There’s a saying on site that the programme is a work of fiction updated monthly. It gets a laugh because it’s often true — and it’s true because of how the plan was made, months earlier, by someone under tender pressure typing durations that felt right.
Good execution is mostly good planning arriving on schedule. Here is what two decades of towers, malls, and airport packages have taught me about doing it properly, and what it looks like when you don’t.
The plan is a set of decisions, not a drawing
A plan is the decisions underneath it: which floors first, which crane serves which elevation, how many crews, what productivity rate you’re assuming per crew, and what has to be true for the work to start at all.
Write those assumptions down next to the schedule. When reality disagrees with the plan later — and it will — the assumptions tell you why, which is the difference between recovering and thrashing.
Durations come from arithmetic, not optimism
If the tower has 1,850 m² of curtain wall and a crew installs 25 m² a day, two crews need 37 working days. Not “about a month.” Not whatever fits the milestone. Thirty-seven days, and now you can defend it — or resource it differently with your eyes open.
Every time I’ve inherited a project in trouble, I’ve found durations that no arithmetic supports. Someone needed the bar to fit, so it fit. The site paid for it later, with interest.
Procurement runs the programme, not the other way around
For a subcontractor, the programme’s real critical path usually runs through a factory, not the site. Glass at 10 to 14 weeks. Custom aluminium profiles at 8 to 12. Add approval cycles before any of it can be ordered — and approval cycles slip more than any site activity.
So plan backwards: installation date, minus delivery, minus production, minus shipping, minus the submittal cycle, gives you the date the shop drawing must go in. That date is usually terrifyingly soon. Better to be terrified in week two than in month six.
Break the work down until you can count it
A work breakdown structure isn’t paperwork; it’s how you make progress measurable. “Tower A façade” can’t be tracked. “Tower A, levels 1–10, curtain wall, 620 m²” can — you can count panels against it every single day. If an element in your plan can’t be measured with a number, break it down further until it can.
Execution is a weekly rhythm, not a monthly panic
The projects that run well all share the same heartbeat: a two-week look-ahead reviewed every week. What’s coming, and what could stop it — materials, access, approvals, the crane. Every constraint gets an owner and a date. Most delays I’ve seen were visible in a look-ahead ten days before they landed; the teams that hold the rhythm catch them there, cheaply. The teams that skip it meet the same problems on the day, expensively.
The second half of the rhythm is measurement. Daily quantities, logged by the people doing the work, rolled up weekly against the plan. If progress data arrives only at month-end, every problem is already four weeks old before anyone in charge sees it.
Guard the scope in writing
Nothing erodes a project like unrecorded change. The site does “one small extra” verbally agreed, then another, and at final account nobody can reconstruct what was instructed by whom. The rule is simple and must be absolute: no change without a record — an instruction reference, a change request, an entry in the log, same day. Not because you distrust anyone, but because memory is not a contract document.
Planning never actually ends
The baseline is a photograph of what you believed on day one. Keep it — you’ll measure drift against it — but keep planning: recovery plans when you slip, revised look-aheads, updated forecasts. The plan that matters is always the current one.
Every document this article mentions — WBS, look-ahead, progress tracker, change log — exists ready-made in the Project Mate library, built the way this article describes because they came from the same projects. Start with the tracker and the look-ahead; those two carry more weight than everything else combined.