Onboarding and Activation

Onboarding is not a tutorial. It is a sales process wearing a friendly hat, and its only job is to get a stranger to the moment where they understand why your product matters, before they lose interest.

The concept

Activation is the point where a new user first experiences real value. Not signup. Not "completed the tour." The moment the product actually works for them.

You have to define it concretely for your product, because you'll optimise it:

Time-to-value (TTV) is how long that takes. It is one of the highest-leverage numbers in your product: shorten it and D1, D7, and conversion all move together.

The activation funnel

   INSTALL / VISIT
        │  ← lose 20-40% before anything happens
        ▼
   ONBOARDING START
        │  ← every screen costs you people
        ▼
   SETUP COMPLETE
        │
        ▼
   ⭐ FIRST VALUE (activation)   ← the number that matters
        │
        ▼
   HABIT FORMED (loop repeated 3+ times)

Every stage leaks. Your job is to know where, which requires instrumenting each step (Chapter 38). Without that data, onboarding optimisation is decoration.

Two philosophies, both correct: for different products

Short onboarding: get out of the way. Minimal setup, land in the product, teach in context. Right when value is obvious and immediate, a note app, a calculator, a utility. Every extra screen is pure loss.

Long onboarding: invest before you ask. 15-40 screens of questions, visualisations, and personalisation before the product proper. Counter-intuitive and it converts, for four reasons:

1. Sunk cost. Five minutes of answering makes abandoning feel like waste.
2. Personalisation. Every answer makes the result feel built for them.
3. Emotional framing. You surface the pain and the possibility before you ask for money.
4. Qualification. People who complete 30 screens are serious.

This pattern powers most high-grossing consumer subscription apps, Noom, Cal AI, Fastic and their imitators, and Duolingo has published that goal-first onboarding moved free→paid conversion from roughly 4% to 9%.

Choose by whether your value needs framing. A calculator doesn't. A 90-day behaviour-change program does, the user cannot evaluate it in ten seconds, so you must build the belief first.

The consumer quiz-funnel, generalised

If you're building a consumer subscription product, this is the pattern to know:

 1. HOOK ─────────── one line: who they become, not what the app does
 2. SEGMENT ──────── 1-2 questions; makes everything after feel tailored
 3. GOAL ─────────── multi-select, in their words
 4. VALIDATE ─────── "most people never start. you did." (in-group + credit)
 5. QUANTIFY PAIN ── one question whose answer becomes a NUMBER
 6. LOSS → GAIN ──── animate THEIR number. Their data, not a statistic.
 7. ASSESS ───────── 8-15 questions. This is the investment. Don't skimp.
    └ every ~4: a progress visualisation ("7 days to sharpen focus")
 8. AUTHORITY ────── why this works. Real names, honestly attributed.
 9. SOCIAL PROOF ─── real testimonials. If you have none, omit — don't invent.
10. COMMITMENT ───── a signed/drawn/spoken promise
11. ANALYSIS ─────── theatre. 5-8 animated steps. Makes the plan feel computed.
12. ACCOUNT ─────── "save your progress", not "sign up". Before the paywall.
13. PLAN REVEAL ─── stats + now→goal + timeline. A concrete deliverable.
14. PAYWALL ─────── annual default, savings %, restore, terms + privacy
15. (optional) ──── honest intro offer on exit. Real offer, no fake timers.

Every question must visibly feed something. A question that changes nothing is pure friction, and users can tell.

B2B onboarding is a different exercise

In B2B, empty state is the enemy. A fresh workspace with no data is unusable and unevaluable. Ship templates, sample data, or an importer, this single thing moves B2B activation more than any tour.

The signup wall

Where you put the account is a major decision:

For consumer, anonymous-first is usually correct (Chapter 21): let people use the product, then ask for the account at a moment where it buys them something, "save your progress across devices" rather than "sign up to continue."

📐 Best practice

Define activation as one measurable event, then instrument every step to it.

Cut every screen that doesn't earn its place. For each, ask: does this reduce time-to-value, or increase belief? If neither, delete it.

