◷ Timeline Thinking · Whitepaper

Temporal blindness:
why work slips out of control

Duration, simultaneity, horizon and your own history go missing when time is a field on a record rather than the surface everything sits on. This document states the problem, defines the model that treats time as the carrier, and describes the mechanisms, the cost, the risks and the way to check whether any of it worked.

About this document

What it is, and what it is not

Genre An analytical document — a whitepaper.
Audience Team and department leads, process owners, and independent practitioners evaluating the approach before applying it.
What it provides A statement of the problem, a description of the model, the mechanisms behind the effect, a pilot programme, metrics, risks, and requirements for a tool.
What it does not contain A product specification, a commercial offer, or promises of productivity.
Companion document The Timeline Thinking manifesto — the declarative version of the same ideas.
Last updated

How to read this. The document stands alone: it requires no acquaintance with the manifesto and does not restate it. The manifesto answers the question “what do we believe”. The long-form account of the approach answers “how is it built and what do I do”. This document answers “what problem does it solve, by what mechanism, what does it cost, and how do I check that it worked”.

Contents

Thirteen sections and four appendices

Abstract

The deficit and the shift

Tools for organising work — task lists, calendars, kanban boards, trackers — inherited a common data model in which time is an attribute of a record: a task has a due date, an event has a date. That model answers the question “what do I do next” and is structurally incapable of answering “how long will this take”, “what will it collide with” and “how long did work like this actually take before”. We call the resulting deficit temporal blindness: the systematic invisibility of duration, simultaneity, horizon, and one’s own history.

Temporal blindness does not feel like missing information. It feels like a run of surprises: “everything landed in the same month”, “we estimated three times more optimistically than it turned out”, “I came back to something I had parked and lost the context”. Each surprise gets a local explanation, so the systemic source goes unnoticed for years.

Timeline Thinking is an approach in which time is taken as the carrier rather than an attribute: activity is laid out on a shared timeline — one axis of time crossed by parallel lanes; work is represented as durations with internal phases and markers; the past stays on the same timeline as the future; and the viewing scale moves freely, from a week to a decade.

The document describes the problem (§1), states the thesis and its boundaries (§2), defines the model and the operations on it (§3), works through the causal mechanisms by which the approach has an effect (§4), gives scenarios (§5), a 90-day pilot programme (§6), metrics (§7), risks (§8) and economics (§9). Requirements for a tool are given separately and are vendor-neutral (§11), along with a list of what remains unverified (§12).

Key claims

Eight statements the rest of the document defends

1

The problem is the data model, not discipline. While time is a field on a card, duration, simultaneity and history are inexpressible; no amount of careful upkeep compensates for that.

2

Four blind spots make up one problem: invisible duration, invisible parallelism, a short horizon, and a lost trace. They reinforce one another.

3

The first shift is duration instead of a due date. A deadline is the end of a span; managing deadlines alone is managing the moment of failure.

4

The second shift is parallel lines instead of a queue. Most failures are born where lines cross, not inside any one of them.

5

The third shift is keeping the past in the system. A retained trace turns estimation from an act of intuition into a statistical procedure on your own data.

6

The approach gives visibility, not speed. It does not promise more work in the same time; it promises that you see overload, collisions and the drift between plan and fact earlier.

7

Upkeep costs on the order of 15–30 minutes per person per week, plus a one-off layout when a large piece of work starts. More than that means an antipattern has kicked in.

8

The main adoption risk is the timeline turning into an instrument of supervision. When that happens the data becomes defensive, and the approach loses its point entirely.

Section 1

The problem: temporal blindness

1.1. Definition

Temporal blindness is a state in which someone taking part in an activity cannot get an answer to questions about time without reconstructing it by hand:

The key word is reconstructing. The information formally exists: the dates are in the tracker, the meetings in the calendar, the correspondence in the mail. But to get an answer out of it, a person has to assemble the picture in their head, and working memory is not built for that. In practice this means the answer is either never obtained at all, or obtained as an impression.

1.2. Four forms

Form 1. Invisible duration. Work is represented as a point (a due date) or as a state (“in progress”). A two-hour item and a two-month item take up the same line in a list. Consequence: planning becomes an enumeration of intentions rather than the placement of volumes into finite space.

