The Estimate That Names Its Assumptions

‘Can we afford to index the games before the next lending season?’ Rosa asked. Farid could have answered with a tidy number. Instead, he asked what ‘index’ meant.

Their community equipment service had 180 board-game boxes. Staff wanted to find a particular box without opening several cupboards. They did not need an assessment of whether every playing piece was present. That distinction changed the work before a calculator entered the conversation.

This fictional case supplies the handbook’s complete worked example. Its people, times and costs are invented. CU means fictional cost units, not a currency conversion, market rate or recommended wage. The model is a simplified internal resource-cost calculation, not a commercial price, tax treatment, investment forecast or invoice.

The estimate has a purpose: help Rosa decide whether to authorise a bounded indexing job within a planning budget of 650 CU. It does not, by itself, authorise spending or prove that the work has been completed. An estimated amount, an approved budget and an actual recorded cost are different things.

Farid writes the proposed result in one sentence: ‘An index of 180 existing game boxes, with asset code, title, shelf location and one front photograph for each box.’ Rosa agrees that this is the result she needs.

They also write what the job excludes: counting pieces, repairing boxes, judging resale value, producing a public image collection or buying a new storage system. These exclusions are not small print designed to surprise Rosa later. They help both people judge whether the proposed result answers the original question.

For your own estimate, name its decision before calculating. Is it comparing two options, requesting a budget or checking whether an existing plan remains affordable? A number can be well calculated and still be unsuitable for the decision someone is trying to make.

Farid separates the work into setup, ordinary records and a final quality check. This is his simple structure for the example, not an official estimating standard.

Setup is two hours. Final checking is two hours. Ordinary recording is initially estimated at four minutes per box. The internal cost allocation is 30 CU per working hour. Supplies are 36 CU and a separate travel allowance is 24 CU. The last two items total 60 CU.

A minute is not an hour. The calculation must show the conversion: 180 boxes multiplied by four minutes gives 720 minutes, or twelve hours. Add two hours of setup and two hours of final checking: sixteen working hours. Sixteen multiplied by 30 CU gives 480 CU. Add the separate 60 CU: the first base estimate is 540 CU.

The travel allowance in this simplified model is a fixed allocation, not another number of working hours hidden in the calculation. The rate is an invented internal allocation; it is not presented as a complete real business’s labour, overhead, profit or tax model. A real estimate would have to identify the resources and obligations appropriate to that organisation.

Write units beside inputs, not just in your memory. ‘180’ could mean boxes, photographs, records or minutes. Here the scope requires 180 records and 180 front photographs, one for each box. It does not mean 180 verified sets of playing pieces.

A checkable calculation should let someone trace the total back to a particular quantity and rate. If Rosa changes the box count, Farid should not have to remember which of several unnamed spreadsheet cells contains it.

For a first review, ask another person to follow one path from a source input to the total. If they cannot tell whether the time is per item or for the whole job, the estimate needs clearer labels before it needs more decimal places.

Where did four minutes come from? Farid has timings for twelve related box records from an earlier fictional job. Their mean is four minutes. They were not a random sample of these 180 boxes, and they do not prove that every new record will take that long.

He writes the basis beside the input: ‘Four minutes per ordinary record, based on twelve earlier related records; current collection not yet timed.’ That is more useful than calling it ‘the standard rate’ when no standard has been established.

An assumption can be reasonable without being a fact. The important next question is what would happen if it were wrong. Farid’s initial model also assumes that all asset codes can be read directly from the boxes. That assumption is easy to overlook because it sounds like an ordinary feature of the job.

Rosa can check it. Before authorising the main work, she inspects the existing collection and finds thirty boxes whose asset codes are missing or illegible. The codes exist in an older internal list, but someone will need to look them up.

The scope has not changed: each record still needs its existing asset code. The evidence about how easily that code can be obtained has changed. Farid should revise the estimate rather than describe the extra effort as an unexpected new service Rosa ordered.

The U.S. GAO’s official summary of its 2020 cost guide includes assumptions, data and later updates among the elements of estimating. The full long guide was not reviewed for this handbook. Our small case is not a claim of GAO certification or compliance.

For your own assumptions, record a basis, a consequence and a possible check. ‘The files will be ready’ becomes useful when it says which files, who confirms readiness, and what part of the estimate depends on that confirmation.

Before revising the missing-code work, Farid examines the uncertainty in ordinary recording time. He keeps the count, setup, checking, rate and fixed allocations unchanged, and tries three, four and six minutes per box.

At three minutes, recording takes nine hours. With setup and checking, there are thirteen hours. Thirteen times 30 CU plus 60 CU gives 450 CU.

At four minutes, the base remains sixteen hours and 540 CU.

At six minutes, recording takes eighteen hours. With setup and checking, there are twenty-two hours. Twenty-two times 30 CU plus 60 CU gives 720 CU.

The three illustrative scenarios are therefore 450, 540 and 720 CU. They are not three supplier quotations. Nor is 450–720 a statistical confidence interval or a guaranteed minimum and maximum. Other assumptions, errors and events could move the actual cost outside those scenarios.

The point of changing one input is to see its effect clearly. If Farid changes the number of boxes and the hourly allocation at the same time, he cannot attribute the whole difference to recording speed. A combined scenario can be useful too, but it should say that several inputs changed.

The base is below Rosa’s 650 CU planning budget, while the six-minute scenario is above it. Hiding the higher scenario would make the base look more certain without making the job less uncertain.

The UK Infrastructure and Projects Authority’s cost guidance discusses ranges and the evidence behind assumptions. It is guidance for infrastructure projects, not a ready-made model for indexing board games. The general caution about uncertainty informs this example; the scenario choices and arithmetic are our own.

