VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

How Estimates Fail | Why False Precision, Optimism and Hidden Work Break Plans

Three secondary students review open books and notes together in a classroom.

How estimates fail is not mainly a mathematics problem. It is a representation problem.

We take something uncertain — a task, a project, a revision plan, a cost, a duration, a future workload — and compress it into a number. That number is useful because decisions need a working expectation. It becomes dangerous when the number begins to look more certain than the evidence that produced it.

An estimate is not a promise. It is a model of what may happen under stated assumptions.

Good estimation makes uncertainty visible enough to improve planning. Bad estimation hides uncertainty behind precision, optimism or authority.

This master guide explains the failure architecture. The three companion explainers go deeper into reference class forecasting, estimate ranges and calibration.


The Core Idea: An Estimate Is a Decision Tool, Not a Fact

A useful estimate helps someone choose, schedule, budget, compare, reserve capacity or decide whether a plan is feasible.

The number is not the goal.

The decision is the goal.

This changes how estimates should be judged. A beautiful number with no connection to the decision can be less useful than a rough range that exposes the relevant uncertainty.

The United States Government Accountability Office treats reliable cost estimating as a disciplined process that documents assumptions, data, risk and uncertainty rather than as a single-number exercise. NASA’s Cost Estimating Handbook similarly treats uncertainty and risk as central parts of an estimate, not as embarrassing details to hide.


Failure Mode 1: False Precision

A duration of 47 minutes can look more intelligent than ‘about 45 to 60 minutes’.

It may actually be weaker.

Precision should be earned by evidence. If the underlying task is unfamiliar, variable or poorly defined, extra decimal places do not create knowledge.

False precision is dangerous because it changes behaviour. People schedule too tightly, allocate too little buffer and interpret ordinary variation as failure.

A better question is: how much precision does this decision actually require?


Failure Mode 2: The Inside View

The inside view estimates from the details of the current case.

We imagine the steps, assume they go approximately as planned and add the pieces.

This feels responsible because it is detailed.

It can still be systematically optimistic.

The missing information is what happened to comparable cases in the real world.

HM Treasury’s Green Book guidance on optimism bias recommends using historical forecast errors and evidence from similar proposals to correct systematic underestimation of costs and duration and overestimation of benefits.

That outside-view mechanism is developed in How Reference Class Forecasting Works.


Failure Mode 3: The Task Is Not Yet Defined

People often estimate labels rather than tasks.

‘Essay’, ‘revision’, ‘website update’ and ‘school project’ are containers. They can hide very different quantities of work.

A student who writes ‘History project — two hours’ before opening the brief is estimating a title, not a task.

eduKateSingapore’s article on diagnosing before estimating makes this explicit: a short early encounter can reveal the output, dependencies, unfamiliar skills and first useful action before time is committed.

Read Why Students Schedule Work Before They Know What the Task Is.


Failure Mode 4: Hidden Work Is Omitted

The visible task is rarely the whole task.

‘Complete the worksheet’ may include finding materials, reading instructions, remembering methods, getting stuck, checking notes, correcting errors and preparing the submission.

‘Publish the article’ may include collision checking, sourcing, drafting, internal linking, QA, metadata, saving, verification and post-publication confirmation.

Omitted work is one of the most common reasons a plan overruns even when the visible activity itself was estimated reasonably.

A strong estimate asks what must happen around the core task, not only inside it.


Failure Mode 5: Decomposition Creates an Illusion of Completeness

Breaking work into parts usually improves estimation.

But a detailed list can still omit dependencies, coordination, setup, rework and integration.

eduKateSingapore’s Work Breakdown Structure guide uses the 100 Percent Rule to make completeness a deliberate test: the children of a deliverable should collectively represent the parent scope.

Read Work Breakdown Structure | How to Decompose a Project Correctly.

Decomposition helps only when the parts still represent the whole.


Failure Mode 6: Dependencies Are Treated as Independent

If Task B cannot begin until Task A finishes, adding their durations is not enough.

If three tasks require the same person, they may compete for one resource.

If one decision blocks five downstream activities, its delay can dominate the schedule.

Estimation therefore needs structure, not only totals.

A set of accurate local estimates can still produce a bad global plan when dependencies are wrong.


Failure Mode 7: Setup and Switching Costs Disappear

Work does not begin at full speed the moment the calendar block starts.

