Why a generic kanban board breaks the moment money is involved

There is nothing wrong with kanban. It does exactly what it was designed to do: make flow visible and limit how much is in play at once. The trouble starts when an agency asks it to also be the system of record for commitments, hours and money, which it was never designed to be.

The failure is not dramatic. A generic board works beautifully for about two months. Then it acquires a column called “Waiting”, a column called “Blocked (client)”, a swimlane per client, a label taxonomy nobody can explain, and a weekly ritual where someone reconstructs reality from it by hand.

What a card cannot tell you

Take any card on any board. It can usually tell you a title, an assignee, a due date, and a column. Now ask the four questions that actually matter to a services business:

• What did we quote for this, and did anyone approve that number?
• How many hours have gone into it so far?
• Is the person doing it owed money for it, and how much is left?
• If the client disputes it in three months, what is the record?

A card answers none of these. So the answers get stored elsewhere: a proposal doc, a timer app, a spreadsheet, an email thread. Four systems for one piece of work, and a human whose job is to keep them in agreement.

The three specific breakages

Columns collapse two different questions

A column can express stage or custody, not both, so teams pick one and encode the other in labels or task titles. We wrote about this at length in “Waiting on client” is not “in progress”. The short version: the convention always decays.

Time lives somewhere else

Most boards bolt on time tracking as an integration, which means hours are associated with a card by a sync rather than by structure. Sync-based relationships fail quietly. A card gets renamed, a project gets restructured, and hours end up attached to nothing in particular. You find out at month end.

There is no commitment object

This is the big one. A board tracks tasks, but client work runs on commitments: a quoted amount, for named work, agreed with a specific person, at a rate. That object does not exist on a board, so it lives in a PDF. Everything downstream — burn, remaining balance, contractor payouts — depends on an object the tool does not model.

You can bolt time and money onto a board. You cannot bolt a commitment onto it, because a commitment is not a card. It is an agreement that cards get measured against.

Keep the board. Anchor it to something.

The answer is not to abandon boards. Boards are the right way to look at what is moving this week. The answer is to make the board a view onto a structure that already knows about estimates, hours and payouts, rather than the structure itself.

In practice that means a piece of work carries three things at once:

1. An estimate with hours or a fixed price, a rate, dates, and an approval state. Draft, submitted, approved, rejected with a reason, re-edit with a note.
2. Time entries logged against that estimate or the task under it — not synced, attached. Edits go through a time-change request a manager reviews, so the record stays auditable.
3. A payout position. When work is marked ready for payment, the payment attaches to the estimate it came from, and amount paid versus remaining rolls up per developer automatically.

Once those exist, the board stops being asked to carry information it cannot hold, and goes back to doing the one thing it is genuinely great at.

When a plain board is fine

Be fair to the simple tool. A plain board is entirely sufficient when:

• You bill a flat retainer and do not need per-project profitability.
• Everyone doing the work is salaried, so payouts are not a variable.
• The team is small enough that “who is owed what” fits in one person’s head.

If two of those three stop being true, the board will keep looking fine while the spreadsheet behind it grows. The spreadsheet is the tell.

Common questions

Can we migrate mid-project?

Yes, and the cleanest cut is at the next estimate. Let running work finish where it is; put the next approved estimate and everything under it in the new structure. Trying to backfill historical hours is rarely worth it.

What about clients who insist on their tool?

Let them have their board for visibility. Keep the commitment, the hours and the payouts on your side. Your operations record should not be hostage to a client’s tooling choices.

Is this just “kanban plus fields”?

Not quite. Fields describe a card. What is missing is a separate object — the approved estimate — that many cards and many time entries roll up to. That is a structural difference, not a cosmetic one.

Boards show you movement. They do not show you whether the movement was worth it.