If you've used Duolingo, Memrise, or any of the streak-based language apps, you've felt the pull of XP. Open the app, do the lesson, see the number go up, feel good. It's a clean loop and it's effective at one specific thing: getting you back tomorrow.
We thought hard about whether to ship something similar in SmartWords. We decided not to. Here's the reasoning.
What XP optimises for
A daily XP target is a near-perfect engagement metric. It's legible, it resets, it's visible from the home screen, and it's easy to push notifications around. From a product-metrics standpoint, it's almost unbeatable for short-term retention.
The problem is that "did the user hit their XP target today" and "did the user actually advance their language" are different questions, and the difference grows over time. A learner who chases XP will reliably:
- Pick the easiest available exercises (XP per minute is highest there).
- Re-do drills on material they already know cold.
- Avoid the part of the app where they'd encounter unfamiliar grammar, because that has worse XP-per-minute.
After three months, the XP score and the actual proficiency gain decorrelate sharply. We've seen this in our own data and in the literature.
What we do instead
SmartWords surfaces two metrics, both deliberately quieter than an XP counter:
- Active vocabulary — words you've successfully produced (not just recognised) at least three times in the last 30 days.
- Course progression — the CEFR-aligned curriculum, with each unit marked complete only when the production-side checks pass.
Neither resets. Neither has a daily target. You can't lose them by skipping a day. The downside is that they're worse engagement metrics for the first two weeks; the upside is that they keep meaning something at month six, which XP doesn't.
Is this the right call for everyone?
No. There's a real cohort of learners for whom XP is the only thing that gets them back to the app, and those learners are better served by a streak-driven product. We're not trying to capture that cohort. We're optimising for learners who want a credible measure of "am I actually getting better at this language" — and that's a different product.