Files must be opened. Materials must be found. Context must be rebuilt. Instructions must be reread. The mind must switch from one subject or task to another.

These setup costs are small individually and expensive in aggregate.

eduKateSingapore’s homework-estimation guide separates focused task time from elapsed time and makes setup, switching and correction visible as real consumers of the evening.

Read Why Students Underestimate How Long Homework Takes.


Failure Mode 8: Rework Is Assumed Away

Many estimates silently assume first-pass success.

Real work includes correction, rejection, misunderstanding, retesting and revision.

If quality matters, rework is not an accidental add-on. Some probability of rework belongs inside the estimate.

The right amount depends on task maturity and evidence, but pretending it is zero is rarely neutral.


Failure Mode 9: The Best Case Becomes the Plan

People often remember the fastest plausible path and promote it into the expected path.

That is not estimation. It is selecting one favourable scenario.

The current UK Green Book explicitly warns against optimism bias in estimates of cost, benefits and duration and recommends empirical adjustments based on historical forecast error.

A useful plan distinguishes what is possible from what is probable.


Failure Mode 10: A Point Estimate Hides the Shape of Uncertainty

‘Three hours’ can mean many different things.

  • Almost always between 2.8 and 3.2 hours.
  • Usually around 2 hours but occasionally 8.
  • Anywhere between 1 and 5 depending on one unresolved dependency.
  • Three hours if the familiar route works; much longer if new research is required.

The same point estimate can sit above very different risk distributions.

Decision quality improves when the estimate communicates the range and the reason for it.

That is the focus of How Estimate Ranges Work.


Failure Mode 11: Uncertainty Is Added as a Decorative Buffer

Adding 20 percent ‘just in case’ can be better than pretending uncertainty does not exist.

But arbitrary contingency can hide weak reasoning.

A buffer should correspond to identifiable uncertainty, historical error or a stated confidence target.

NASA’s estimating guidance uses structured risk and uncertainty analysis, including ranges and probability distributions, because the point is not merely to add padding. The point is to represent what could move the outcome.


Failure Mode 12: Double Counting Risk

The opposite failure also occurs.

Risk may already be included inside task estimates, then added again as contingency, then added again as management reserve.

The final number looks conservative but the logic is opaque.

A good estimate keeps base work, known risk, residual uncertainty and contingency conceptually separate.


Failure Mode 13: The Estimate Is Used as a Commitment

The moment an estimate becomes a promise, people become reluctant to update it.

New information arrives, but the old number remains because changing it feels like admitting failure.

This is a governance problem.

An estimate should become more accurate as evidence improves.

If the system punishes updating, it rewards stale numbers.


Failure Mode 14: Scope Changes but the Number Does Not

A changed task requires a changed estimate.

If output requirements, quality standards, dependencies or resources change, the old estimate no longer represents the same object.

Preserving the number after changing the scope creates fictional precision.

eduKateSingapore’s Project Estimation guide treats change control and re-estimation as linked: approved scope change should trigger estimate impact analysis where relevant.

Read Project Estimation | Effort, Duration, Cost & Uncertainty.


Failure Mode 15: The Estimate Ignores Capacity

Duration is not only about how much work exists.

It is also about who can do it, when, with what energy and alongside which other obligations.

A student may have four nominal hours available after school and far less high-quality attention.

A tutor may have one lesson slot and several competing needs.

A team may have ten tasks and one specialist.

Capacity turns effort into calendar time.


Failure Mode 16: Every Task Is Estimated as If It Were Familiar

Familiar work and exploratory work behave differently.

For familiar work, past data can be strong.

For unfamiliar work, the first phase may be discovery: learning what must be done before estimating the rest.

A mature system does not force both into the same precision.


Failure Mode 17: The Error Direction Is Ignored

An estimate can be wrong high or wrong low.

Those errors may have different consequences.

Underestimating an exam essay can leave no editing time. Overestimating it can waste planning capacity but may be less damaging.

Underestimating a project budget can threaten feasibility. Overestimating can block worthwhile work.

The estimate should therefore be designed around the cost of being wrong in each direction.


Failure Mode 18: The Decision Threshold Is Hidden

Sometimes the exact estimate matters less than whether it crosses a boundary.

Can the task fit before bedtime without sacrificing sleep? Can this article batch fit inside the publication window? Can the revision plan cover the essential weaknesses before the exam?

When the decision is threshold-based, estimate effort should concentrate near the threshold.

