How to Split Capacity Between Discovery and Delivery
How to split capacity between discovery and delivery: name a share of time, protect it in writing and expect to defend it, because spare time never appears.
To split capacity between discovery and delivery, name the share of the team's time that goes to discovery, plan around it, and write down what happens to it when delivery is under pressure. The exact number matters less than the fact that it exists and is protected.
Most teams that abandon dual-track do so because of a decision that was never made at the start. This is one of the most important of those decisions, and the one that is most often left implicit.
How to split capacity between discovery and delivery: why a number is needed
The usual arrangement is unspoken. The team does discovery "when there is time". It sounds reasonable, and it quietly guarantees that no discovery happens, because the time never appears. There is always something urgent, and urgency is a permanent condition rather than a temporary one.
The reason is structural. Delivery has deadlines, outside commitments and an obvious cost to stopping. Discovery has none of those, so it can always be deferred to next quarter. That deferral is what "we will do proper research once this release is out" means, said in good faith, every quarter, forever.
An unprotected share is therefore the same as an unallocated one. It goes to whatever is loudest. Naming the share is the only way to make discovery compete on equal terms with work that already has a date attached.
Decide the share, not the perfect share
The guide does not prescribe a proportion, and neither should you agonise over one. Different teams and different quarters will land in different places. What the guide insists on is a short list of properties.
| Property | What it means in practice |
|---|---|
| Named | Someone can say out loud how much of the team's time goes to discovery |
| Planned around | Delivery commitments are made with that share already taken out |
| Protected | It does not quietly shrink each time something urgent arrives |
| Explicit | It is written down, not understood |
If you can tick all four, the number you chose is good enough. You can adjust it later, in daylight, with evidence. What you cannot recover is a share that was never there.
Expect to defend it
Accept from the start that this allocation will be the first thing sacrificed when delivery is under pressure. That is not a sign of bad people. It is how the incentives are built. Delivery pressure is visible and discovery's value is not, so in any given week the argument for taking the time back sounds sensible.
This is precisely why the share needs to be explicit and not implicit. An implicit share can be eroded without anyone ever deciding to erode it. An explicit one forces a visible decision, and a visible decision can be questioned.
Agree in advance what happens when pressure arrives. The guide's checklist puts it plainly: we know what happens to that allocation when delivery is under pressure. Possible answers include who is allowed to raid it, for how long, and what must be said out loud when they do. The point is that the answer exists before the pressure does.
The argument that persuades sceptics
Stakeholders will ask why capacity should go to something that does not ship. The best answer is about cost, not philosophy.
- Building is by an order of magnitude the most expensive way to test an assumption.
- Discovery answers cheaply the questions that would otherwise be answered expensively.
- Every idea killed in discovery after a few days of effort frees a quarter of build capacity for something with better evidence behind it.
Framed this way, the discovery share is not a tax on delivery. It is the part of the budget that decides whether the rest of the budget is well spent. For a fuller account of why the two tracks need each other, see what is dual track agile.
What the split is not
A capacity split can be misread in two damaging ways.
| Misreading | Why it fails |
|---|---|
| Some people do discovery, others deliver | This recreates the handoff problem. Insight arrives as a specification and the builders were not part of the reasoning |
| Discovery gets whatever is left over | Nothing is left over, so nothing is done |
The split is a division of the same team's time. The same people are involved in both tracks, and the total capacity is fixed. That is also the honest limit of the practice: anything given to one track comes from the other, so a share that is too large will starve delivery just as surely as a share that is too small starves discovery.
Keep both tracks alive over time
Setting the share is a start. Keeping the balance takes a monthly look. The guide describes two directions of failure, and both feel locally reasonable at every step.
- Discovery starves when delivery pressure absorbs everything. The symptom is a delivery queue full of items with no evidence behind them and nobody able to say what outcome they serve.
- Delivery starves when discovery becomes precious about certainty. The symptom is shipping slowing down while research documents accumulate.
The check is a direct question each month: what did discovery decide, and what did delivery ship? Two consecutive months with a blank on either side is the signal to rebalance.
What to do when the buffer runs dry
A subtler version of the capacity question appears at the boundary between the tracks. Delivery needs a steady supply of validated work, but discovery cannot produce it on a fixed schedule, because questions do not resolve on demand.
The guide's answer is a buffer: enough validated work queued that delivery is never blocked, but not so much that discovery is being rushed into approving things to fill the queue. When the buffer is empty, the honest response is to let delivery pick up technical work or reduction of debt, rather than to lower the discovery bar. Lowering the bar turns dual-track back into a feature factory with extra ceremony. More on that handover is in how to hand off discovery to delivery.
Give the share a direction
Protected time spent on nothing in particular is still wasted. Attach the discovery share to the quarter's objectives, so that it has somewhere to go. Objectives say what should be different, discovery finds the routes that might achieve it, and delivery builds the routes that survive.
An outcome-based objective is what makes the share defensible. It commits to a result rather than to a predetermined feature, which means finding a cheaper route to the same outcome counts as a success and not a deviation from the plan.
Also give each question a limit. Time inside the discovery share still expands to fill what is available, so a timebox for each question keeps the share from turning into open-ended research. Tips for running the loop within it are in how to run product discovery.
Questions to settle in the capacity conversation
A single meeting can settle most of this. Bring the following questions, and write the answers down before anyone leaves.
| Question | Why it matters |
|---|---|
| What share of the team's time goes to discovery? | An unnamed share is an unallocated one |
| Who is allowed to reduce it, and what must they say when they do? | It turns erosion into a visible decision |
| What does delivery do when the buffer of validated work is empty? | It prevents the discovery bar from being lowered to fill the queue |
| Who decides what graduates? | Without an answer, either everything or nothing moves across |
| Which objectives does the discovery share serve? | It gives the time a direction |
Why leftover time fails, in three steps
It helps to walk through the logic once, because it is the argument you will be repeating.
- Delivery has visible deadlines and an obvious cost to stopping. Discovery has neither.
- So when the two compete for the same hours, delivery wins every time, in good faith.
- Therefore any discovery that depends on spare time will be deferred each quarter, and the team ends up in the feature factory it was trying to avoid.
The only way to break the chain is to take the decision out of the weekly argument. The share is settled once, deliberately, and then defended. That is also why the guide asks you to say it out loud more than once.
A checklist for the capacity decision
- A share of capacity is explicitly allocated to discovery.
- The share is written down and protected.
- We know what happens to it when delivery is under pressure.
- Discovery is attached to the quarter's objectives.
- Each question has a timebox.
- Neither track has been blank for two consecutive months.
- We said all of this out loud, more than once.
Frequently asked questions
How much of the team's time should go to discovery?
The guide gives no fixed proportion. It says the number matters less than the fact that it exists and is protected. Choose a share, write it down, and adjust it later with evidence.
What if we are in a crunch and cannot spare anyone?
Expect that moment to come, and decide in advance what happens to the allocation when it does. If it is raided, make that a visible decision. An allocation that disappears silently was never really there.
Should we have a separate discovery team instead?
No. A separate team recreates the handoff problem, with insight arriving as a specification and the builders absent from the reasoning. The same people should hold both tracks.
How do we know the split is wrong?
Ask each month what discovery decided and what delivery shipped. A blank on either side for two consecutive months means one track is starving.
The full method, with the checklists, is in The Dual-Track Playbook, a free 34-page playbook.