Top-down Estimation
Top-down estimation, which PMI calls analogous estimation, prices the whole first, then apportions it across its components. The total is set early, from a high-level view, drawing on a comparable past project whose actual cost is known, or on expert judgement. That total is then adjusted for known differences and split across the components by proportion, percentages that sum to one hundred, so the parts gather back exactly to the whole. The technique is fast, cheap and rough, the right tool when the detail a component-by-component build would need does not yet exist. Its natural contrast is bottom-up estimation, which estimates every low-level element and sums upward. BABOK lists top-down estimation among the estimation methods, beside bottom-up, parametric, rough order of magnitude, the Delphi method and three-point estimation.
Goal
Top-down estimation starts from the whole rather than the parts: the cost or effort of the whole is judged in one stroke, in the light of a comparable project already run or of expert opinion, then that total is carried down to the components by a simple proportional split. It produces that number early, when nothing yet allows it to be built from the bottom and the commitment decision already calls for it: how many francs, how many person-days, how many months. The problem it solves is early costing: giving a defensible estimate at the moment when the only information available is an order of magnitude and the memory of what a similar piece of work actually cost.
The decision it supports is a judgement made upstream, often before a detailed breakdown of the work exists. Opening a feasibility budget, passing a decision gate, quickly comparing two or three solution options at portfolio level, all these decisions need a fast, cheap number, not a precision the project stage cannot yet deliver. The technique excels there: it turns a judgement about the whole into a costed, apportioned estimate in a few operations, where a component-by-component estimate would cost weeks of analysis the schedule does not grant.
The deliverable is a costed total, with its analogue and its assumptions, split across the components and stated as a range rather than a single point. Top-down estimation fills that deliverable from the top down, by analogy and proportion, where bottom-up estimation builds it from the bottom up by summing the detail.
Usage
When to use
- Feasibility or decision-gate costing: an initial budget is needed before any detailed breakdown of the work.
- Component detail unavailable: no basis yet allows an element-by-element estimate.
- Well-documented comparable project: a similar past piece of work exists, with clean, usable actuals.
- Order of magnitude defensible and enough: the decision is content with a fast, cheap number traceable to its analogue.
- Quick comparison of options: several options are separated at portfolio level without costing them in detail.
When not to use
- No usable analogue: nothing grounds the proportion, converge expert judgement through the Delphi method instead.
- A firm, definitive commitment is expected: too coarse for a contractual budget, build the number with bottom-up estimation.
- Component-level accuracy required: each separately contracted or risk-bearing part needs its own estimate, size the sensitive components with bottom-up estimation.
Description
The method: establish the whole, adjust it, apportion it
The technique runs in four steps, and its meaning lies in the direction of travel: from the whole toward the parts. The first step establishes the whole. You pick a completed, comparable project whose actual cost or effort is known, and take that actual as the starting number. A reliable analogue assumes archived actuals, total cost, effort and distribution: without those kept records the technique has no solid starting point. Where no usable analogue exists, a panel of experts sets this first-cut total from experience. What is compared matters more than the domain: two projects in the same sector can have opposite cost structures, and two projects in different sectors can share the same cost drivers. The first pitfall sits here, in the convenient analogue: a project that resembles the new one only on the surface imports its whole cost structure, and the estimate inherits a reality that is not its own.
The second step adjusts the analogue for known differences. A past project is never the exact twin of the new one: size, scope, complexity, team maturity, technology stack or regulatory load differ, and each of those gaps warrants correcting the starting total. PMI describes analogous estimating as a gross value drawn from a similar project, sometimes adjusted for known differences in complexity. This is where a 400-person-day analogue becomes a 480-person-day estimate for a scope judged twenty percent larger. The second pitfall is skipping this step: taking the analogue's total without correcting it amounts to copying a number, the adjustment is what makes it an estimate. It is the moment judgement enters an otherwise mechanical calculation.
The third step apportions the adjusted total across the components by proportion. You apply percentages, drawn from the analogue's own actual split or from an organisational cost-distribution model, so the shares sum to one hundred percent and the effort and cost gather to the whole. This is the apportion method: allocating a known total to segments by percentage. The third pitfall lies in those percentages: an apportionment grid nobody has revisited encodes an outdated way of working. The testing share shrinks as automation advances, the balance between back-end and front-end shifts with every platform change and a frozen ratio makes the estimate carry the structure of a project that no longer exists.
The fourth step states the result as a range. A top-down number is order-of-magnitude, and BABOK places that class, the rough order of magnitude, at roughly plus or minus fifty percent, where a definitive estimate, once the detail is known, tightens to ten percent or less. The fewer the data, the wider the confidence interval. The fourth pitfall is false precision: reporting four hundred and eighty person-days or CHF 576'000 as a definitive number when the method yields only an order-of-magnitude estimate. A total to the exact franc looks precise because it is written that way, and that apparent precision invites a commitment the technique cannot support. The range, with its assumptions, is what you deliver.