Form 2. Invisible simultaneity. Tools show one line: a project, a board, a sprint. A person, however, takes part in several lines at once, and their capacity is shared across all of them. Consequence: commitments are taken independently of one another and executed out of a single resource.

Form 3. A short horizon. A calendar is readable at the week, a board at “today”, a sprint at two weeks. No mainstream tool gives a readable quarter or year. Consequence: seasonality, cycles and the accumulation of commitments stay invisible until they arrive.

Form 4. A lost trace. Finished work is archived, hidden, moved to “Done”. Consequence: the only available source of estimates is memory — and memory of duration is systematically biased toward optimism.

1.3. How it shows up

Below are the typical observable symptoms and their actual source. It is precisely the mismatch between symptom and source that makes the problem durable: every individual case gets its own particular explanation.

SymptomUsual explanationActual source
“Everything landed in November”an unlucky coincidenceform 2: commitments were taken one line at a time
“We estimated three times too optimistically”hard piece of work, bad luckform 4: no access to the durations of past work
“The deadline arrived out of nowhere”we lost track of itform 1: the end was being managed, not the span
“Half the quarter went on waiting”the people upstream let us downform 1: the waiting was never drawn as a duration
“I came back to the project and lost a day”context switchingform 4: there is no trace, so there is nothing to restore from
“Status meetings take a day a week”too much communicationforms 2 and 3: the state is visible nowhere, so it has to be said out loud
“Plans are useless, everything changes”the environment is uncertainform 1: dates were planned, not volumes and reserves

1.4. Why the tool landscape does not close this

Each class of tools is optimal for its own question and stays silent about time in its own way: a list answers “what is next”, a calendar “where am I busy”, a board “what is happening now”, a tracker “what state is this unit of work in”, notes “what do I know”. None of them answers “how does this unfold, and what does it land on top of”.

The project Gantt chart — the one mainstream tool in which time really is the carrier — has degenerated inside organisations into a reporting artefact: excessive detail, a rigid network of dependencies, upkeep by an administrator, and a plan treated as a promise. What is left is Gantt theatre: the chart is stale by its second week and lives separately from the work.

Section 2

The thesis: time as the carrier

2.1. Statement

Activity becomes manageable when it is represented not as a set of records with dates, but as the arrangement of durations on a shared axis of time across several parallel lines, with the past retained and the scale freely changeable.

Everything that follows is derived from that sentence. It contains exactly four requirements: durations (not points), a shared axis (not separate projects), parallelism (not a queue), and a retained past (not an archive). The fifth — free scale — is a technical condition of readability.

2.2. Working definition

Timeline Thinking is a practice of organising understanding and control in which:

  1. the unit of description is a span — a piece of work, a period, a state — with a start, an end and internal phases; openness at either end is allowed;
  2. spans are placed on lanes: stable parallel lines that answer the question “whose rhythm is this”;
  3. lanes lie on the shared axis of one timeline, which makes the crossing of lines observable;
  4. finished work stays on the timeline and is used as a source of estimates;
  5. the scale of viewing moves from a week to decades, and each scale has its own class of questions;
  6. the supporting elements are the buffer (an unstructured inbox) and a regular review (a weekly ritual).

2.3. What the thesis does not claim

This section exists to cut off inflated expectations in advance: they are the main reason approaches like this get rejected after the first month.

Section 3

The model

3.1. Primitives

PrimitiveThe question it answersNotes
Timeline “which area of activity is this” the whole plane: the axis of time with every lane on it. You start a separate one when its contents need no comparison with the neighbouring one
Lane “whose rhythm is this” a person, a team, a department, an area of life, a storyline
Span “what lasts” a project, a phase, a state, a period; open ends are allowed
Phase “what kinds of part is it made of” internal structure of a span, without spawning new entities
Marker “how will we know this work is going as intended” a significant point inside a span or a phase; may carry task state
Milestone “what is this line’s significant stage or goal” a significant point on the lane as a whole, belonging to no individual span
Point (event) “what happened at a moment” a meeting, a publication, a payment, an incident; may carry task state
Buffer “where do I put this until I know” an unstructured inbox

A marker, a milestone and a point are three different things, and it is worth telling them apart from the very beginning. A marker lives inside a piece of work and speaks about how that work is going: did we get to where we meant to get. A point is simply something that happened at a moment. A milestone sits on the lane and marks a stage or a goal of the line itself — it belongs to no span, and it stays meaningful even when the work underneath it changes.

