Plan before everything is settled - start rough, get measurably better
Estimate in the unit that matches the uncertainty, keep slack visible, test assumptions safely and learn from every actual: planning as a learning process, not a promise.
From rough to precise
Five estimation units from T-shirt to hours - one per project, enforced server-side.
Why early planning fails today
Rarely for lack of knowledge - mostly because not-knowing isn't allowed.
False precision to two decimals
Planning demands hours although nobody knows more than an order of magnitude. So numbers get invented - and then defended.
The slack is invisible
Where the plan has air and where it doesn't shows in no bar. Things get moved on suspicion - often in the wrong place.
Trying means breaking
Whoever wants to test an assumption changes the real plan. Afterwards nobody knows what was baseline and what was experiment.
Progress is a matter of taste
“Are we at 60 or 80 percent?” Every work package is measured differently - and none is comparable with another.
From rough to precise - planning as a learning process in five stations
This is how orders of magnitude become a plan that holds: fitting units, open buffers, a sandbox for assumptions, clean progress measurement - and actuals that improve the next estimate.
The unit adapts to the uncertainty - not the other way round
At the start, T-shirt sizes or Fibonacci points from 1 to 21 are enough; later they become hours or days - five estimation units are available, and exactly one applies per project: the server rejects deviating estimates so numbers stay comparable. And on the card, a buffer computes automatically from the time window and the effort - no side calculation.
Card: data migration - the same work in three units
One unit per project - the server rejects deviating estimates.
And the card does the maths: buffer from time window minus effort, automatically.
The plan shows its own tolerance
The network calculation delivers earliest and latest dates per task - and with them the corridor it may move in. Total float sits openly in the plan instead of hiding as padding in the estimates; where the corridor shrinks to zero, the critical path runs. Fixed dates set hard anchors amid the estimated.
Earliest to latest date per task - the leeway sits visibly on the bar.
Test assumptions without damaging the plan
In planning mode you enter probe values - a longer duration, a later start - and watch the effect ripple through, while the real plan stays unchanged. A bar counts the modified tasks; at the end you adopt in one go or discard everything. Frozen states are archived by the baseline - browsable back to the first draft.
The real plan stays untouched until you decide - baselines archive every frozen state.
Measured the way the work package allows
Seven progress methods per container: manual 0-100, quarter steps, 0/50/100, 0/100, custom - or derived from the Kanban cards or the sub-containers. New containers inherit the project default automatically. Progress becomes a measurement question instead of a matter of taste - and packages stay comparable.
WP 3.1 - Training · choose progress method
Seven methods per container - new containers inherit the project default.
Every actual makes the next estimate better
Booked hours flow back to the work package as actual effort and sit right next to the planned value - analysed per person, team and project up to budget context. The deviation is not an accusation but your teacher: the next package gets planned one notch more realistically.
Plan and actual on the same work package - the deviation is the teacher, not the accusation.
And where uncertainty becomes risk, risk management takes over: three-point estimates and Monte Carlo simulation quantify the necessary reserve on data.
One learning process, five modules - one data foundation
In WORKSPACE.PM, heuristic planning is not a slide method but the platform working in concert: estimates, buffers and actual hours live on the same work package.
Project management
Resources & time
AI & automation
One data foundation - from estimate to actual
What you learn about your project stays in the system, not in someone's head: estimate, buffer, progress and actual effort hang on the same work package.
And if we don't know yet?
That's exactly what this way of working is for - rough is allowed, arbitrary is not.
Rough is allowed - chaos is not
One estimation unit applies per project, and the server enforces it: T-shirt sizes stay comparable with T-shirt sizes. Orders of magnitude yes, number sprawl no.
Refine without rebuilding
Placeholders first, names later; an AI structure draft as a starting point; probe values in planning mode. The plan grows with the knowledge - wave by wave, without starting over each time.
The estimate stays with you
There is no AI auto-estimation - deliberately. The AI drafts structure and proposes milestone dates; efforts and commitments remain a decision of your team.
Related use cases
From rough to precise - this is what comes next.
Frequently asked questions about heuristic planning
What project leads and teams want to know before switching.
Five: story points by Fibonacci (1-21), linear story points (1-10), T-shirt sizes (XS-XXL), hours and days. Per project you pick one unit - and the server rejects deviating estimates so all values stay comparable.
No - deliberately not. The AI drafts structures from a briefing and proposes milestone dates; automatic effort or duration estimation doesn't exist. Estimating remains a team decision - the platform makes sure it is documented and comparable.
In waves: T-shirt sizes and skill placeholders instead of names first, an AI structure draft as a starting point, probe values in planning mode. With every wave you refine - the unit gets finer, placeholders become people, and the baseline freezes each state.
From the plan itself: the network calculation shows the corridor between earliest and latest date and the total float per task, and every card carries its automatically computed buffer from time window and effort. Reserves are visible - not hidden inside estimates.
That's risk management's job: risks carry three-point estimates, the Monte Carlo simulation runs 10,000 iterations, and you adopt the recommended reserve as a cost position in the WBS.
Start rough - get measurably better
Start for free: pick a unit, estimate the first orders of magnitude, make buffers visible - and in 30 days you'll know whether it fits.