What Is Dual Track Agile? Two Streams, One Team
What is dual track agile? Discovery and delivery run at the same time as two continuous streams of work, done by one team, instead of phases that take turns.
Dual track agile is a way of working in which discovery and delivery are two continuous streams of work, performed by the same team at the same time, rather than two phases that take turns. At any moment the team is both learning what is worth building and building what has already been learned.
The idea is widely known and rarely run properly. This post explains what the two tracks are, why they exist, and where the practice honestly stops working.
What is dual track agile, and what problem does it solve?
Most product teams swing between two failure states. In the first, the team ships continuously. The roadmap is full, releases go out on time, velocity looks respectable, and after two quarters remarkably little about the business has changed. In the second, the team decides to be more rigorous, pauses to do research, produces a pile of insight and ships almost nothing while that insight goes stale.
Both come from the same belief: that discovery and delivery are phases. Once you accept that framing, one of them always loses, and in practice it is discovery. Delivery has deadlines, outside commitments and an obvious cost to stopping. Discovery has none of those, so it can always be pushed to next quarter. "We will do proper research once this release is out" is said in good faith, every quarter, indefinitely.
Dual-track is the correction. The two tracks run at different rhythms, produce different artefacts and answer different questions, but they never stop for one another.
The two tracks side by side
| Discovery track | Delivery track | |
|---|---|---|
| The question it answers | Should this exist at all? | Did we build it well? |
| Its rhythm | Irregular: some questions resolve in an afternoon, others take a month | Defined units of predictable duration |
| What it produces | Evidence, decisions and, often, a decision not to build | Shipped work that has already earned its evidence |
| Who does it | The one team | The same one team |
The last row matters as much as the others. The tracks differ in clock and in question, not in people.
Why delivery alone is not enough
A team that optimises purely for output becomes what the guide calls a feature factory. It produces a great deal of confidently shipped work with no reliable connection to outcomes. From the inside it looks healthy: the board is full, everyone is busy, the pace is steady. It is also the most expensive way to run a product organisation, because the real cost is not the work you did but the work you could have done instead.
The discovery track attaches evidence to the queue. What gets built is then what is worth building, not what was most recently requested.
Why discovery alone is not enough
The opposite failure is rarer and just as real. A team that learns continuously but ships too slowly cannot test any of its conclusions against reality. Insight that is not implemented does not compound. It decays, because markets move and confidence in a finding drops with every month it goes untested.
Delivery is therefore not only the reward for good discovery. It is the mechanism that validates it. A conclusion nobody shipped is a hypothesis with better manners.
Six jobs the two tracks do together
The practice earns its place through six specific mechanisms. None of them is simply "talk to users". Each prevents a particular failure.
| Mechanism | Failure it prevents |
|---|---|
| Delivery needs discovery | A feature factory with no link to outcomes |
| Discovery needs delivery | A research project whose insight decays |
| Different rhythms, different evidence | Discovery items looking permanently overdue on a delivery board |
| De-risking before build capacity is spent | Answering cheap questions in the most expensive way |
| A legitimate way to say "we don't know yet" | Building something the team doubts, or objecting without evidence |
| Closing the loop after shipping | Never learning which releases actually worked |
Three of these deserve a closer look.
Different clocks
Forcing both tracks into the same cadence, the same board and the same definition of done is one of the main reasons teams give the practice up. Discovery items sit on a delivery board and appear to be failing when they are behaving normally. Recognising that the tracks keep different time is most of what makes running them together possible.
The economics
Building is by an order of magnitude the most expensive way to test an assumption. Discovery exists to answer, cheaply, the questions that would otherwise be answered expensively: will anyone use this, do they understand it, is this the real problem, will they pay. Every idea killed in discovery after a few days of effort leaves a quarter of build capacity free for something with better evidence behind it. This is the argument that persuades sceptical stakeholders, because it is about cost rather than philosophy.
Permission to not know
In a delivery-only culture, uncertainty has nowhere to go. A team asked to build something it doubts can build it anyway or object without evidence, and building is easier and more common. A discovery track offers a third answer: that is a real question, here is the cheap test, we will know in a week. The conversation changes from opinion against authority to a question with a date attached.
The honest limit
Dual-track is not a cure for everything, and the guide is direct about where it stops working.
- It is not two teams. Splitting into a discovery group and a delivery group recreates the handoff problem the practice was meant to solve. Insight arrives as a specification, the builders were not part of the reasoning, and the context is lost in translation.
- It is not a licence for indefinite research. A discovery track with no forcing function toward decisions turns into an internal research function that produces documents nobody implements.
- Capacity is fixed. The same people are involved in both streams and the total is finite. Anything given to one track comes from the other.
The practice only works when both tracks are staffed by the same people and both are held to producing something that changes what happens next.
Three misreadings to avoid
- Reading it as two teams. The people who will build the thing need to be present in the discovery that shapes it. Not every interview, but enough that the reasoning is shared and not merely passed on.
- Reading it as a stage gate with a new name. If the boundary between the tracks collects templates, approvals and queues, you have rebuilt waterfall with new vocabulary. The hand-over is covered in how to hand off discovery to delivery.
- Letting one track starve the other and calling it pragmatism. When delivery pressure absorbs everything, the queue fills with items nobody can tie to an outcome. When discovery becomes precious about certainty, shipping slows while research documents pile up.
All three misreadings are about protection and follow-through, not about method. The method itself is easy to describe in ten minutes.
What running it looks like in outline
The full detail lives in sibling posts, so here is only the shape.
| Stage | What happens | Where to read more |
|---|---|---|
| Before you start | Settle one team, the bar for evidence, the capacity share and who decides | Splitting capacity |
| While running discovery | Test the riskiest assumption with the cheapest test that could change your mind | Running product discovery |
| At the boundary | Move work across with its reasoning attached | The hand-off |
| After shipping | Check whether the outcome moved | Closing the loop |
Notice that nothing in the table requires special software. Dual-track can be run with two boards, a shared document and a disciplined team. The hard part is holding the two tracks together while a quarter is under pressure.
A quick test of whether you are really doing it
Ask three direct questions about the last month.
- What did discovery decide?
- What did delivery ship?
- Were the same people involved in both?
Two consecutive months with a blank on either side is the signal that one track is starving. If the answer to the third question is no, you have two teams, whatever the organisation chart says.
Frequently asked questions
Is dual track agile the same as doing research before you build?
No. Research before building is a phase, and phases take turns. Here the learning continues while building continues, and it carries on after release, when the team checks whether what it shipped produced the outcome it predicted.
Do we need a separate discovery team?
No, and creating one is the most common way the practice fails. The engineers and designers who will build the work need to be present in the discovery that shapes it, so the reasoning is shared rather than transmitted as a document.
Why do discovery items look overdue on our board?
Because they are being measured on delivery's clock. Discovery is irregular by nature, so it needs its own rhythm and its own definition of done. Items that look late on a delivery board are often behaving normally.
What is the point of discovery if we have a full roadmap?
A full roadmap tells you the team is busy, not that the work is worth doing. The discovery track exists to put evidence behind the queue, so that the team builds what is worth building and not simply what was asked for most recently.
The full method, with the checklists, is in The Dual-Track Playbook, a free 34-page playbook.