Case · M4 × Return (consumer subscription)
Duolingo: a streak with a safety valve
Duolingo made the daily streak its core return mechanic, then let learners equip Streak Freezes that protect the streak on a missed day. Letting people hold up to two freezes nudged daily active learners up instead of lowering practice, per the company (2022).
The outcome · 2022
Letting learners equip up to two Streak Freezes raised daily active learners by 0.38% (company A/B result, Jan 2022). Learners who reached a 7-day streak were 3.6x more likely to finish their course, a correlational figure.
CorrelationalCompany-reported
Read the sourceStage cell: M4 × Return (consumer subscription).
What they did. The daily streak is Duolingo's core return loop, backed by a late-day "streak saver" reminder. A Streak Freeze, bought with in-app gems, protects the streak on a missed day, and Duolingo tested letting learners hold two at once.
The mechanism. The streak is stored value that grows every day HK-INV27 (Hooked Ch.5), and the fear of losing it is Loss & Avoidance CD8 (Actionable Gamification Ch.12, p.307). Each day also extends a visible commitment record INF-T-CC-11 (Influence Ch.3, p.58). The freeze is forgiving lapse recovery HK-RWD34 (Hooked Ch.4): one bad day no longer wipes out months of investment. Streak research points the same way: streaks with a repair option hold up better F22 (On or off track: how broken streaks affect consumer decisions, Silverman & Barasch, Journal of Consumer Research, 2023).
The outcome. Equipping up to two freezes: +0.38% daily active learners (2022) Case grade B. Company-reported numbers.. A new streak animation lifted D7 retention for brand-new learners by 1.7%. Learners on a 7-day streak were 3.6x more likely to complete their course (correlational).
What it means at your stage. Ship a streak only together with its forgiveness mechanic. Define the streak on a real unit of value (one lesson, not one app open), then watch the streak-break → churn rate, freeze usage and the share of actives on a 7+ day streak.
Caveats. Company-reported; method not disclosed; the completion figure is correlational; the effect is small and was measured at Duolingo's scale. Mechanism link Evidence grade B. Moderate evidence: solid studies with some limits., from the company's own experiments.
Would you have called it?
This case is also a round of Predict the Outcome: read the situation, pick what you think happened, then see the evidence.
Play this roundTry this
The techniques behind it
Each one is a starting point, not a recipe. Gamify It turns a technique into an experiment card for your product: the change, the metric, the guardrail and how long the test needs at your traffic.
- TechniqueEvidence grade B. Moderate evidence: solid studies with some limits.
Streaks and completion chains with a safety valve
a visible chain you can repair
Mark each finished unit with a clear 'done' moment and a visible chain of days, and ship forgiveness with it: freezes, repair windows, rest days, partial credit and a running total that survives a break. Never let one miss erase everything.
- TechniqueEvidence grade C. Limited evidence: few or small studies.
Make stored value visible, and exportable
show what users have built
Show users what they have built in the product (their library, history, notes, reputation, working setups) in summaries and anniversaries, and let them export it. Visible value raises the felt cost of leaving; export keeps it a gift rather than a hostage.
- TechniqueEvidence grade D. Practitioner framework or anecdote: plausible, untested.Practitioner heuristic
Grace systems and small executable losses
forgiving loss rules
Threaten a large loss but apply only small ones, and add forgiveness users value but cannot abuse: capped streak freezes, payment grace periods, a second chance when users expect the worst. Chou's practitioner sizes: 2-5% of what users invested is usually enough, 15% at most, never above 30%.
- TechniqueEvidence grade C. Limited evidence: few or small studies.
Honest loss framing
show what the user would lose
When a real loss is at stake, frame the message around it: "$210 in credits expire on June 30; apply them to your next invoice", "On the free plan, 14 of your projects would become read-only." People respond more to a possible loss than to an equal gain, but only true, specific losses hold up.
Limits
What we don’t claim
- The Hook Model is a practical design tool, not a validated or complete theory of behavior.Hooked, Introduction
- Octalysis is a practitioner framework built from observing games, not a validated scientific model.Actionable Gamification Ch.15, pp.399-400
- The evidence is correlational until it is tested in your product.
- No success is attributed to a single technique: real outcomes have many causes.
- Company names are used editorially; no endorsement is implied.
Sources
Influence · Robert Cialdini
Ch.3, p.58
Hooked · Nir Eyal
Ch.4 · Ch.5
Actionable Gamification · Yu-kai Chou
Ch.12, p.307
How Duolingo's streak builds habit · Duolingo blog, 2022
Case grade B. Company-reported numbers.On or off track: how broken streaks affect consumer decisions · Silverman & Barasch, Journal of Consumer Research, 2023
Evidence grade B. Moderate evidence: solid studies with some limits.
Source IDs: CASE-M01HK-RWD34HK-INV27INF-T-CC-11CD8F22
Book content is paraphrased, never quoted at length. Last verified October 24, 2026.
Your product
Find the stage your product is stuck in
Your product has its own version of this story. Find the stage it’s in and the move that fits it. Connecting is free; a Pro subscription starts it.