The analogue is a hypothesis
A top-down number reads as a starting point. It rests on the idea that the new work will behave like the old, corrected for what is known to differ, and that idea is a hypothesis to revise as soon as better data arrive. An analogous estimate set at the start of a project is meant to be refined later, by a bottom-up pass once the work breakdown is built, or by a parametric pass once a calibrated model is available. The proportional split, too, is a forecast: saying the back-end will weigh thirty percent of the total because the analogue weighed it so commits to nothing until a finer estimate confirms it. Holding the analogue as a hypothesis, and saying so, guards the technique against its commonest failing, which is letting a rough number harden into a commitment by mere repetition.
AI considerations
The calculation itself needs no artificial intelligence: an adjustment and a proportional split fit in a few spreadsheet cells. A model earns its place around the number, on the choice of analogue and the checking of assumptions.
Upstream, a model helps find the right analogue. Searching a project history for those whose cost drivers truly resemble the new work is research the machine shortens. It can rank past projects by profile similarity, surface the cost splits typical of a project family and flag that a candidate analogue strays too far from the population to serve as a basis. On the split, a model fed the organisation's actuals proposes a first grid of percentages the analyst revises and detects when that grid has drifted across recent projects.
What the machine must not decide belongs to judgement and to data. The adjustment for known differences, size, complexity, technology, regulatory load, remains an expert appraisal that the project context informs and a model cannot settle alone. The statistical similarity of two projects does not guarantee they share their true cost drivers, and an analogue chosen by an algorithm demands the same verification as one chosen by hand. Finally, the history of costs and effort is sensitive financial data: past-project actuals are not handed to an external tool without the controls that apply.
Examples
A retailer in the canton of Vaud is scoping the rebuild of its customer portal. A comparable portal, delivered and measured by a sister brand, cost 400 person-days. The new scope is judged about twenty percent larger, a loyalty module and a second language being added, which brings the adjusted analogue to 480 person-days. At a blended rate of CHF 1'200 per person-day, the top-down total comes to CHF 576'000, apportioned across the six components using the analogue's historical distribution.
| Component | Share | Effort (p-d) | Cost (CHF) |
|---|---|---|---|
| Scoping and requirements | 15% | 72 | 86'400 |
| UX/UI design | 10% | 48 | 57'600 |
| Back-end development | 30% | 144 | 172'800 |
| Front-end development | 20% | 96 | 115'200 |
| Data integration and migration | 15% | 72 | 86'400 |
| Testing and quality assurance | 10% | 48 | 57'600 |
| Total | 100% | 480 | 576'000 |
The total is fixed first, then divided. The share column is what the technique has of its own, and it is the one to read with suspicion. These percentages are the analogue's actual split: if the sister project spent thirty percent on the back-end, the new one is estimated to do the same, until a bottom-up pass confirms or corrects it. Nothing is measured on the new project: everything is inherited from the analogue and adjusted, which is at once the strength of the method, its speed and its limit, an accuracy bounded by how alike the two projects are.
Visualizations
Top-down estimation runs in one direction only: the whole is fixed first, the adjusted analogue, then split across the components by ratio.
Cost
| Phase | Level | Justification |
|---|---|---|
| Preparation | Moderate | This is the real cost of the technique. Finding a comparable analogue, drawing its archived actuals, total cost, effort and distribution, then adjusting it for known differences takes time and judgement. |
| Execution | Low | Once the analogue is set, the calculation fits in a few spreadsheet cells: a scaled total, then a proportional split. It is this speed that justifies the method at a stage where a detailed estimate would cost too much. |
| Documentation | Low to moderate | The total and its split fit on a page. The care goes into the assumptions, the percentage grid and the order-of-magnitude label, without which the band is misread and the number slides toward the status of a budget. |
Tools
A spreadsheet is enough to carry a top-down estimate: one cell for the adjusted total, a column of percentages, and the shares compute themselves. At this stage nothing heavier is warranted for the arithmetic itself, and richer tooling helps only around the number.
The quality of the estimate depends above all on the source of the analogue. A database of past projects, a repository of historical costs or a base of internal comparables are what turn a total set from memory into an analogue anchored on actuals; the richer and closer the history is to the work being estimated, the better the starting point. Estimation tools and project-management platforms that keep historical baselines make the analogue easier to find and its split reusable from one project to the next, which pays off above all when the same team estimates neighbouring work repeatedly.
Sources
- IIBA, A Guide to the Business Analysis Body of Knowledge (BABOK Guide) v3, §10.19 Estimation: the placing of top-down estimation among the estimation methods, the information sources it draws on, analogous situations, organisational history and expert judgement, together with the principle that fewer data give a wider confidence interval, hence stating the result as a range.
- PMI, A Guide to the Project Management Body of Knowledge (PMBOK Guide), 8th edition, analogous estimating: the method's own definition, a gross value drawn from a similar project through its historical data, sometimes adjusted for known differences in complexity, applied to a total or to its segments, less costly and less accurate than other estimation methods.