Whitepaper: declared stages, checked by people

The hill knows where the work is

A lightweight way to see dozens of workstreams at once — and to notice when the written record and the people doing the work quietly disagree.

Large companies do not lack status. They drown in it: red-amber-green cells, percent-complete bars, weekly decks. What they lack is status that says whether the hard part is behind us. This paper proposes a hill chart whose dots are placed automatically from a declared stage and the work recorded against it — and which treats every human correction as a clue rather than an overwrite.

1The problem

Picture a programme with five workstreams and forty deliverables, owned by people in four time zones. Each one has a ticket, a page, or a spreadsheet row. Each one reports progress. Almost none of those reports answer the only question a leader actually has: is this going to land, and if not, where is it stuck?

Percent-complete is the usual answer, and it fails for a structural reason. It assumes work is a straight line from zero to done. Real work has two phases. First there is uncertainty: nobody yet knows the shape of the solution, what systems it touches, or who has to agree. Then there is execution: the approach is known and it is a matter of doing it. A task can be "sixty percent complete" and still have its biggest unknown sitting in the remaining forty.

At scale this becomes invisible work. People carry unknowns in their heads. Blockers live in chat threads. The status page says "on track" until the week it says "red", and by then the decision that could have helped was needed a month ago.

2Why a hill

Basecamp's hill chart names the two phases directly. The left slope is figuring it out; the crest is the moment nothing unknown remains; the right slope is making it happen. Each dot is a piece of work, and its position is a person's honest sense of where that work stands.

That honesty is the strength of the original. It is also what makes it hard to run across a large organisation. Someone has to remember to drag forty dots every week. Positions drift toward optimism. Nothing connects the dot to the evidence behind it, so a dot on the downhill slope and a dot on the uphill slope look equally credible. And there is no way to see a whole workstream as one dot made of many.

3Options we considered

There are three honest ways to decide where a dot goes. We compared them on the four things that matter at scale.

ApproachHonestyEffort to maintainRolls up to streamsResistant to gaming
A. People drag every dotHigh while people careHigh; decays within weeksNoLow — optimism wins
B. Computed from ticketsLow: tickets measure activity, not certaintyNoneYesLow — close tickets, move dots
C. Declared stage, placed by the record, corrected by peopleHigh, and the corrections are visibleLow: update the stage and tick items offYesMedium — disagreement surfaces

Option C keeps the human judgement Basecamp was right about, but takes the chore away. It also creates something neither of the others has: a measurable gap between what is written down and what people believe. That gap is the most useful signal in the whole system.

4The proposal

What we record

Each deliverable has a name, the outcome it should produce, the workstream it belongs to, its owners, the system components it touches, a stage, and a short list of items. Items come in five kinds, and each answers a different question about the work:

The record is expected to be incomplete. With dozens of owners updating asynchronously, much of what happens is never written down, and the chart is designed around that rather than against it.

Where the dot goes

The stage chooses a band on the hill. Inside the band, the record moves the dot. On the uphill stages it moves as questions are answered and decisions are made; on the downhill stages it moves as scope parts are built. Next steps never move it: they describe activity, not certainty — the same reason option B was rejected. An open blocker holds the dot at the start of its band and draws a short red bar across the slope in front of it, like a chock under a wheel. The blocker changes only where the record places the dot; it stops no one from doing anything.

The six working stages mapped onto the hill. Answered questions and made decisions move a dot through the uphill bands; built scope parts move it through the downhill bands. Completed work leaves the hill on the right.

Knowns and unknowns

Open questions and waiting decisions are known unknowns. The system reads them in two directions. A deliverable still clarifying its problem with none written down is suspiciously certain: either the problem is genuinely simple, or its unknowns have not been found yet. A deliverable on the downhill slope with questions or decisions still open is flagged: the task list may be incomplete. This is a signal, not a rule — the dot is not held back, because a partial record is normal.

