A teardown of the first five minutes of Words With Friends, checked against 450 recent App Store reviews and public benchmarks, and a redesigned first match you can play on the right.
No sign-up. We'll pair you with a player who's online now.
Psst, the tiles above are draggable.You're 2 turns in. Save your progress so this game follows you to any device.
Before your first word: 0 sign-up screens, 0 mode pickers, 0 ads, 1 button.
Getting in is already fast: no sign-up wall, guest play, and an opponent labeled "Plays Back Quickly." The problem is what each screen asks of someone who has never played.
123
45
678
91011Screenshots from a fresh install. Other players' photos are blurred.
I pulled the 450 most recent US App Store reviews (Jul 21 – Oct 3, 2026) from Apple's public review feed and keyword-coded the 243 one- and two-star reviews by theme. A review can match more than one theme.
Few reviews mention onboarding. That fits: people who quit in the first five minutes rarely write reviews, so retention data has to answer that question. That's what the A/B test is for.
The first screen asks a new player to choose a mode, answer an invite, or explore events. Every option slows down the one thing that matters: placing a first word.
Evidence: screen 1 (callouts 1–3).
Fix: one button, "Play your first word." Modes, events and invites appear after the first word is played.
No coach marks for dragging tiles, bonus squares or the Play button. Boosts and coins are shown with no explanation. The opponent has already moved.
Evidence: screen 3 (callouts 6–8).
Fix: the new player moves first. Three short tips appear at the moment each one is needed, and each can be dismissed.
An invalid word still shows a point preview, and a rack of three Es offers no way forward. "All vowels" racks are a recurring theme in reviews about rigged tiles.
Evidence: screen 4 (callouts 10–11); 11 reviews about unfair tiles.
Fix: a balanced opening rack that can always make a 4-letter word. No score preview on invalid words. A free hint is offered after two misses.
A casino banner shows during the first turn, and players describe an ad after every move. A new player meets the monetization before they've enjoyed the game.
Evidence: screen 4 (callout 9); ads in 57.6% of negative reviews.
Fix: no ads in the first match, and no real-money gambling ads before the player's age is known. Test the revenue cost against D7.
The prototype at the top of this page is the redesigned flow. It's built to feel like a native app: tiles track your finger 1:1, carry your throw velocity into a spring, and can be grabbed again mid-flight.
"Play your first word" replaces the mode picker, invite banner and event tiles. Sign-in lives behind a text link for returning players.
Three tips, shown in context: drag to the star, bonus squares, tap Play. The real board, scoring and bonus colors are kept, so nothing has to be re-learned later.
The point badge appears only when every word is valid. Errors name the problem ("Leave no gaps between letters"), and a free hint is offered after two misses.
Saving progress and notifications come after turn 2, tied to a reason ("Tell me when Sam plays"). Simulated opponents are labeled.
New players meet choices, unexplained systems and ads before they've played a word. US active users fell about 36% across the second half of 2025, so every new install has to count.
Raise D1 retention of new installs by 2 points or more, with a median time to first word under 60 seconds and no loss in D7 retention.
Given a first-time install, when the app opens, then one primary button ("Play your first word") is shown, with no mode picker, invite banner or event tiles.
Given the player taps it, then a board loads within 3 s (p90), the player moves first, and tip 1 points to the center star.
Given the opening rack, then it has 2–4 vowels and can make at least one common 4-letter word. No all-vowel or all-consonant rack in the first 3 turns.
Given tiles are placed, when any formed word is invalid, then no point badge shows and the message names the word. After 2 invalid attempts, the free Hint is highlighted.
Given a tip is showing, when the player taps "Hide tips", then all remaining tips are suppressed for this install and ftue_tips_hidden logs the step.
Given matchmaking for a first match, then the opponent was active in the last 5 minutes and is within 2 levels. A simulated opponent is labeled "Practice partner."
Given arm C, when fewer than 10 turns of the first match have been played, then no interstitial or banner ad renders.
Given the player's second word is played, then the save sheet appears once, can be swiped away, and guest progress is kept either way.
| Metric | Definition | Target / benchmark | Role |
|---|---|---|---|
| D1 retention | New installs who open the app on calendar day 1 ÷ new installs | +2 pts vs control | Primary |
| Time to first word | Median seconds from first open to first valid word played | < 60 s | Secondary |
| First-word rate | Installs who play a valid word in session 1 ÷ installs | Up vs control | Secondary |
| Tip completion | Players who finish all 3 tips ÷ players shown tip 1 | ≥ 80% | Secondary |
| D7 retention | New installs who open the app on day 7 ÷ new installs | No drop > 0.5 pts | Guardrail |
| Revenue per install | Ad + IAP revenue per new install, first 7 days | No drop > 5% | Guardrail |
| Crash-free sessions | New-user sessions without a crash ÷ sessions | No drop > 0.2 pts | Guardrail |
If new players get one clear button, learn inside a real first match, and don't see ads until it ends, D1 retention rises by at least 2 points, because they reach a played word sooner and leave with a game in progress.
| Arm | Experience | What it isolates |
|---|---|---|
| A · Control | Today's flow | Baseline |
| B · Guided first match | Single-button open, player moves first, 3 tips, balanced rack, free hint, save prompt after turn 2 | Effect of the onboarding changes (friction points 1–3) |
| C · Guided + ad-free | B, plus no ads until the first match ends or 10 turns | Added effect of delaying ads (point 4), and what it costs |