Lull: from installs to protected daysScroll to read, or present one slide at a time.
Similarweb - B2C Product Manager take-home
Lull: from installs to protected days
Why half of new users never let Lull work, what I built to fix it, and how Lull grows and earns from here.
Martin GregorNovember 2026Prototype, deck and research board
Starting assumptions
What we optimise for, and for whom.
Goal: daily active users
Assumption: the goal is daily active users. A blocker does its job when it is not opened, so I define active as protected:
DAU = people Lull protected that day they tried to open a blocked app and the block held. Averaged over 7 days; app opens are tracked next to it.
Counting only app opens would miss 23% of the days Lull did its job: it blocked, and the user never opened Lull.
The design reduces opens on purpose: no timers, a held block closes the app. A blocker that works should need opening less.
Audience: iPhone users losing hours a day to their phone
68% say 3-7 hours a day. 48% US, 16% UK. Two main jobs, one app:
Goal picked (today's wording)
Share
The job
Less social media
33%
An all-day fight with the feed
Focus at work
33%
Protect a work window
Better sleep
22%
Protect the evening
Time with family
12%
Protect set moments
The market: 45% of US adults tried to cut back on their phone last year, and 3 in 4 did not fully succeed (Pew, 2026). Appendix E.
Onboarding already asks the goal. I use the answer to build the whole plan, not to build two products.
Terms used throughout: held block = someone opens a blocked app and does not get round the block. Activation = a first held block within 24 hours of install. Lull data: 1,545 installs, 10 Oct - 8 Nov 2026.
The DAU equation
DAU = installs x activation x retention. Installs are fixed, so the levers are getting people to a first held block, then keeping them protected.
Out of every 100 installs
1. Activation fills the pool
Two in three installs never get a single held block. Without one, almost nobody is protected in week 2 (1.5%).
2. Retention keeps it full
Of those who activate, half are still protected in week 2, and the pool keeps draining after that.
~1,200 non-paid installs a month. Week 2 = days 7-13, users with 13+ days. Day 21 from a fixed cohort of 143 activated users.
Problem 1 - activation
Problem 1: two in three new users never have Lull block anything for them. Most are lost before Lull is even allowed to.
Too long before the ask
6 screens come before Apple's Screen Time ask, and 5-8% of people quit on each one.
"No" is final
After a "no", Lull never asks again by itself: you must tap "Try again" or find it in Settings. Only 112 people did, and 59% of them said yes. So only 3% of everyone who says no ever allows it.
Weekend setups wait
Most first plans block work hours, Monday to Friday. Set one up Friday to Sunday and nothing is blocked until Monday: 38% get a held block in 24h, vs 71% on a weekday.
n = 1,502 users with a full 24h. Exclusive split of permission outcomes within 24h. The 59% is from people who chose to see the ask again, so for everyone it will be lower.
Fix 1 - activation
Fix 1: get more people to allow Screen Time and reach their first held block on day 1.
Where people drop today
What the prototype changes
25% quit before the Screen Time ask
Fewer steps: 3 screens before the ask instead of 6 (a goal, a plan, a promise)
29% tap Don't Allow
A reason to say yes: they commit to their plan first, then a guided ask shows exactly what to tap
Only 3% of them say yes later
Lull asks again itself. A/B test: an easy "Not now" vs asking until allowed
8% quit while picking apps
The plan says what to block: a checklist shows which categories to tick in Apple's list
Weekend setups wait until Monday
2 protected hours right away, plus a practice block (it doesn't count toward the metric)
What Fix 1 could add, and how we would prove it
If 42% instead of 33% get a held block on day 1, 8-24% more people are protected in week 2. The A/B test needs about 6 weeks.
Competitors seen doing parts of this: Opal (personal stake, drawn dialog), ScreenZen and Unrot (ask again), BePresent and SocialLite (hold to commit). Sizing and test design in the appendix.
Fix 1 - in the prototype
From the iPhone home screen to a live plan in about a minute.
1One questionA short chat, a tap moves on
2The planTime on the phone, what an hour back gives, the plan in two lines
3Hold to commitOne slow breath fills the screen
4Coached askShows what Apple will ask, where it will appear
5Apple's list, coachedChecklist ticks as you go
6Try it onceA real practice block, then "you're set"
Real screens from prototype/lull.html. It opens on a simulated iPhone home screen with a 2-second welcome. Screen Time, Face ID and blocking are simulated.
Problem 2 - retention
Problem 2: after activation, protection fades. Social and sleep users fade fastest.