The set is deliberately small. Expanding the ontology is the classic way to kill models like this: every new entity demands a decision at the point of entry, and decisions at the entrance kill entry.

3.2. Operations

A model is defined not only by what it contains but by what you do with it. There are five operations, and each is tied to a rhythm.

OperationWhenResult
Capture the moment the thought appears — seconds an entry in the buffer, or a place on the timeline straight away if there is enough to go on
Layout when a large piece of work starts, 20–40 minutes a span with phases, markers, and the “worst week” located
The weekly review weekly, 15 minutes an emptied buffer, boundaries moved to match fact, one decision
The zoom cycle monthly and quarterly answers to the questions of scale: direction, collisions, capacity
The trace review quarterly, and when a large piece of work finishes the plan-versus-fact coefficient, applied to the next estimate

The buffer is insurance, not a mandatory stop. If at the moment a thought appears you already have enough to go on — you know what it is, how long it lasts and which lane it belongs to — placing it straight away is better than routing it through the buffer. The buffer exists for the other case: when there is no time or no material to decide with, and the only alternative is losing the thought.

The operations matter more than the composition of the model. A timeline that never gets reviewed degrades faster than a list, and does more harm than one, because it goes on looking trustworthy.

3.3. Invariants

Testable statements; a violation means the model has drifted.

  1. Everything that lasts is represented as a duration; only the instantaneous is represented as a point.
  2. A lane answers exactly one question — “whose rhythm” — and a “Miscellaneous” lane does not exist.
  3. The boundaries of past spans move to match fact and are never adjusted to fit the original plan.
  4. A span with no foreseeable end carries a revisit date rather than infinity.
  5. Detail never exceeds the horizon: a yearly timeline has no half-hour units on it.
  6. Not knowing is expressed as an open interval, not as a plausible date.

3.4. Properties

Section 4

Mechanisms behind the effect

This section answers “by what mechanism, exactly”, without appeals to belief. Each entry is a causal chain plus a way of testing it on your own data.

4.1. Volume instead of intention

What changes
Work acquires a duration and starts taking up room.
Why it works
A list of intentions has no capacity — there is always space for one more item. A span occupies finite space, so an attempt to add work immediately shows exactly what no longer fits.
Observable consequence
Some commitments are not taken at all — at the discussion stage, rather than at the failure stage.
How to test it
Count, over a quarter, the refusals and deferrals decided before the work started. A rise in that number is the effect itself.

4.2. Early detection of collisions

What changes
Parallel lines lie on one axis, and their crossings are visible.
Why it works
A collision is a property of mutual arrangement, not of any single record; it is fundamentally inexpressible in a tool that shows one line at a time.
Observable consequence
Conflicts are found weeks out, when they can still be resolved by moving something, rather than days out, when only heroics are left.
How to test it
Record the distance between the moment a conflict is discovered and the moment it lands. A growing distance is the effect itself.

4.3. Reference-class estimation

What changes
Past spans of the same type are available and measurable.
Why it works
The systematic underestimation of the duration of one’s own tasks — the planning fallacy, described by Kahneman and Tversky — responds badly to willpower and well to a switch to the outside view: estimating from statistics on similar cases. The technique is known as reference-class forecasting (Bent Flyvbjerg’s work on large projects). A retained trace gives you a reference class made of your own cases.
Observable consequence
The fact-over-plan coefficient stabilises and becomes a usable correction multiplier.
How to test it
Compute the coefficient for every finished piece of work; after 6–10 observations, look at the spread.

4.4. External memory instead of working memory

What changes
The picture is externalised onto the timeline and read with the eyes.
Why it works
Working memory is small and especially poor at holding simultaneous processes; the visual system, by contrast, detects density, gaps and crossings cheaply. Moving the load frees attention for decisions.
Observable consequence
The time it takes to get back into context after a pause shrinks, and the share of discussion spent restoring state falls.
How to test it
Measure the share of status meetings that open with a retelling of “where we are”.

