A milestone is a defined point in a project at which a particular, verifiable state must have been reached. Unlike a task it has neither duration nor effort — it is a moment, not a work package.
Its actual purpose
Milestones are often understood as date markers. Their purpose is different: they are decision points. At a milestone, what has been achieved is checked against expectations, and a decision follows — continue as planned, continue with adjustment, or stop. A milestone without that option to decide is merely a date in a calendar.
What makes a good milestone
A usable milestone is unambiguously verifiable: you can establish without discussion whether it has been reached. "Concept completed and approved by the client" is verifiable. "Concept phase largely complete" is not. That unambiguity is the difference between a control instrument and a statement of intent.
Phase gates
In formal process models, milestones are shaped as phase transitions at which the client side explicitly approves continuation. That is particularly common in public-sector and regulated settings and has a double value: it forces a deliberate decision, and it documents that one was taken. On large, risky projects in particular, that structure prevents a project running on uncontrolled.
Why too many become ineffective
A project with thirty milestones effectively has none. When one falls due every week, the check becomes routine and the decision a formality — nobody stops at milestone seventeen. What works is a few, at substantively meaningful transitions: where the risk profile changes or where a substantial investment is released.
Handling a missed milestone
A missed milestone is not a catastrophe but exactly the information it was set up to provide. The damaging response is to move it quietly. The right one is to name the deviation, establish its cause and adjust the plan. Anyone moving milestones instead of deciding has abolished the instrument without noticing.
Practical consequence
Three to five milestones with an unambiguously verifiable state and a real option to decide steer a project better than a long list of dates. The useful test question is: what happens if this milestone is missed? If the answer is nothing, it is not one.