6. Feasibility as sequence

§The hypothesis: the first version can be sequenced so that every act depends only on acts that have already held, and on operations that have already shown they can deliver.

§The previous two chapters produced a list of what the product needs from people, partners and operations, and whether each has been shown or only said. This chapter orders the product by that list.

§ 6.1Three columns#

§Depends on no one. Elements the team can build and the customer can use without any partner delivering, any operation changing, any integration landing. These go first, not because they are most valuable, but because they can be true on a date the team controls.

§Depends on something shown. Elements that need an operation or partner to do something it has already demonstrated. These go next, and their date is the partner's date, not the team's.

§Depends on something that failed or was never shown. Elements that rest on a check-in that failed six of six, on a server's glance nobody has watched, on a data-sharing agreement between brands that has never existed. These wait, with the criterion that would promote them written down, and are not discarded.

§Goldratt's insight from manufacturing transfers cleanly: the sequence is set by the constraint, and the constraint is whatever the team does not control. A roadmap that ignores this puts the team's best idea first and discovers in month two that the idea was waiting on a partner's month six.

§ 6.2The roadmap from the other side#

§The fourth whitepaper built the roadmap top-down: mission to bets to goals to items with promotion criteria. This chapter builds it bottom-up: from what can actually be delivered, in what order, by whom. When the two agree, the plan is sound. When they disagree, the disagreement is the finding: a bet the strategy wants first that feasibility says cannot go first is either a bet that needs a smaller first form, or a bet in the wrong position.

§ 6.3Doors, again#

§Bezos's one-way and two-way doors apply to sequence. A dependency on a two-way door, a partner that can be swapped, an integration that can be replaced, can be taken early and unwound. A dependency on a one-way door, a data model, a contractual exclusivity, a brand promise made to customers, should come only after the evidence has earned it. The technical path recommendation from chapter 4 listed the one-way doors. The sequence puts nothing behind them until the bet they serve has held.

§ 6.4Appetite per act#

§Each act in the sequence gets an appetite, in weeks, set before the work and held. Ryan Singer's reason is the right one: an act without an appetite absorbs every good idea and ships late; an act with one is shaped to fit. Act one in the case was four weeks. Anything that did not fit in four weeks, including several good ideas, went to the second or third column with a note.

§ 6.5What has to be true#

§Added to the running list: each act, its column, its dependencies, and for every dependency whether it is shown, said, or failed. The promotion criterion for everything in the third column. The appetite for each act. And the one-way doors behind which nothing is scheduled until its bet has held.