Small steps, big vision: why industrial leaders never aim for the best product on the first try

Starting this new long-format podcast, "DASSAULT: France's Defense Flagship" (French-language)¹, I wasn't expecting to walk away with something positive to build a blog post on — a defense company, led by men whose politics, among other things, I don't share. And yet, part of that story caught my attention and resonated with what I try to do every day: what they call "the policy of small steps."

The principle is simple enough: prototypes are built on known elements into which innovations are progressively integrated² — you never change the airframe and the engine at the same time³.

Here's the more interesting part: this principle was independently validated by a study funded by the U.S. Air Force, which was trying to understand why its own industry was being outpaced by a much smaller company in building fighter jets. That's the R-1148-PR report, "A Dassault Dossier: Aircraft Acquisition in France," Robert L. Perry, RAND Corporation, 1973⁴, which confirms the method of "gradual, incremental, and evolutionary design."

I had Claude put together a summary and fact-check of what the podcast reports on this policy of small steps, and you can find the summary sheet at the bottom of the page.

Today's thesis: industrial leaders built to last haven't necessarily aimed for the best product on the first try — they move forward in small steps that validate, at every stage, that the trajectory still holds. This method has a documented limit (the blindness of whoever optimizes without a vision), and its condition for success is a structured dialogue between whoever holds the course and whoever executes. Software agility didn't invent anything: it's the modern version of that same dialogue — execution that validates, sprint after sprint, a product vision that stays elsewhere.

Movement 1 - The principle, beyond Dassault

What struck me in this story was the incremental nature of innovation at a cutting-edge industrial company; the ambition of continuous improvement on a solid base; shipping a first version with real strengths — enough to sell a few aircraft — without resting on those laurels: chasing the best version of the product by changing one element at a time, validating its contribution before moving on to the next improvement.

So industry doesn't systematically lock itself into finding the right product in one shot and then milking it for all it's worth — which is what I thought I'd understood.

One case never makes a rule, so we did some digging to check whether other industrial leaders shared this pattern — and spoiler alert: they do. I picked two other cases:

• At Intel, they pushed this to its limits with the "tick-tock" model⁵: either a smaller process node, or a new processor architecture — never both major changes at once. That strategy ran for 10 years before a new model offered a better answer to the technological challenges.
• At Airbus, which documents its own doctrine⁶, the A320neo delivers a 20% fuel saving over its direct predecessor, the A320ceo, while sharing 95% of its parts and 98% of its tooling with its family cousin, the A321neo — to the point that integrating an A320neo into an existing A320ceo fleet only takes half a day of self-study for already-qualified pilots.

This second example is close to Dassault's context, so we might conclude this only works for heavy industries building things piece by piece. But then Intel comes along and contradicts that conclusion.

What we can take away from these first examples:

• They isolate the variable that changes, which makes diagnosis easier when something goes wrong;
• They validate a hypothesis before industrializing it at scale, which has the advantage of not industrializing the mistake;
• Finally, thanks to these first elements, they can react fast (on diagnosis, and on applying the fix to future production).

Movement 2 - The limits of the small-steps policy

The small-steps policy can have a blind spot: by focusing on one improvement at a time, we might lose sight of the bigger picture, or end up optimizing for one specific customer. Christensen documents this in The Innovator's Dilemma⁷: the very mechanism that makes small steps powerful is what structurally blinds you to the disruption coming from below.

On top of that, there's also the risk Diane Vaughan called the "normalization of deviance" (Challenger)⁸ — when the quality criteria themselves drift, one small step at a time. The next case illustrates that risk well.

Boeing used to be an engineering company driven by product and technology. But for years now, it has followed a small-steps policy aimed at avoiding recertification, rather than mastering technical risk⁹. That shift in governance dates back to the McDonnell Douglas merger: power moved from engineers to finance, and headquarters relocated from Seattle to Chicago in 2001, explicitly to distance leadership from the day-to-day of the factories¹⁰.

The 737 MAX disaster traces back to the moment Boeing's leadership decided to divorce its own culture. Boeing didn't misuse small steps — it changed hands on who was holding the course.

The whole point of the small-steps policy is that the vision has to be carried by strong product leadership — not by a technical or financial execution team. Toyota's "catchball" (Hoshin Kanri) is a good model for this¹¹: the leader sets a goal, each level translates it into smaller goals and sends it back up — an iterative dialogue until everyone's aligned.

Movement 3 - The continuity in agility

Putting all these pieces together, we can trace a lineage between our agile values and practices and these industrial practices — something we've systematically attributed to Lean, but which turns out to run broader and deeper than that.

This small-steps technique we've been talking about since the start of this article — we've implemented it:

• in the Agile Manifesto and its values (working software, responding to change...)¹², and in its most widespread implementation, Scrum (a usable increment, the feedback loop)¹³: this echoes Movement 1 above, and lets us validate before scaling;
• with Lean Startup, we first seek MVP validation and learning¹⁴ — a direct echo of "not industrializing the mistake";
• and finally, the link back to Toyota is just as direct, through Lean Startup and Lean Software¹⁵ practices that carry these industrial best practices forward.

But Movement 2 also warned us about this small-steps policy and its possible consequences: Agility carries execution and delivery (the Manifesto opens with "We are uncovering better ways of developing software by doing it and helping others do it."¹²); it's the product practice that, in complement, holds the vision and the course.

So it's essential to embed this small-steps culture in a context with a strong course, backed by real product leadership — one that takes field feedback and folds it into the long-term vision: the goal being to avoid becoming a "feature factory"¹⁶ or a "feature team," and to actually be a "product team"¹⁷ instead. A team that runs sprints with no product practice holding the vision above it is doomed to follow Boeing's example.

We can run into this risk every day, and we have to fight it every day too. Christensen's Jobs-to-be-Done examples¹⁸ and Rich Mironov's "product sprawl"¹⁹ both echo things I deal with in my own day-to-day work (that's a lot of "every day"s in one paragraph...). But that topic deserves — and will get — its own dedicated article.

Conclusion

This example of Dassault's "small steps" policy is a reminder that Agility was never really a break from the past — at least not once you step back and look at what was already being practiced across industry, on the topics that concern us today. It's the continuation of a discipline already practiced by leaders built to last.

To pick up the thread from my previous article on the death of Agility (French): by embedding Agility's values, and the practices whose value we've measured through the examples above, deep into our company and team values, we could stay on a long path of improvement toward better products.

In the end, what matters isn't aiming for the best product on the first try, or even on every try — it's aiming for victory at the end of the road. The steps that get you there should let you validate hypotheses as early as possible, catch mistakes before they get industrialized, and add a pinch of value every single time.

Small steps aren't a reactive pile-up — they're a structured dialogue that validates the trajectory at every stage, while following a clear course.