Parametric vs Analogous Estimating
Both techniques produce a number quickly, early, before anybody could possibly build up an estimate from the actual work. That is their purpose and it is a legitimate one: projects have to be funded before they are designed, and a top-down figure with its basis stated is a great deal more useful than a refusal to estimate.
Section 5 of the PMBOK® Guide Eighth Edition includes both among the estimating techniques available, and choosing between them comes down to what evidence you actually hold.
What each one relies on
Analogous estimating relies on a comparable whole. The last job of this kind cost this much and took this long, this one looks about a quarter bigger, so here is the figure. It needs somebody who genuinely knows both jobs, and its accuracy lives entirely in the comparability: an analogy drawn by a person who worked on the previous project is worth several times one drawn from a spreadsheet of past totals.
Parametric estimating relies on a measurable relationship. Cost per square metre of fit-out, hours per metre of harness, days per kilometre of duct, pounds per kilowatt installed. It needs a driver that genuinely drives the cost and a dataset broad enough to support the rate, and when both exist it is the stronger of the two, because it scales and because the assumption is visible as a number somebody can argue with.
Where each one breaks
Comparability drift. The previous job is remembered as it was planned rather than as it happened, or it differed in a way that nobody thought worth mentioning: a different site, a client who made decisions quickly, a team that had just done three of them. Analogies are most confident when the person drawing them knows one of the two jobs well and the other only by reputation.
Extrapolation past the range. A rate derived from jobs between forty and ninety metres says almost nothing about a job of three hundred, because at some point the work changes character: access becomes the constraint, or a different installation method applies, or the thing no longer fits on one trestle. The arithmetic gives no warning, and this is the most common way a parametric estimate goes badly wrong.
A driver that does not drive. Cost per square metre works for fit-out until the driver is really the number of rooms, because partitions, doors and services scale with rooms and not with area. Where the relationship has never been tested, a parametric estimate is an analogous estimate with better presentation.
A harness rate and three hundred metres
An aerospace manufacturer estimated harness installation using an hours-per-metre rate, built from installations on an existing platform. The rate was sound, it had been refined over several years, and it had predicted the last four programmes within a few per cent.
A derivative variant came along with a substantially larger harness: about three hundred metres against a historical range of forty to ninety. The estimate was produced the usual way, the rate was applied, and the number went into the programme budget.
The installation took about seventy per cent longer than the estimate. Nothing had gone wrong in the workshop. At that length the loom no longer dressed out on a single trestle, so it was built in three sections and joined, which added interface work and two inspection points. Routing it through the airframe needed two people at positions that had previously needed one, because the weight made it unmanageable alone. And the branch count had risen faster than the length, which the rate had no way to capture because branch count had never varied much in the source data.
What the manufacturer changed was the estimating process rather than the rate. Every parametric estimate now carries the range of the underlying data on the same page, and any application outside that range requires a note saying why the relationship is expected to hold. On the following programme that note could not be written for the harness, so a bottom-up estimate was built from the routing drawings, which took three weeks and came within nine per cent.
The old rate is still in use, and it is still good, for harnesses between forty and ninety metres.
Using both honestly
State the basis with the number. What it was derived from, what range the source data covered, what was assumed about comparability, and how confident the estimator is. An estimate travels far from its author, and without the basis attached, a figure produced as an early indication will be quoted six months later as a commitment.
Replace them when definition allows. Both techniques exist because the work cannot yet be described in detail. Once it can, a bottom-up estimate is more accurate and the top-down figure becomes a cross-check, which is a genuinely useful role for it: a bottom-up total that differs sharply from a sound parametric estimate is telling you something about one of the two.
For a PMP® candidate, the reading that helps is that both are top-down techniques resting on evidence about past work, so a scenario where an early estimate proved badly wrong is asking what the estimate was based on and whether that basis applied. A response that treats the overrun as a delivery failure has skipped the question. A structured PMP exam preparation course gives real room to situations where the technique was applied correctly and outside its range.
Find the largest single estimate in your current budget and ask where its number came from. If the answer is a rate, ask what range of jobs the rate was built from, and where your job sits in that range. If nobody can tell you, that estimate is carrying more uncertainty than the budget shows.
Interested in going further?
Early estimates decide what gets funded and are rarely revisited with the same seriousness later. Omega's PMP® Exam Preparation works through estimating techniques and the evidence each one needs.
The estimating techniques behind both approaches are described in the PMBOK® Guide Eighth Edition.
Ad · Amazon affiliate link.
References
A167: Project Charter Explained: Purpose, Content and Common Misunderstandings
A168: Business Case vs Project Charter: What’s the Difference?
A169: Assumption Log vs Risk Register
A170: Decision Log: The Forgotten Project Control
A171: Change Log vs Issue Log
PMP and PMBOK are registered marks of the Project Management Institute, Inc.
