PHX

How I decide what gets made.

Six stages. The same six whether what comes out is a video or a system.

Most of what stops me building the wrong thing happens in the first two stages, before anything is made. A problem with nothing specific against it does not become a task. It stays a question.

The two that stay open

Neither of these is a phase that finishes. They are the reading the rest of the work is done against, and they get revisited for as long as an engagement runs.

Needs and market

Work out what the business actually needs before deciding what to make.

Hands on

A prioritised set of audiences and the problem each one has.

Inside it

  • What the business is trying to move, in its own words
  • Who the product is genuinely for and who it is not for
  • The distinct audiences worth separating rather than averaging
  • What each of them is already doing instead

Competitor and reference research

Find out what the category already looks like.

Hands on

A read on the category and a shortlist of what to avoid looking like.

Inside it

  • What the direct competitors are running and what they are claiming
  • What the category has trained the audience to expect
  • Reference collected from outside the category. That is where the good ideas usually are
  • The conventions worth keeping and the ones worth breaking on purpose

The four that turn

These are the pass. Make the idea, plan it so a team can build it, build the versions that let it be tested, then read what came back and start again from what won.

The innovation

Make the actual idea.

Hands on

Named concepts, each with an audience, an angle and a reason to exist.

Inside it

  • A new way to show the product that has not run before
  • The angle for each audience, mapped to the problem it answers
  • Ideas argued with while they are still cheap to kill
  • The concept named so it can be tracked and beaten later

Storyboard and production

Turn the concept into something a team can actually make.

Hands on

A shot plan a team can execute without asking what was meant.

Inside it

  • The stage plan: what happens, in what order and the job of each beat
  • The messaging and copy strategy for the piece
  • The hook and the scroll stopper that has to survive the first second
  • Caption strategy, including the spoken style a real person would use
  • What has to be in frame at every stage, and why

Variations

Build the versions that let one idea be tested properly instead of guessed at.

Hands on

A set where every difference is deliberate and therefore legible.

Inside it

  • The same idea composed for each placement, never resized into it
  • Length cuts that are different edits rather than trims
  • The one variable changed at a time so a result can be read
  • Enough versions that the test has something to say

Champions

Read what came back and name what won.

Hands on

A champion and a sharper brief for the next pass.

Inside it

  • What actually performed and the version it beat
  • Why it won, separated from why it was expected to
  • The winner promoted into the next round as the thing to beat
  • What the result changes about the concept and occasionally about the audience read

The same six on a system

None of this was invented for software. It is what I ran on paid social, and it is what happens before I specify a build. The stages did not change. What comes out of them did.

Needs and market

The same conversation before a system is specified: what the business needs, who uses it and what they do today instead.

Competitor and reference research

The same sweep before I design an interface: what the incumbents do, what users already expect and where the convention is worth refusing.

The innovation

The feature itself. What it does that nothing else does, named and specified before anyone builds it.

Storyboard and production

The build spec: states, edge cases, copy and the order a person meets them in, written before the first component.

Variations

The same discipline in a build. One change per release where it matters so I can tell what caused what.

Champions

Ship, read, iterate. The release informs the next spec and the loop is the same loop.

Where it has run

The stages do not move when a new engagement files under them, and that is the only test of a method worth anything. Two are written out in full.

New customer acquisitionConcept to finished asset