ratio is a proposed screening device, not an industry acceptance threshold.

Uncertainty then needs a separate treatment. GAO recommends schedule-risk analysis because variable durations, converging paths and correlated risks can affect completion confidence in ways a single deterministic date does not capture. A positive float value is not a guarantee that the remaining programme will finish on time. 3

For a significant carryover decision, test plausible adverse conditions: a missed access window, an unavailable specialist or an unresolved design response. Record whether the recovery plan still works and which intervention would be needed if it does not. The purpose is to expose fragility before the work is deferred.

Compare whole-project consequences before authorising a change

A useful commercial comparison considers both the disruption avoided now and the consequences introduced later. We recommend estimating the difference between the feasible alternatives, rather than assuming that an earlier release is automatically cheaper.

On one side sit the costs avoided by not interrupting a productive sequence or delaying a successor. On the other sit additional mobilisation, protection, access, supervision, reinspection and potential interference with subsequent work. Keep sunk costs out of the comparison unless they affect a future decision. Document uncertain costs as ranges or scenarios rather than presenting unsupported precision.

Safety, regulatory obligations and mandatory acceptance conditions are not ordinary cost items to trade away in this calculation. The analysis should compare alternatives that already satisfy those constraints. Contractual dates, notice requirements and authority to approve changes must also be checked against the actual contract; a programme revision alone does not establish permission to change them.

For each proposed carryover, we suggest one explicit decision record: what remains, why the sequence changes, who is responsible, when the work is needed, which resources and access are reserved, and what evidence will establish closure. Link that record to the live schedule and the relevant quality-assurance requirements.

A 4D planning review can help examine the spatial assumptions behind the proposal. It should support, not replace, the review of resource availability, acceptance criteria and schedule logic.

Measure whether the next stage is becoming more reliable

To complement milestone dates, we propose monitoring three different dimensions of completion.

Remaining demand: resource-hours still required, by discipline and access window, with revised estimates when scope or conditions change. This is more informative than an unweighted count of open items.

Verified readiness: whether each planned release has the agreed physical condition, approved information and inspection evidence needed by its successor. A release should count as ready only when all of its defined prerequisites have been checked, not when their average percentage looks satisfactory.

Recovery performance: planned versus verified closures, overdue recovery dates, and any new restrictions created by later work. A shrinking item count is not sufficient if the difficult items remain and their available windows are disappearing.

These are proposed management measures, not validated predictors with universal limits. Their value depends on clear definitions and consistent records. They should help the team challenge a favourable milestone report when the underlying delivery conditions are worsening.

The central distinction is between meeting a date and demonstrating a usable transition. Good milestone management connects the two. It does not reward finishing every activity prematurely, and it does not conceal unfinished work behind an achieved event. It makes the remaining obligations, dependencies and recovery assumptions explicit enough to be tested.

The objective is a more reliable path to completion, not a more attractive snapshot of progress.

References and evidence scope

The sources below combine a review of empirical goal-setting research, a critical management analysis, technical scheduling guidance, construction production research and an observational field study. They support the mechanisms discussed; they do not establish a universal optimal backlog or a measured financial effect for Konstraff. The worked example, recovery-load screen and reporting measures are original applications with stated assumptions.

1. Locke, E. A., & Latham, G. P. (2002). Building a practically useful theory of goal setting and task motivation: A 35-year odyssey. American Psychologist, 57(9), 705–717. doi:10.1037/0003-066X.57.9.705.

2. Ordóñez, L. D., Schweitzer, M. E., Galinsky, A. D., & Bazerman, M. H. (2009). Goals gone wild: The systematic side effects of overprescribing goal setting. Academy of Management Perspectives, 23(1), 6–16. doi:10.5465/amp.2009.37007999.

3. U.S. Government Accountability Office. (2015). Schedule Assessment Guide: Best Practices for Project Schedules (GAO-16-89G). See especially Best Practices 3, 6, 7 and 8. Report and full-text access.

4. Ballard, G., & Howell, G. (1998). Shielding production: Essential step in production control. Journal of Construction Engineering and Management, 124(1), 11–17. doi:10.1061/(ASCE)0733-9364(1998)124:1(11).

5. Koskela, L. (2004). Making-do: the eighth category of waste. Proceedings of the 12th Annual Conference of the International Group for Lean Construction. Original paper and abstract.

6. Amaral, T. G., Braga, P. B., Elias, K. V., & Brandão, C. M. (2021). Dynamic method to identify and analyze waste by making-do in construction sites. Gestão & Produção, 28(3), e5676. doi:10.1590/1806-9649-2021v28e5676.