Who fades fastest: week-2 protected, by goal
Goal
Week 2
Less social media
41% (n=121)
Better sleep
44% (n=82)
Focus at work
63% (n=81)
Time with family
65% (n=40)
Where people drop off: of 137 users who went quiet, 85 did so in the first 3 days; 20% turn all blocking off by day 14; 19% of block screens end with a 5-minute break.
Fixed cohort: 143 activated users with 21+ days, so the curve is not a mix effect. Goal table: activated users with 13+ days.
Improvements 2 and 3 - retention
A reason to come back every day, and a block screen that suggests something to do instead.
2. The path: four weeks, one small step a day
27% of activated users go quiet by day 3. The path gives one small step and one sourced fact a day.
HomeHeld count, today's step
Tap the pathThe 4 weeks explained
Day-2 nudgeWritten for sleep
3. The block screen: an instead, not just a no
Today 19% of block screens end in a break. A/B test: "I'll hold" vs a button with something to do instead.
A: I'll holdToday's button
B: Stretch insteadChanges each block
Apple's block screen allows an icon, a title, a subtitle and two buttons, and cannot open Lull. Notes cite their study (Wood 2002, GrĂ¼ning PNAS 2023, Lally 2010 and others).
Monetisation
Charge after a proven week, never before: the users who can pay are the ones Lull has already protected.
Who can pay
~1 in 6 installs is still protected in week 2. Asking for money before the first held block would scare people off at the step everything depends on.
What the market does
The top earners use a hard paywall at the end of onboarding. It is their top complaint: 70% of Unrot's and 67% of BePresent's bad reviews.
The offer, built
Lull Plus: a second path (e.g. better nights alongside deep work), your own rules beyond the path, deeper weekly insights. 7 days free, then $29.99 a year, a reminder 2 days before.
Free forever
Blocking, breaks, editing apps and hours, strict mode and one full path. Never charge for the protection itself. Offered at the week-1 report, never in onboarding.
Reviews: latest up to 1,000 App Store reviews per app (US + GB), tagged by keyword. Price compared with Opal, BePresent and Unrot paywalls. Test: appendix F.
Test plan
Four tests, in an order our traffic allows: one onboarding test at a time, others alongside only when they reach different users.
Test
Primary metric
When
1. Path vs today
Held block in 24h (33% today)
First, ~6 weeks
2. Block screen: hold vs instead
Blocks ending in a break (19% today)
In parallel, ~2 weeks
3. After a no: "Not now" vs ask until allowed
Share who allow Screen Time, among those who first said no
After test 1
4. When to offer Plus
Free trials started per protected user
Step 3 of the plan
Why this order: about 1,200 non-paid installs a month buys one activation test every six weeks. The block-screen test runs on people already protected, and block screens are frequent, so it does not compete for the same users.
Hypotheses and guardrails: appendix F2. Every test is switchable in the prototype (side panel, or You > Prototype on a phone); every event carries its variant.
1-year plan
Year 1: four problems, in the order users hit them.
1
Get people protected on day 1
Problem
2 in 3 installs never get a held block.
Build
New onboarding (A/B test)
Alert if Screen Time access is lost
Session recordings
Goal
Day-1 held block: 33% to 40%+
2
Keep them through week 1
Problem
27% go quiet by day 3; 20% turn blocking off by day 14.
Build
Block-screen A/B test
Day-3 and day-10 milestones
Weekly report, widget, emergency unblocks
Goal
Fewer quiet by day 3, fewer switched off by day 14
3
Fit real phone use, start charging
Problem
Social-media users want limits, not total blocks.
Build
Daily opens limit for social apps
Strict mode, offered in week 3
Plus, after a proven week
Goal
Social users protected in week 2: 41% to 50%
4
Grow through friends
Problem
Word of mouth is the only free way to more installs.
Build
One accountability friend
Do the path together
Referrals
Goal
More installs from word of mouth
No fixed quarters: each step starts once the one before is moving. Built from the feature map: 35 features from 7 competitor recordings, 4 teardowns and App Store reviews. Also in the prototype: You > Where Lull goes next.
How I worked with AI
Seven steps. AI did the reading and the heavy lifting; I did the trying and the deciding.
1I used the prototype myself first, to get a feel for it and what could already be better.Me
2AI ran the data analysis and the market research, down to a shortlist of the top apps worth trying.AI
3I installed and used those apps myself: their focus, their audience, what they do well and badly.Me
4AI looked across my recordings and their reviews for patterns.AI
5I had AI take some of the shipped apps apart (static reading only) to see what runs underneath.Both
6I picked what to apply: competitor patterns that work (why reinvent the wheel), ideas from other apps, and my own.Me
7From all of that I set the direction, and built and tested the prototype with AI.Both
Tools
Claude Code: SQL on 112k events, review mining, the prototype and this deck. Claude Design for screens.
