If Kelly ran Kmart Innovation
Fourteen rules transposed onto your operating reality — in their unfiltered, Skunk-Works-faithful form. You decide what survives contact with retail.
I have not pre-edited these for political viability.
Kelly's rules were designed to be structurally isolated from the host organisation. Most translations into corporate innovation teams fail because the rules get individually softened until they're indistinguishable from normal operating procedure. The point of this document is to keep them sharp.
Each recommendation states what a faithful translation would look like and flags the friction you'll hit. Three priority bands: NOW (already compatible with your operating model, just do it), NEXT (requires negotiation but achievable in 1–2 cycles), ASPIRATIONAL (requires Kmart structural change you may never get). Adopt, modify, or reject. The triage matrix at the end summarises.
The Fourteen, Translated
14 recommendations / Kelly Johnson's original rules in greyA rotating project lead with no real authority is a project coordinator. Skunk Works leads ran their programs end-to-end. For each of the 4 projects per cycle, the rotating lead should be the single decision-maker on scope changes, technical direction, and ship/no-ship calls. They report to you. Stakeholders (Kylie, Josh, Sam, Emma, Terry, Treena) get input at defined moments, not approval at every fork.
The Skunk Works model is contractor + customer in paired structure. For Kmart Innovation, this means each of the 4 projects per cycle needs one stakeholder who is structurally accountable for the outcome on their side, can make decisions in real time, and is in the working channel daily. If no stakeholder will sign up to that, the project should not be in the cycle. Treat this as a gating criterion at intake.
"Almost vicious" was Kelly's word, not a hedge. If a project genuinely needs more than three, it is not an innovation project; it is a delivery project and should be handed off. The three Innovation Program Managers — covering Business & Ops, Design, and Tech & Data — map cleanly onto a three-person team archetype: one lead + two practice contributors. Pull additional people in for finite consultations only, not as standing members. Visibility into the project is not the same as membership in it.
Replace fixed-stage deliverables (discovery deck, concept deck, recommendation deck) with a single canonical document per project that the team edits continuously and stakeholders read async. The document changes shape as the work changes shape. Heavy decks are inspection theatre — expensive to produce, slow to update, optimised for the meeting in which they're presented rather than for the work. Keep decks for board-level moments only.
Distinguish between two genres that get conflated: status updates (what was done last week — mostly worthless) and decision records (what was decided, by whom, on what evidence — institutional gold). Cut the first to zero. Invest in the second. Pair this with the decision archaeology skill so the rationale survives the project's lifecycle. The Wednesday/Friday sweep cadence is the natural rhythm for this.
Innovation teams hide costs because they don't carry P&L responsibility. That's a comfort, but it's also why innovation teams lose budget the second the business tightens. Voluntarily attribute fully-loaded team time, vendor spend, and infrastructure to each of the 4 projects, and report it monthly to your stakeholder counterparts. Kelly's rule was "don't surprise the customer with sudden overruns" — even when there is no customer paying directly, the discipline protects you.
Wesfarmers / Kmart corporate procurement is built for risk reduction at scale and is fundamentally hostile to short, exploratory vendor engagements. The faithful translation is a pre-approved innovation procurement carve-out: a delegated authority to spend up to $X per engagement on vendors who've been pre-cleared through a lighter process. Without this, the team will continue to be slowed by procurement cycles that exceed the projects themselves.
The three-practice structure (Business & Ops, Design, Tech & Data) already implies domain authority. Make it explicit: work signed off by the practice lead doesn't get re-reviewed by a wider panel before it ships. Reviews exist to catch errors; if the practice lead can't catch errors in their own domain, you have a hiring problem, not a process problem. Pair this rule with Rule 14 — the practice leads have to be rewarded for shipping quality, not for being cautious.
The team that designed the intervention must be the team in the stores measuring it — physically present for at least the baseline measurement and compliance phases of any trial. If you outsource the testing to store ops, you get cleaned-up data, no contextual signal, and you lose the design competency that makes the next iteration better. Kelly's word was "rapidly" — he meant within one program cycle.
This is the single highest-leverage rule that costs almost nothing to adopt. At cycle kickoff, each project brief includes a section listing the corporate standards, IT policies, brand guidelines, governance procedures, or methodological norms the project will knowingly not comply with, with one-line justifications. Stakeholders sign off on the exemptions at intake, not at the end. This converts implicit corner-cutting into a negotiated exception — which is what protects the team when the standard's owner finds out.