Personalise with the user's own data. "Based on your 2 hours a day, that's 730 hours a year" beats any statistic you could quote, because it's theirs.

Ask for permissions in context, never upfront. Request notifications when the user sets a reminder, not on launch. Upfront prompts get denied, and denial is often permanent.

Show progress. A progress indicator meaningfully reduces abandonment in long flows.

Make the first win happen in the first session, even if it's small.

Defer the account until it buys them something.

Instrument per-screen drop-off. You cannot optimise a funnel you can't see.

Design the empty state as a real screen. It's the most-viewed screen in most products and usually the least designed.

💀 Common mistakes

A carousel of feature screenshots. Nobody reads it; everyone taps skip. It delays value without building belief.

Asking for data you don't use. If you collect it, act on it visibly, or don't ask. Asking and ignoring is a small broken promise at the exact moment you're establishing trust.

Permission prompts on launch. Notifications, location, contacts, all denied, often permanently.

Signup wall before any value (consumer). The largest single drop-off available to you.

A tour instead of a doing. Guided tours teach less than one real completed action.

Empty states that are actually empty. "No items yet" with no path forward.

Long onboarding with no payoff. Length only works if it ends in something that feels computed and personal.

Not resetting animated screens on re-entry. Navigate back and the animation shows its end state, looks broken.

Fake urgency and fabricated social proof. Countdown timers that reset, invented review counts, testimonials attributed to people who don't exist. Beyond the ethics, these violate app-store rules (Chapter 36) and are trivially detectable.

Optimising onboarding while retention is broken. You'll get more people to a product they still leave.

The professional workflow

 1. DEFINE ACTIVATION — one measurable event

 2. MEASURE CURRENT TTV and per-step drop-off (Chapter 38)

 3. CHOOSE YOUR PHILOSOPHY
    Value obvious in 10 seconds → SHORT
    Value needs belief/framing  → LONG

 4. MAP THE MINIMUM PATH to first value. Delete everything else.

 5. FOR LONG FUNNELS, build in this order (highest ROI first):
      a. loss→gain moment using THEIR number
      b. commitment device
      c. plan reveal / analysis theatre
      d. segmentation + progress visualisations
      e. authority + social proof (real only)

 6. HANDLE THE STATES
    empty · error · permission denied · returning user ·
    re-entry into animated screens

 7. PLACE THE ACCOUNT where it buys them something

 8. INSTRUMENT EVERY SCREEN

 9. TEST ON 5 PEOPLE WHO'VE NEVER SEEN IT. Watch, don't help.

10. A/B TEST length and order — once you have volume

Tools, websites & costs

Alternatives & trade-offs

Progressive onboarding, teach features as they become relevant, over days, rather than upfront. Lower initial friction, higher long-term comprehension. Requires more engineering (contextual triggers) and analytics to know what's been seen.

Checklist onboarding, a persistent "getting started" list. Very effective in B2B (Slack, Notion). Uses completion bias; visible, non-blocking.

Templates / sample data, skip setup entirely by giving them something real to react to. The single strongest B2B activation lever.

Concierge onboarding, a human sets it up on a call. Terrible economics, superb learning. Right for early B2B and high-ACV products.

Reverse trial, full access to premium features for N days, then downgrade to free (rather than a card-gated trial). Increasingly popular; users see the ceiling before deciding.

Trade-off: long funnels convert better and cost more to build, maintain, translate, and test. Short funnels are cheap and leave conversion on the table for products whose value needs framing.

Checklist

• Activation is defined as one measurable event
• I know my current time-to-value and per-step drop-off
• I chose short or long deliberately, based on whether value needs framing
• Every question feeds something visible
• Personalisation uses the user's own data
• Permissions are requested in context
• The account is placed where it buys them something
• Empty state is designed as a real screen
• Animated screens reset on re-entry
• Returning users can skip to sign-in
• No fake urgency, no fabricated proof
• Every screen is instrumented
• Tested on 5 people who've never seen it

📓 Case Study: copying the mechanics, refusing the tricks

Project: SOLIS, rebuilding a consumer onboarding funnel.