A practical review should ask why a scenario differs, not merely admire the width of a range. A decorative percentage added to a total cannot tell Rosa which part of the work needs checking.

The thirty missing codes now have a separate line. In this invented case, Farid checks the older list and allows two minutes per lookup. Thirty multiplied by two minutes is sixty minutes: one additional hour, costing 30 CU.

That hour is not another final quality check. The original two-hour check covers completeness and consistency of the finished records and photographs. The lookup line covers obtaining thirty existing codes before those records can be completed. Ordinary recording time is unchanged in this revision.

The updated base becomes seventeen hours at 30 CU plus the same 60 CU, or 570 CU. Add the same lookup hour to the low and high recording scenarios: 480 CU and 750 CU. The updated illustrative scenarios are 480, 570 and 750 CU.

Farid saves version two with a date and explanation: ‘Thirty asset-code lookups added after collection inspection. Other inputs unchanged.’ He preserves version one instead of quietly replacing it and claiming he had always estimated 570 CU.

This also tells Rosa what has not changed. There are still 180 boxes, one front photograph each and no piece-counting service. A separate request for two photographs per box would be a different scope decision. It is not secretly included in the missing-code adjustment.

Do not add an allowance twice because it appears in two conversations. If the same cost is already included in a rate or another line, name where it lives. Conversely, do not erase a real cost merely because you lack the data to calculate it confidently. Mark the gap and decide how to investigate or present it.

Versioning is useful when the reason survives with the number. An unexplained total of 570 CU is only thirty units more informative than an unexplained total of 540.

Rosa understands that 570 CU is the current base, not a promise that the higher scenario cannot occur. She authorises the stated indexing job with a 650 CU internal budget limit. They agree to review progress after sixty completed records, and to pause for a decision if the forecast exceeds that limit.

Approval does not erase uncertainty. A pause rule does not guarantee that every surprise can be detected before any cost is incurred. It does identify who must decide rather than allowing an estimator’s spreadsheet to become an open-ended instruction to continue.

At the sixty-record review, Farid compares the work remaining with the recorded time. He forecasts a total cost of 600 CU, still below the authorised limit, and reports that forecast to Rosa. The scope remains unchanged. He does not substitute an optimistic 570 merely to make the latest estimate appear exact.

If you use a similar review, specify what information will be available: completed units, recorded effort, remaining work and known exceptions. ‘Halfway through the week’ may be a useful meeting time, but it is not automatically halfway through the work.

Keep a cost forecast separate from a calendar promise. Seventeen working hours do not necessarily fit into two days if the person is available only for short periods, access is limited or another person must provide information. This example does not pretend to derive a delivery date solely from the cost calculation.

Likewise, internal cost approval is not a customer payment. The case contains no claimed sale, paid invoice or profit. The number answers the bounded resource question it was made to answer.

Farid finishes the index: 180 records with asset code, title, shelf location and one front photograph each. Rosa checks the agreed result. No claim is made that all game pieces were counted or the boxes valued. Those remained outside the job.

The fictional actual effort is eighteen hours: two for setup, thirteen for ordinary records, one for code lookups and two for final checking. At 30 CU an hour, that is 540 CU. Supplies are 36 CU and travel 24 CU, so the actual recorded resource cost is 600 CU.

Compared with the revised base of 570 CU, the difference is 30 CU. Ordinary recording took thirteen hours rather than the estimated twelve. The separate lookup hour matched the added allowance. The actual is 50 CU below the authorised budget of 650 CU, but that does not retroactively turn the estimate into a guarantee.

Farid keeps three things distinct: the 540 CU first base, the 570 CU revised base and the 600 CU actual. He adds the explanation to the record rather than rewriting both earlier versions to match the ending.

For your own estimate, a compact record can include the decision it supports; the included result and exclusions; inputs with units and bases; assumptions and checks; scenarios and limits; version changes; approval and review conditions; and the eventual comparison with actuals. These are original editorial headings, not a substitute for an organisation’s required financial controls.

You do not need to pretend that every difference is a personal failure. Nor should you use uncertainty to excuse arithmetic errors or hide omitted work. A useful comparison separates the cause: changed information, different actual effort, scope change, a missing cost or a calculation mistake.

Rosa can now use the index, and Farid has a better account of this job. He has not discovered a universal price per game. He has completed one bounded estimate, made its assumptions visible, revised it when evidence changed and compared it honestly with what happened.

Research consulted 3 November 2026. All people, service details, timings, 650 CU budget and costs are fictional. No real wage, price, tax treatment, earnings, purchasing or investment recommendation is supplied. The arithmetic was separately checked. Scenario ranges are not probability intervals or guaranteed bounds.

U.S. Government Accountability Office (2020), Cost Estimating and Assessment Guide, GAO-20-195G. Official HTML summary consulted for scope, assumptions, data and updates; the complete 476-page report was not reviewed. https://www.gao.gov/products/gao-20-195g

UK Infrastructure and Projects Authority (2021), Cost Estimating Guidance. Official introduction, scope and selected assumption, data and uncertainty passages consulted. Its stated focus is infrastructure; this independent small-case exercise is not a certified application of the full guidance. https://www.gov.uk/government/publications/cost-estimating-guidance/cost-estimating-guidance

Complete original short digital handbook in seven chapters and one research unit, not a printed page-count claim. Original worked example and editorial record; no copied source diagrams. English original and five complete reading translations prepared with AI assistance and checked for aligned meaning, not independently certified by native-human reviewers. Original generated English front and back covers; no ISBN or print-proof claim.