OpenASO: App Store keyword rankings and competitor listings (US + GB).
REA: static teardowns of 4 competitor builds, pulled with Apple Configurator.
Second opinions from Claude and Gemini. A simulated user panel (8,000 real reviews), labelled: not user research.
All research in one place: Research board (market, seven apps first-hand, teardowns, feature map, note sources).
In one line
Get people protected on day 1, give them a reason to stay, and charge only once Lull has proven itself.
1. Day 1
New onboarding: a goal, a plan, a promise, and a guided ask that Lull repeats after a no.
Activation 33% to 40%+
2. Week 1
One step a day, sourced notes, an instead on the block screen.
Fewer silent by day 3
3. Money after value
Blocking free forever. Plus as a second path, offered after a proven week.
Once protection grows (step 3)
My first two weeks
Build the alert for lost Screen Time access, launch test 1 (new onboarding vs today), and watch how many allow Screen Time and how fast they reach the ask.
Who: new installs, paid social excluded (20% vs 37% activation, a one-off campaign).
Sample and duration
~830 users per arm to detect +20% relative (33% to 40%) at 80% power, 5% significance.
~1,660 users in total vs ~1,200 non-paid installs a month: about 6 weeks.
So: test the bundle first; split the parts later. Use the grant rate as an early read, and first-run sessions with 5 users before launch.
Two-proportion test, 33% baseline. Installs per month from the 10 Oct - 8 Nov data excluding the paid social test.
Appendix E - Market
Big demand, young money: 120 million Americans tried to cut back last year, and one app makes $10 million a year.
Demand (US adults)
Adults 18+
~270 M
Say they spend too much time on their phone
53% (70% of 18-29)
Tried to cut back in the last 12 months
45%, about 120 M
...and did not fully succeed
3 in 4
Of them on iPhone (estimate)
~65-70 M
That is Lull's job: helping the attempt stick.
Money
Opal, the leader: over $10 M a year and over 1 M daily users (founder, Apr 2026). Moving to free-first cut download-to-paid from 20% to 9% and still grew revenue.
Concentrated: only 2 of 84 apps in top search spots stay on the top-grossing chart.
Opal top 10 in 17 of 28 keyword lists; Unrot (Jun 2025) grows fastest with "earn your screen time".
Vendor "category size" reports range $0.5-2.8 B for 2025 by definition. Context only, low confidence.
Census Vintage 2025; Pew Research, 1 Sep 2026 (n = 9,750); RevenueCat Sub Club with Opal's founder, 29 Apr 2026; App Store US scan, 15 Nov 2026 (nichestillopen.com); OpenASO, 12 Nov 2026. iPhone base and per-app revenue other than Opal's are third-party estimates. Links: market sizing note, Research board.
Appendix E2 - Lull's slice
At today's traffic any paywall is a rounding error. Protected users and installs are worth more than price, so DAU stays first.
Year 1, bottom-up
Installs
~1,500 a month (Lull data)
Protected in week 2
~20 per 100 after Fix 1
Start a Plus trial at the week-1 report
15% (10-25%), assumption
Trial to paid
~40% (health and fitness median)
Price
$29.99 a year
Paid per 100 installs
~1.2 (Opal: 9)
Revenue, year 1
~$6.7 K ($4-11 K)
What it means
Even Opal's 9% would make only ~$50 K on this traffic. Growth in protected users and installs comes first.
The ceiling: Opal's $10 M at ~$0.36 per install needs ~2.3 M installs a month. Later, Lull needs more value per user too.
Price check
$29.99 sits at the low end: Opal $60-100, Unrot $70, BePresent $100 (family) a year. Fine for launch; test $29.99 vs $49.99 once Plus is live.
Trial to paid: RevenueCat State of Subscription Apps 2025, health and fitness median 39.9% (via Athletech News). Week-2 protected: analysis/04_retention_dau.md. The trial-start rate is my assumption; Lull data is synthetic and 30 days long.
Appendix F - Monetisation - test design
Plus offer timing: day 7 vs week 3. Trial starts as the fast read, protected users as the guardrail.
The test
Hypothesis: offering Plus after a proven week converts better than offering it in onboarding, and costs no protection.
Variants: offer at the day-7 report vs at the start of week 3. Never in onboarding, never before the first held block.