4.5. Emptiness as reserve

What changes
Gaps are read on equal terms with what is filled.
Why it works
In a list model unoccupied time does not exist — it cannot be seen, and therefore cannot be defended. On a timeline the absence of gaps looks like the absence of reserve, which is to say like risk.
Observable consequence
Deliberate buffers appear ahead of milestones, and cascading failures become rarer.
How to test it
The share of major milestones with an unoccupied interval in front of them.

4.6. Drift as data

What changes
Fact is recorded separately from plan, and the plan is not rewritten after the event.
Why it works
An organisation that punishes divergence gets adjusted data and loses its feedback. Separating the planned line from the actual line makes divergence a neutral observation.
Observable consequence
Divergences get discussed rather than hidden, and knowledge of your own systematic error appears.
How to test it
The share of work whose original planned line has been retained.

4.7. Summary

MechanismWhat it makes visibleMetric to check it
Volume instead of intentioncapacityrefusals taken before the start
Early detectioncrossings between linesthe discovery distance
Reference classreal durationthe fact-over-plan coefficient and its spread
External memorycurrent statetime to get back into context
Emptiness as reservemissing buffersshare of milestones with a buffer
Drift as datasystematic errorshare of work with the plan retained

Section 5

Scenarios

Important. The scenarios below are synthetic illustrations built out of typical situations, not reports from adoptions. The numbers in them convey scale; they are not measurements.

5.1. A product team

Situation. Three areas of work, fifteen people, quarterly planning in a tracker, a status meeting once a week. The recurring complaint: “everything converges at the end of the quarter”.

What they do. One lane per area, plus a lane for external dependencies. Initiatives become spans with phases — “work it out → build → ship → observe”. Markers inside the phases are the dates by which something has to be true; the lane’s milestones are the major goals of the area that this work leads to. Everything that depends on other parties — legal approvals, deliveries, partner replies — becomes separate points.

What becomes visible on the very first layout. Three initiatives from different areas all lay claim to the same month of the same person. The “observe” phase after shipping was never planned, yet always took two weeks. Half the dependencies on other parties fall inside the holiday period.

The decision that gets taken. One initiative moves out by a month — not because it matters less, but because it is now visible that three do not fit in parallel.

5.2. An editorial desk and content production

Situation. Several channels, a publication plan in a spreadsheet, and a chronic sense of scramble before each piece goes out.

What they do. One lane per channel, one span per production cycle of a piece (“research → draft → edits → production”), with publication dates as points.

What becomes visible. The publications are evenly spread; the production cycles are not — they overlap at the tails. The real load falls not on publication days but on the intervals between them, and those intervals do not exist in the publication calendar at all.

The decision. What moves is not the release schedule but the starts of the cycles.

5.3. An education programme

Situation. A course of several modules, several cohorts, teachers who overlap.

What they do. One lane per cohort and one lane per teacher. Modules become spans; checkpoints become markers inside them.

What becomes visible. For students, the third week of every module is the point where assignments from three courses converge. For a teacher, the peak of marking coincides with the start of the next cohort.

The decision. Checkpoints move by a week — a change that is impossible without looking at all cohorts at once.

5.4. An independent practitioner

Situation. Freelance or research work: several clients or topics, no external structure, the horizon held in the head.

What they do. One lane per client or topic, plus lanes for personal areas — health, study, family. Commitments become spans.

What becomes visible. The total load taken on exceeds the available time by 30–40% — systematically, not in the odd month. The personal lanes are empty for half a year at a stretch.

The decision. Declining one commitment during negotiation — the cheapest refusal available.

5.5. An incident review

Situation. You need to understand what happened: an outage, a conflict, a long story with several parties.

What they do. Chronology mode — there is no future. Every dated fact goes into the buffer without judgement, then gets laid out on lanes by source: the system, the team, the external party.

What becomes visible. The gaps — intervals where nothing was recorded; the substantial thing is usually in there. And coincidences in time between lanes, which get lost in a textual account.

The decision. The chronology becomes the primary artefact of the review: a sequence shown answers more questions than a sequence narrated.

Section 6

A 90-day adoption programme

Designed for a team of up to fifteen people, or for a single practitioner. It is built deliberately so that each phase is useful on its own and could be the last: an approach that requires full adoption before the first benefit does not survive its third week.