A range that sits entirely below the limit may be enough. A range that straddles the limit deserves deeper investigation.


Failure Mode 19: The Estimate Is Never Compared With Reality

An estimate that is not scored cannot learn.

People often remember dramatic misses and forget ordinary ones. That makes intuition hard to calibrate.

Record the prediction before the outcome. Then compare.

Were estimates systematically low? Too wide? Too narrow? Better for familiar tasks than unfamiliar ones?

This feedback loop is the focus of How Estimate Calibration Works.


Failure Mode 20: Accuracy Is Judged From One Case

One lucky estimate does not prove a good estimating process.

One bad miss does not prove incompetence.

Calibration is a property of repeated predictions and outcomes.

The pattern matters.


The Three Pillars of Better Estimation

1. Use the Outside View

Ask what happened to comparable cases before trusting the story of the current case.

Read How Reference Class Forecasting Works.

2. Express Uncertainty as a Range

Show how much the outcome can move and why.

Read How Estimate Ranges Work.

3. Calibrate Against Outcomes

Compare predictions with reality and update the method.

Read How Estimate Calibration Works.


A Practical eduKateSG Estimate Test

  • What exactly is being estimated?
  • Is the scope sufficiently defined?
  • Which comparable cases exist?
  • What hidden work surrounds the visible task?
  • Which dependencies can dominate the outcome?
  • What assumptions matter most?
  • What range is plausible?
  • What confidence or decision threshold matters?
  • What buffer is evidence-based rather than decorative?
  • Who has the capacity to do the work?
  • When will the estimate be updated?
  • How will predicted and actual outcomes be compared?

If these questions are visible, the estimate becomes a working model rather than a ceremonial number.


Example: Student Homework

A student estimates that Mathematics homework will take thirty minutes.

A stronger method records the task type, number of questions, topic familiarity, setup time, expected correction time and recent outcomes for similar homework.

If the last four comparable assignments took 45, 50, 35 and 55 minutes, a new thirty-minute promise needs evidence.

The outside view does not eliminate judgment. It disciplines it.


Example: English Examination Writing

A student may estimate how much time an essay paragraph deserves.

SETC’s exam time-management work shows why the whole paper matters: planning, drafting, finishing and editing compete for the same fixed clock.

Read Essay Time Management for English Exams.

A paragraph estimate is only useful if it protects the final state of the whole essay.


Example: Mathematics

Mathematical estimation has a different but related role.

Before exact calculation, the learner predicts magnitude or range so the final answer can be challenged by independent evidence.

Bukit Timah Tutor’s Primary Mathematics guide develops estimation as an answer neighbourhood rather than a replacement for exact work.

Read Primary Mathematics: Estimation, Reasoning and Checking.


Example: Revision Planning

A revision plan estimates how much learning work can fit into a week.

eduKate Punggol’s exam-preparation and A-Math planning material repeatedly protects buffer, recovery and capacity rather than filling every visible hour.

Read Exam Preparation in Singapore and How to Balance Additional Mathematics With School, CCA and Other Subjects.


Example: Academic Recovery

Recovery estimates should be especially cautious because the first diagnosis may change the plan.

eduKate Yishun’s recovery architecture starts by identifying the first broken dependency and choosing the smallest justified intervention rather than estimating an entire rebuild from a vague label such as ‘weak in Mathematics’.

Read Academic Recovery | Repair the First Broken Dependency.


Across the eduKate Ecosystem

eduKateSG owns the general estimation mechanism. eduKateSingapore carries project estimation and knowledge-library depth. eduKate Sengkang applies estimates to study planning. eduKate Punggol applies them to family and exam-year capacity. Bukit Timah Tutor develops mathematical estimation and verification. SETC applies time estimation to English exam execution. eduKate Yishun applies the same evidence-first logic to recovery and rebuild decisions.


Sources and Further Reading

U.S. Government Accountability Office — Cost Estimating and Assessment Guide

NASA — Cost Estimating Handbook

HM Treasury — The Green Book 2026

HM Treasury — Optimism Bias Supplementary Guidance


Continue the Series

How Reference Class Forecasting Works | Estimate From Comparable Outcomes Before Trusting the Inside View

How Estimate Ranges Work | Represent Uncertainty Without Hiding Behind False Precision

How Estimate Calibration Works | Compare Predictions With Outcomes Until the Method Learns

eduKateSG