Not every signal deserves the same attention. The chart sorts them into three tiers. Red flags are contradictions that turn into rework or late surprises if left alone: every scope part built while a question is still open, a blocker older than two weeks, a decision waiting three weeks, or someone moving work back over the crest. Ask about covers what needs a conversation soon. Keep an eye on is context, folded away until asked for.

Below the signals, the same record is read two more ways. Four kinds of knowing counts what has been learned in the last two weeks, what is still open, what was written down only after the crest, and the notes people leave when they move a dot by hand — the unknown knowns. It describes rather than ranks: nothing in it is flagged. Decisions waiting lists every open decision across all deliverables, oldest first, with who decides. It is the steering view of decisions without a separate list for anyone to maintain; recording one opens its deliverable, so the moment of resolving still asks what came out of it.

Two more readings follow from scope parts. A deliverable in design with no parts listed doesn't yet say what it will build. A part added after the crest means the scope grew during execution — the useful half of traceability, without anyone recording dependencies.

The moment of resolving

Answering a question or making a decision often creates work. When someone ticks one off, a single line appears beneath it: anything new come out of this?, with buttons to add a question, decision, scope part or next step. Nothing links the new item back to the answer. The prompt builds the habit; a dependency graph nobody maintains would not.

How a workstream becomes one dot

A stream's position is the average of its deliverables, with one rule on top: a stream cannot pass the crest while any part of it is still uphill. Five finished deliverables and one unresolved design do not make a stream that is "figured out". So while anything is uphill, the stream's dot measures only how much of it has been figured out — downhill parts count as reaching the crest, not beyond it. Hover a stream and a coloured band on the hill shows the spread from its least to its most advanced deliverable, so a stream with a long tail is visible at a glance.

When a person disagrees with the record

Anyone can drag a dot. The dot moves and gains a dashed ring; hover or tap it and the place the record put it appears as a hollow ghost, joined by a dotted line. Ghosts stay hidden otherwise, so a chart with many corrections stays readable. If the gap is large, the chart opens a short conversation instead of silently accepting the new position.

Each prompt ends in an action — add a question, decision, blocker, scope part or next step; split the deliverable; move the stage — so the correction flows back into the record and the ghost and the dot meet again.

Sometimes the honest answer is not a structured item: just a feeling, the team seems stuck. Keeping the placement accepts a one-line note instead, shown in the tooltip, the signals and the export. It respects people who won't structure their knowledge, and still captures it.

Time

The chart takes one automatic snapshot a day and lets you name snapshots for occasions like a steering review. Comparing against an earlier snapshot draws where each dot was, with a thin trail. A trail pointing left means work rolled back uphill. A dot with no trail for two weeks is listed as still.

Stillness has a sibling. A record nobody has touched for ten days is quiet. A dot can be still because the work is stuck, or because nobody is writing anything down; those need different conversations, and telling them apart costs nothing.

Getting it into Confluence and email

Both tools strip most styling from pasted HTML and discard live graphics. So the export produces a high-resolution image of the chart and a plain, inline-styled table grouped by stream. Paste them separately into Confluence; paste them together into an email.

5What this does not do yet

It stores data only in this browser. It does not import from Jira or spreadsheets, and it does not share state between people; those are deliberate next steps once the placement rules have been tried on real work.

Prototype

A sample programme is loaded. Change a stage in the table, answer a question or mark a scope part built, or drag a dot along the hill and see what the chart asks you. Everything is saved in this browser.

Click or tap a workstream to see its deliverables. Drag any dot to say where you think it really is. On a touch screen, the first tap shows details and the second opens.

Worth a conversation

Where the record, the blockers or the calendar suggest something is off.

Four kinds of knowing

Across everything on the chart right now.

Decisions waiting

Oldest first, from every deliverable on the chart.

    Deliverables

    Edit in place. Open a row to manage its questions, decisions, scope parts, next steps and blockers.