Estimation
Estimation forecasts the cost and effort of a course of action under consideration and sometimes its value, its risk or its benefit, to inform a decision taken under uncertainty. Two ideas connect the seven methods BABOK groups under this name. The first: an honest result is stated as a range rather than a single number, an interval with a minimum, a maximum and a probability, and that interval is the wider the less is known. The second: estimation is iterative, revised as information arrives and often calibrated against the organisation's past performance. Each of the seven methods sits at a different point on the trade-off between what is already known and the precision one can promise without self-deception.
Goal
An estimate provides the justification for a budget, a deadline or a size assigned to a piece of work. Without it, a team is handed budgets and schedules that are unrealistic, taken as given for want of any reasoning to support them. BABOK notes that a small team following a defined technique forecasts more accurately than a lone individual, and that revising the estimate as the elements sharpen brings the new knowledge into it. Estimation therefore serves as much to produce a number as to make it debatable: an explicit range, with the assumptions that carry it, can be contested and corrected, where a single number thrown out without justification is simply endured.
Two limits frame its use. An estimate is worth no more than the knowledge of what it quantifies: applied to poorly understood work, the most rigorous method produces a number that is precise and wrong. And confining oneself to a single method keeps unrealistic expectations alive, because each method has its domain of validity and its blind spot. That is what justifies choosing the method to suit the situation rather than adopting one out of habit and knowing to switch when the situation changes.
Which method to choose when
The factor that governs the choice is the amount of information available at the moment of estimating. Early on, when almost nothing is defined, only a wide range is honest and the fast methods suffice. Later, when the work is decomposed and measured, a costlier method tightens the range and commits to it. The seven methods line up along this axis, from the least informed to the most detailed, each with its mechanism, the data it demands, the precision it promises, its effort and the phase where it earns its place.
| Method | Mechanism | Data required | Typical precision | Effort | Project phase |
|---|---|---|---|---|---|
| Rough order of magnitude | A single, deliberately coarse number, paired with a wide range. | Almost nothing, an analogue or an expert opinion will do. | The widest, on the order of ±50%. | The lowest. | The earliest: pre-study, opportunity screening, the decision to commit to what follows. |
| Top-down estimation | Estimate the whole from a comparable past project, adjust it, then allocate it across the components pro rata. | A well-documented analogue, with actual costs. | Order of magnitude. | Low and quick. | Feasibility, a decision gate, before any detailed breakdown of the work. |
| Delphi method | Several rounds of anonymous questionnaires, the coordinator returns the group's median until convergence. | The judgment of a panel of experts, with no historical base. | That of the experts, cleared of rank and bandwagon bias. | High, several rounds. | A new or contested estimate, high stakes, experts dispersed. |
| Parametric estimation | A measurable driver multiplied by a calibrated rate, plus adjustment factors. | A stable, measurable driver, together with the organisation's own historical data to calibrate the rate. | Good and repeatable where the model holds. | Low once calibrated, the calculation reruns itself. | Early budgeting when one parameter dominates, comparison of options. |
| Three-point estimation (PERT) | Three values per component, optimistic, most likely and pessimistic, whose weighted mean and spread accumulate into a confidence interval. | A breakdown and experts able to name three credible values. | A quantified range rather than a point. | Moderate, the elicitation is its cost. | Definition, when uncertainty has to be given a number. |
| Bottom-up estimation | Decompose down to the smallest components, estimate each, then sum upward. | A fully defined deliverable, measurable activities, ideally historical unit costs. | The finest, to within 10%. | The highest, the deliverable must be defined before it can be summed. | A firm budget, a fixed-price bid or confirmation of a top-down number. |
| Rolling wave | Definitive estimation for the near term, order of magnitude for the rest, re-estimated at the end of each delivery. | A delivery in phases or in iterations. | Tight near, wide far, honest on both horizons. | Recurring, one re-estimation per wave. | Long or multi-year programmes, run in iterations. |
The seven methods combine and relay one another as information arrives. Rolling wave is the explicit example: it assembles a definitive estimate for the near term and an order of magnitude for the rest, two methods of the same family held together. A total obtained by top-down estimation is confirmed later by a bottom-up pass over a representative slice. An order of magnitude gives way to parametric estimation as soon as the driver becomes measurable. A Delphi method brings experts to converge on values that then feed a three-point calculation. Practice consists less in electing a method than in chaining them: start with the one the current information allows, then move to the next when there is enough to feed it.
Alongside this family, relative estimation, described in the Agile Extension to the BABOK Guide, compares the items of a backlog against one another on a unitless scale, story points, and the team's velocity then turns those points into iterations. A duration therefore appears only second, from the measurement of completed work, whereas the seven methods deliver a value in time, cost or effort from the outset.
Sources
- IIBA, A Guide to the Business Analysis Body of Knowledge (BABOK Guide) v3, §10.19 Estimation: the definition of estimation, the list of methods, the accuracy values (rough order of magnitude ±50%, definitive estimate to within 10%) and the confidence-interval frame.
- PMI, A Guide to the Project Management Body of Knowledge (PMBOK Guide), 8th edition: the anchor for the estimation family, with analogous estimating, parametric estimating, bottom-up estimating, rolling wave planning and three-point estimating under the name multipoint estimating, the five methods of the family it defines outright.