Phase 0. Preparation

Before the start · 1–2 hours

  • Pick one pilot unit: a team, an area of work, or yourself. Not the whole organisation.
  • Name an owner of the pilot — someone who does the work, not someone who watches it.
  • Record the baseline for later comparison: the accuracy of the last five estimates, hours per week spent on status meetings, the number of missed deadlines last quarter.
  • Agree on the rule without which the pilot is pointless: timeline data is not used to appraise people (§8).

Phase 1. Capture and durations

Weeks 1–2 · Goal: stop losing incoming things and learn to see volume

  • Set up a buffer. Everything undigested goes there, with no structure and no dates.
  • Move all current work onto the timeline as spans. Not tasks — pieces of work. No more than 20 spans.
  • Put in honest ends; where the end is unknown, leave an open interval.

Passing criterion: nothing on the timeline that lasts longer than a day is represented as a point.

What already pays off: the visible volume of commitments taken. As a rule, this step reveals 1.5–2 times more work than anyone assumed.

Phase 2. Lanes and rhythm

Weeks 3–6 · Goal: see the crossings and establish a rhythm

  • Separate the lines into lanes: 3–7 of them, one per stable rhythm.
  • Introduce the weekly review — 15 minutes at a fixed time: empty the buffer → move boundaries to match fact → look four weeks ahead → take one decision.
  • Run the first layout for the nearest large piece of work (§3.2), and take it all the way through to finding the “worst week”.

Passing criterion: three reviews happened in a row, and each one ended in a decision.

What already pays off: the first collisions found in advance.

Phase 3. Feedback from the trace

Weeks 7–12 · Goal: turn the past into an instrument of estimation

  • Do not archive what is finished.
  • For every finished piece of work, record both the planned and the actual duration.
  • At the end of the period, run a trace review: take 3–5 pieces of finished work of the same kind, compute the fact-over-plan coefficients, and apply the median to the next estimate.
  • Run a quarterly zoom cycle: year → quarter → month → week, top down.

Passing criterion: the next estimate is made with an explicit correction for your own coefficient, rather than by eye.

Decision points

MomentSign of successSign that it is time to stop
End of week 2volume has become visible, the buffer is in usenobody opened the timeline once after setting it up
End of week 6the review has become a habit, 1+ collision foundthe review has turned into a report for a manager
End of week 12there is a coefficient, and estimates have changedupkeep takes more than an hour a week

Stopping is a legitimate outcome. The approach delivers value in layers, and the second one — durations instead of items, with no lanes and no trace yet — is already self-sufficient.

Section 7

Measuring the effect

7.1. Principle

Metrics exist to test the hypothesis that this is useful, not to appraise people. The moment measurement starts to affect pay or reputation, the data becomes defensive — and with it the only source of feedback disappears. This is not an ethical footnote but an engineering constraint: a measurement system that influences what it measures stops measuring.

The regularity of the review is not an indicator but a precondition: until the ritual is a habit, every number below is unreliable, because the model lags reality. Computing “reviews held ÷ reviews planned” is meaningless — a continuous flow of work has no finite number of planned reviews. A qualitative sign is enough: skips happen less often than once a month.

7.2. Indicators

IndicatorHow to compute itExpected direction
Estimation coefficientmedian of fact ÷ plan across work of the same kindstabilises; the spread narrows
Discovery distancedays between finding a conflict and the conflict landinggrows
Share of work with a durationspans ÷ (spans + points that last more than a day)tends toward 1
Refusals before the startnumber of commitments declined during discussiongrows (and that is a positive result)
Time spent on status discussionhours per week spent retelling statefalls
Share of milestones with a buffermilestones with an unoccupied interval in front of themgrows
Overdue itemsnumber of overdue commitments in the periodfalls slowly, with a quarter’s lag

7.3. What not to measure

7.4. Honest interpretation

For the first 4–6 weeks the indicators typically get worse: overdue work that nobody used to track becomes visible, and volume that nobody used to measure becomes measurable. That is the effect of visibility appearing, not of anything degrading. A meaningful comparison is not possible any earlier than a quarter in.

