Learning a Language in 12 Months, Product-Management Style

In May 2021 I picked up Benny Lewis's book Fluent in 3 Months, and the claim seemed wild enough that I wanted to test it. I chose French, a language I knew nothing of beyond "bon appétit". Reading it, I kept hitting ideas that sounded lifted from Lean Startup: focus on basic conversations first, experiment, attack your weak points. So I asked what happens if you run language learning the way you run a product. Twelve months later I passed a B2 exam with 90 out of 100.

Here are the five product techniques that got me there, and the failure that started it.

Why does grinding an app not teach you to speak?

Because you are optimizing an output that does not transfer to the outcome you want. If you wanted to build a house, your first move would not be picking up a hammer, it would be drawing a plan. Yet in language learning we go straight from "I want to learn French" to translating sentences about a cat eating pizza. No plan, just hammer swings.

That was my first month. I ground away at Duolingo without any learning plan, got frustrated and unmotivated, and then went a full month without touching French at all. Project failed.

The research explains why. Transfer learning, the assumption that getting good at task A (an app) delivers task B (speaking), does not hold here. If you want to learn to speak, you have to learn by speaking.

How do you research before you start?

The same way you would before building a product: look for the best existing solutions to the problem instead of reinventing one. Before we build, we survey what already works. I had skipped that step entirely.

When I did the research, I found the Comprehensible Input theory from linguist Stephen Krashen, which holds that acquisition comes from constant exposure to content you understand about 70% to 80% of and that genuinely interests you. That single finding reshaped what I did daily. Motivation returned, but motivation alone does not get you there. You still have to define what success is.

What is the north star metric for learning a language?

Something you can measure this month, not next year. "Learn a language" is vague enough that plenty of unproductive activity can feel like progress while you go nowhere.

I made the goal specific: hold a 60-minute conversation with a native French speaker. Researching how language ability is actually measured, I found the CEFR scale that runs A1 to C2, and my goal mapped to roughly B1 to B2. The problem was that those exams run once or twice a year. Working for twelve months to discover at the end whether I had made it is the opposite of how I set goals as a PM.

So I tracked monthly speaking minutes instead, as the leading indicator for that 60-minute conversation. At the start the number was zero.

How do you get from zero to your first conversation?

Build the loop before you optimize it. My first goal was small: introduce myself, ask basic questions, understand the answers. I deliberately avoided grammar lessons and long explanations of language structure, and focused on forming simple sentences about myself.

In Atomic Habits, James Clear makes the case for standardizing before optimizing: build the small habit loop and run it, rather than designing a perfect process and starting later. With each conversation loop I added two or three new details about myself, which got harder once I had exhausted family, country, and my own strange language-learning goals.

Those conversations are also your test environment. Try new sentences and watch whether your partner nods along or looks startled at how badly you just used a word.

To track it I logged minutes across my two main activities: listening, mostly podcasts, and speaking, mostly hours on iTalki with a handful of excellent teachers. The more listening I banked, the easier it became to move into real conversation and grow the speaking-minutes number.

Why do outcomes beat outputs here?

Because an output is something you did, and an outcome is something that changed in the world because you did it. As PMs we know that "we shipped 10 features this month" is the road to product failure; the job is to change users' lives, not to generate activity.

Memorizing the thousand most frequent French words is an output. At the end you have a thousand words in your head, which is a different thing from using any of them with a person. Phoning a French newspaper to take out a subscription, and reading out my credit card number to do it, was an outcome. My first Duolingo month produced impressive outputs and no outcomes at all.

You can make these small. Cooking from a recipe in your target language counts, though it helps to learn the difference between salt and sugar first.

What do you do when you hit a wall?

Treat it as evidence, the way you would a failed experiment. As PMs we want to find out quickly that an idea does not work, and we celebrate that, because we just avoided wasting months. Learning a language will fail on you repeatedly. Failure is not a detour from the route to fluency, it is the route. Hitting a wall means you reached the edge of what you had, which is the only way to push it further.

The insights from my first failed month are what found the routes that worked. And when the wins come, small or large, stop and mark them. That is the fuel for the next wall.

In May 2022 I sat the B2 exam at the Institut Français and scored 90 out of 100. I deliberately did not study for it, because I wanted it to measure my French rather than my exam preparation. The formal result mattered less than what the language actually opened up: people, stories, and experiences I would never have reached otherwise.

Key takeaways

• Plan before you swing the hammer. A month of unplanned app grinding produced nothing but burnout. Research your problem the way you would before building a product.
• Pick inputs you understand and care about. Comprehensible input, content you follow about 70% to 80% of and actually enjoy, beats drilling.
• Choose a metric you can move this month. Monthly speaking minutes was the leading indicator; the once-a-year exam was not a usable goal.
• Standardize before you optimize. Run a small, repeatable conversation loop and add to it, rather than designing the perfect study system first.
• Chase outcomes, not outputs. A thousand memorized words is an output. Ordering a newspaper subscription over the phone is an outcome.
• Walls are progress. You only hit the boundary by reaching it, and every failure narrows where the working route must be.

Products are never done, and neither are languages. My next goal is to speak at a French tech or product conference, which still looks like a mountain from here. If you are applying product thinking to a goal outside work, or want a sounding board for the one inside it, that is what I do in product mentoring.