If you have to seek incremental approval for spend within an already-approved cycle, you're paying a tax in calendar time on every adjustment. Negotiate a single cycle budget envelope, with deployment authority across the 4 projects at your discretion. Stakeholders see the report; they do not approve every line. The implicit deal: you guarantee they won't be surprised (Rule 6 covers this), in exchange for not requiring micro-approvals.
Monthly steercos exist because daily contact has been outsourced to a meeting. Invert: stakeholder counterparts are in the team's Slack / Teams channel daily, see the work in progress, and intervene in real time. Steerco becomes optional or quarterly. The trust isn't built in the steerco; it's built in the daily exposure. Pair this with Rule 2 — if a stakeholder won't commit to daily channel presence, they probably weren't a real counterpart anyway.
Skunk Works secrecy wasn't paranoia; it was a productivity tool. The team needed long uninterrupted work blocks. Corporate equivalent: each project has a named buffer (likely the rotating lead) who fields all external questions, status pings, and "got a sec?" interruptions. The other two team members are protected. This is consistent with your existing morning-calibration / offence-defence split: enforce it at the team level, not just personally. The async-first philosophy is the structural defence.
Most enterprise HR comp bands reward people for headcount under management. That is the exact opposite of Rule 3. The faithful translation requires explicit advocacy at the HR / comp level: practice leads should be banded by impact, not by reports. Until that changes, the second-best version is your own behaviour — you visibly reward shipping and learning, you visibly do not reward "I'd like another headcount." The team reads this fast.
The Unwritten 15th
For your discretion / not in the original listKelly's 15th rule wasn't about competence. The Navy didn't know what it wanted and couldn't make decisions quickly. The Skunk Works thrived on customers who could (CIA on the U-2, USAF on the SR-71). It failed with customers who couldn't. The faithful translation for Kmart Innovation is harder to say out loud: there are stakeholders whose function in your operating model is "the Navy." They are likely senior, well-intentioned, and important. They commission projects, change their minds, slow decisions, and reduce the team's hit rate.
- Map your stakeholders by decision velocity, not by seniority. Some senior people decide fast; some don't.
- Use the cycle intake to quietly de-prioritise work commissioned by slow-decision stakeholders, even when it looks strategically valuable.
- If you can't refuse the work, compress the surface area — smaller scope, more explicit "Will Not Do" (Rule 10), tighter time-box.
What Would Actually Have to Change
Hard truths / for your eyesSkunk Works was structurally sealed off. You're not.
Kelly reported to a division president and was kept outside Lockheed's normal approval flows by design. Your team sits inside a retail enterprise that has legitimate reasons to want consultation rights, governance compliance, and information flow. Most of these rules don't transplant unless you and your direct executive co-sponsor explicitly negotiate the isolation.
The honest version: You can probably get 60–70% of the operating discipline without the isolation. The remaining 30–40% requires structural authority you may not have.
The rules are a system, not a menu.
Picking three favourites and ignoring the rest will not work. Rule 3 (small team) only survives because Rule 14 (no headcount incentive) removes the empire-building drive. Rule 5 (minimal reports) only survives because Rule 12 (daily liaison) gives stakeholders information through other channels. Adopt the related rules together or expect the isolated rule to erode within a cycle.
The pairs: 3+14, 5+12, 1+2, 8+9, 6+11.
Speed is the validation, not the goal.
The XP-80's 143-day delivery wasn't the point of Skunk Works; it was the consequence of the operating model. If you adopt these rules and projects don't measurably accelerate, the rules aren't working — or you're applying them in form but not substance. The honest metric is cycle time from intake to validated learning, not adherence to the rules themselves.
Baseline now, measure in two cycles.
The hardest rule for you specifically is 13.
Based on how your operating philosophy is already wired — async-first, deep-work-protected, morning-calibration — you've already adopted Rule 13 personally. Translating it to the team means letting the Innovation Program Managers also be structurally protected from interruption, including from you. The buffer needs to absorb your own pings as well as stakeholder pings.
This is the smallest behavioural change with the largest team-level impact.