Section 8

Adoption risks

RiskEarly signCounter-measure
The timeline as supervision a manager demands detail about someone else’s time write down the Phase 0 rule; an aggregated overview instead of a personal one; the timeline is owned by whoever does the work
A reporting artefact the timeline is updated before a meeting rather than during the review it is kept by whoever does the work; the review must end in a decision
Over-modeling weeks go into structure rather than content the invariant “detail never exceeds the horizon”; a cap on the number of lanes
Duplicating systems calendar meetings get copied over by hand draw the line: the calendar holds other people’s claims on your time, the timeline holds your own work
Fading after the start reviews get skipped from the second week a fixed time for the review; the rule “15 minutes, no more”; durations alone as an acceptable stopping point
Punishing divergence the boundaries of past spans get adjusted to fit the plan keep the planned and the actual line separate; discuss divergence as data
Total rollout the approach is introduced organisation-wide by decree a pilot on one unit; spread by example, not by regulation
False precision every date is known and every span is closed make open intervals legitimate, along with the phrasing “not before”

The first two risks stand apart: they do not reduce the effect, they zero it. Both turn the timeline from an instrument of understanding into a subject of reporting, after which what gets written into it is the safe version rather than the real one.

Section 9

The economics of the approach

9.1. Costs

ItemOne-offRecurring
Initial layout of current work1–2 hours per person
Layout of a new large piece of work20–40 minutes at the start
The weekly review15 minutes per person
Quarterly zoom cycle and trace review1–2 hours per quarter
Learning itan hour of reading plus one review done together

The recurring load comes to roughly 15–30 minutes per person per week. That is the upper bound of normal: exceeding it almost always means an antipattern has fired, not that the work is objectively complex.

9.2. What pays off

The cost is compared not with “productivity” but with the price of the specific events the approach prevents:

The break-even point is reached on the first collision prevented. That is exactly why the pilot is built around the layout and the review rather than around completeness of data.

9.3. Where it will not pay off

For single-threaded work with a horizon of a few days the recurring cost is not recovered: there is nothing to prevent, no reference class accumulates, and the horizon is empty. In that case the correct conclusion is not to adopt it (§12).

Section 10

Relation to existing practices

The approach adds a dimension; it does not replace systems. Each of the practices below covers one of its weaknesses.

PracticeWhat we take from itWhat we add
GTDcapture into a single entrance, a regular reviewdurations, parallel lines, a horizon
Kanbanlimiting work in progress, flowan axis of time, memory, a horizon
The calendarcommitments and remindersyour own work and its volume
OKRs and goalsdirection and prioritiescapacity: how much room there actually is
Gantt / PERTthe form: spans and lanes on a shared axis, milestonessurvivability: minimal detail, the plan as a hypothesis, upkeep by whoever does the work
Note-taking systemscontext and knowledgethe processual dimension: how it unfolded

The combining formula: capture from GTD, flow from kanban, commitments from the calendar, direction from goals, and the shape of time from Gantt.

Section 11

Requirements for a tool

This section is vendor-neutral: it is a checklist for evaluating any solution, including a spreadsheet and a sheet of paper.

11.1. Mandatory

  1. Spans with open ends. Without them the ability to state uncertainty honestly disappears.
  2. Parallel lanes on a shared axis. Otherwise collisions — the main source of value — are not observable.
  3. Free scale. Readable at the week and at the year alike; switching takes one gesture, not a rebuild of the view.
  4. Internal structure inside a span. Phases and markers without spawning separate entities.
  5. An unstructured entrance. Capture in seconds, with no required fields.
  6. The past is kept. Finished work stays on the timeline; no forced archiving.

11.2. Desirable

  1. Lane milestones separate from markers inside work. Otherwise the goals of a line and the checkpoints of a particular span merge into one stream of points.
  2. Direct manipulation. Dragging boundaries — with undo, which is mandatory, because without it direct editing becomes dangerous.
  3. Separate planned and actual lines — for the mechanism in §4.6.
  4. Publishing. The ability to show a timeline outward: chronology as a way of explaining.

11.3. Signs of unsuitability

11.4. One implementation

The Timelines product was designed as a direct embodiment of the model described here: a timeline is a Timeline (a board), a lane is a swimlane, a span is a bar (including an open one), a phase is a segment, a marker inside work is a marker, a lane milestone is a milestone, a point or a task is a point, and the buffer is the inbox. This does not mean the approach requires the product: the mapping is given as an example of full coverage of the checklist, not as a condition of practising the approach.