Primary: trial starts per protected user. Revenue is too slow to read in a small app.
Starts when Lull begins charging (step 3 of the plan), after the activation tests: the same traffic limit, so in sequence, not in parallel.
Duration is set from the trial-start baseline once Plus is live; before that there is no rate to size on.
Decision rule: keep the later offer unless it lowers trial starts and leaves protection flat.
Plus is built in the prototype at the week-1 report (Jump to: Day 7 + Plus).
Appendix F2 - Test plan detail
The four tests in full: hypothesis, metric, guardrails, timing.
Test
Hypothesis
Primary metric
Guardrails
When
1. Path vs today
A goal-built plan, a promise and a coached ask get more installs to a held block
Held block in 24h (33%). Fast read: grant rate, seconds to the ask
Week-2 protected, day-1 uninstalls
First, ~6 weeks
2. Block screen
An instead on the button turns a "no" into a decision
Blocks ending in a break (19%); re-tries in 5 min
Pauses, path turned off
In parallel, after activation, ~2 weeks
3. After a no: "Not now" vs ask until allowed
Asking again until allowed wins back more of the 29% who say no
Share who allow Screen Time in 24h, among those who first said no
Drop-off, uninstalls, App Review
After test 1, or early on grant rate
4. When to offer Plus
An offer after a proven week converts better without hurting protection
Free trials started per protected user
Week-2 protected, path off
Step 3, day 7 vs week 3
Fast reads for test 1: Screen Time grant rate and seconds to reach the ask (both logged in the prototype).
Appendix G - What iOS allows
What I designed around: Apple decides what a blocker can show and when.
No app list, no pre-ticks
Apps cannot list what is installed or pre-tick Apple's picker: each pick is a sealed token only Apple's list can create. So the plan names categories, the picker is coached, and later edits open pre-ticked.
A fixed block screen
Apple's shield is an icon, a title, a subtitle and two buttons, and cannot open Lull. So the instead lives on the button and the win is counted, not shown in a new screen.
No block before permission
Nothing can be blocked until Screen Time is allowed (then Face ID). So: a promise before the ask, a real practice block after. A drawn copy of Apple's dialog is an App Review risk; Opal, ScreenZen and BePresent ship it.
Apple FamilyControls, ManagedSettings and DeviceActivity frameworks; competitor patterns from screen recordings, 13-14 Nov 2026.
Appendix H - Assumptions
What I assumed, and what would change my mind.
Assumption
Why
What would change it
DAU-first at the start
Starting assumption; for a blocker I define active as protected
A target for revenue in year 1
Small team, ~1,200 non-paid installs / month
From the data
More traffic = more, smaller tests
Keep the app-blocking focus
The brief
-
Activation drives retention
50% vs 1.5% week 2, holds in every channel
Fix 1 test lifts activation but not week-2 protected
Paid social is excluded from tests
20% vs 37% activation; one-off campaign
A plan to scale paid social
Fewer opens is fine
Lull works when it is not opened (23% of protected days)
Protected users fall while opens fall
The data is synthetic (per the brief). All effects are correlations until tested.
Appendix I - How I checked the numbers
How I checked the numbers
Cohort: users with a full 24h of follow-up (n = 1,502). One rule in every script.
Every held block has a matching block screen 5 minutes earlier (31,046 / 31,046).
No events before install or after the data end.
Retention curve re-run on a fixed cohort (143 users, 21+ days): same shape.
Confounders: activation gap holds in every channel and both app versions.
Weekday bias: D7 sits on the install weekday; week 2 used instead.
Small n flagged with 95% confidence intervals (late activators: 17-49%).
Retracted: "1.1.0 raised grant rate" (51.5% vs 51.4% on non-paid users, p = 0.62).
Appendix J - Other findings for Q&A
Other findings for Q&A
Paid social test: 20.2% vs 36.9% activation (p < 1e-7); 55% of installs in the 19 Oct week. Its activators also stick less (34% week 2 vs 42-67%).
"Start a session" without permission: 67 of 67 refused; 1.5% ever grant. A dead end. In the prototype, every blocked action without Screen Time opens Apple's ask instead.
Weekend dip: protected users -17% at weekends, all from Mon-Fri schedules.
Most blocked apps: Instagram and TikTok (~4.4k held blocks each), WhatsApp (3.7k), Snapchat, Reddit, Facebook, X, YouTube.
Breaks: 19% of block screens, flat across apps. Highest for Telegram and Discord (22%).
Strict mode: 100% of its block screens hold, but only 7% of all block screens.