Section 12

Limits and open questions

12.1. Where the approach does not apply, or is redundant

12.2. Known weaknesses

12.3. What remains unverified

An honest list of what we do not know and what would be worth checking before asserting anything stronger:

  1. The size of the effect. The mechanisms in §4 are described causally but have not been measured; it is unknown how much the spread of estimates narrows when you work from the trace.
  2. Durability of the habit. It is unknown what share of practitioners still hold the weekly review after six months, and what helps them do so.
  3. The threshold for the number of lanes. The “3–7” rule comes from practice, not from measurements of readability.
  4. Team dynamics. It is unknown at what team size a shared timeline stops being readable and has to be split.
  5. Transfer between domains. It is unclear whether the skill acquired in a work context transfers to personal planning, or the other way round.
  6. Comparison with a strong alternative. The approach has never been compared head-to-head with disciplined kanban plus cycle metrics.

These questions are a programme for later work, not caveats. A document claiming more than this without data would be advertising.

Section 13

Conclusion

The problem described in §1 is not solved by effort, and it is not a consequence of poor discipline. It follows from a data model in which time is an attribute of a record. In that model duration, simultaneity, horizon and history are inexpressible, and no amount of careful upkeep will change that.

Timeline Thinking proposes one change, carried through consistently: treat time as the carrier. From it follow the primitives (§3), the operations (§3.2), the mechanisms behind the effect (§4) and the practices of adoption (§6). The approach does not promise speed — it gives visibility: overload, collisions and the drift between plan and fact become observable before they turn into events.

Testing this is cheaper than debating it. The minimal experiment takes two hours: lay your current work out as spans on a shared axis, and find the week where everything converges.

The result is usually unpleasant exactly once — and useful every time after that.

Appendix A

Questionnaire: do you have temporal blindness?

Answer yes or no. The questions are about the last three months.

  1. Has it happened that several commitments unexpectedly converged in the same period?
  2. Can you say within a minute how long the previous piece of work of the same type took, without relying on memory?
  3. Do you have any idea what the third month from now will be occupied with?
  4. Can you see at a glance what is running in parallel with your current work?
  5. Do you know your own coefficient of divergence between estimate and fact?
  6. When you come back to something you parked, does it take noticeable time to restore the context?
  7. Do your status discussions open with a retelling of “where we are”?
  8. Are there deliberately unoccupied intervals in your coming quarter?
  9. Is a piece of work that lasts a month represented differently in your tools from a piece of work that takes two hours?
  10. Do you still have access to finished work together with its actual durations?

Interpretation. “Yes” to 1, 6 and 7, and “no” to 2, 3, 4, 5, 8, 9 and 10 are signs of temporal blindness. Five signs or more mean the deficit is structural rather than situational: it will not be removed by effort and calls for a change of model. One or two signs are situational — targeted measures are enough. Zero signs mean the approach is probably not for you, and that is a perfectly normal result.

Appendix B

Glossary

Temporal blindnessthe inability to get an answer about duration, simultaneity, horizon and history without reconstructing it by hand
Timelinethe whole plane of “axis of time × lanes” on which activity is laid out; not one line and not one lane
Lanea parallel line; it answers the question “whose rhythm is this”
Spansomething that lasts; open ends are allowed
Open-ended spana span with an unknown start or end; the form honest uncertainty takes
Phasea part of a span that differs in the character of the work
Markera significant point inside a span or a phase: it shows whether the work is going as intended
Milestonea significant point on the lane as a whole: a stage or goal of the line, belonging to no individual span
Point (event)something that happens at a moment and has no duration
Bufferthe unstructured inbox for thoughts that have not been placed yet
Layoutplacing a piece of work in time at the moment it starts
The weekly reviewa weekly ritual, 15 minutes: the buffer, boundaries moved to match fact, a look ahead, one decision
The zoom cyclereading the same material at several scales, top down
The tracethe retained past, used as a source of estimates
Gapsignificant emptiness; reserve
Collisionseveral lines crossing over the same stretch of time
Estimation coefficientthe ratio of actual to planned duration across work of the same kind
↑ Back to top

Timelines is one way to do this

Boards, lanes, spans with phases, markers and milestones, an inbox and a past that stays put — the primitives of this document, as an app. It’s completely free.

Works on paper — no tool required