The method, worth stealing. Rather than guessing, a competitor doing ~$20k MRR was studied directly:

"go through a competitor app who is doing 20k MRR, and watch this recording of their onboarding process, I want all of the process replicated into our onboarding but with our theme and not copy paste, any extra page they have, all I want (because its proven and working)"

Three steps: install and complete the competitor's funnel with a screen recorder running (5m36s, ~40 screens) → transcribe every screen into a pure-observation document → write a separate adaptation plan mapping each mechanic into their own product.

📐 Keeping observation and adaptation in two separate files is the discipline that makes this work. If they're in one document, the competitor's copy is sitting next to your plan, and reusing it is easier than rewriting it. Separate files force a translation step.

What the teardown extracted, twelve reusable mechanics, none of them copy: a long invested quiz; segment-first; mid-quiz validation; loss→gain visualisation with a personalised number; progress visualisations; authority quotes before the ask; social proof; a signed commitment; "analysing…" theatre; account before paywall; plan reveal with stats; annual-anchored paywall.

The translation, not transcription. The competitor's commitment device was "Let's make a contract" with a drawn signature. SOLIS's equivalent became The Oath, a Roman sacramentum, the oath soldiers swore on enlistment, with four affirmations and a signature drawn on screen that gates the continue button.

Identical psychology (Cialdini's commitment-and-consistency, made physical). Completely different execution. The user isn't signing a contract with an app; he's enlisting. That's what "copy the mechanic, not the copy" means in practice: find the thing in your world that does the same psychological work.

Personalisation from the user's own data. One quiz question asked how many hours a day vanish to distraction. The answer became an animated counter, grayscale painting, −730 hours a year at 2h/day, under a Seneca line, then flipped to colour and +730 hours reclaimed.

The number is theirs, computed as hours × 365, not a statistic. Honest and more persuasive for it.

⚠️ What was deliberately refused, the most instructive part.

The teardown documented four mechanics that were not copied, each recorded with a reason:

1. Fake urgency. The competitor showed "One-Time Offer! 80% OFF, expires when you leave this screen." Rejected: "a fake countdown is not [fine]. Keep the downsell honest." Beyond App Store Guideline 3.1.1 risk, it's a lie told in the first five minutes of a relationship you want to last 90 days.
2. Fabricated testimonials. The competitor's read as invented. SOLIS was pre-launch with zero reviews, so the rule became: clearly-illustrative quotes now, real TestFlight testimonials before launch, and, notably, no invented review-count badge. Inventing "4.8/5 (1,194 reviews)" is fraud and trivially disprovable.
3. In-app rating prompt during onboarding. The competitor fired it before the user had used the app. Refused on practical grounds too: ratings from users who haven't experienced the product are uninformed, and iOS limits how often you can ask.
4. Attribution beyond a simple "how did you hear about us." This one had a concrete technical consequence, the social-login SDK ships the Facebook SDK by default, and it was explicitly disabled to keep it out of the binary, because its presence would have forced a heavier App Store privacy label whether or not it was used.

⚠️ Deviation 1: a gender question was built, then cut.

It was copied from the competitor's segmentation, with sub-copy promising to "tailor your path." Nothing was tailored. It was removed with the right reasoning: don't ask for data you won't act on. Asking and ignoring is a small broken promise at the exact moment you're building trust, plus it contradicted the product's own validated positioning (Chapter 06).

⚠️ Deviation 2: the funnel has no analytics.

A 19-question funnel with 8 interstitial screens, two paywall placements, and zero instrumentation. Per-screen drop-off is unknown. Which of the twelve borrowed mechanics helps and which adds friction is unknown. Whether 19 questions is better than 12 is unknown.

The plan itself flagged the open question, "DoReset is ~40 screens. Long converts, but is a lot to build/maintain. Full set, or a leaner ~20-screen cut? Recommend building P1-P2 first, then judging length live." You cannot judge it live without measurement. (Chapter 38.)

🚩 Unvalidated. No users, so no conversion data. The mechanics are borrowed from a product with real revenue; their effect here is unmeasured.

Lessons

1. Onboarding is a sales process, not a tutorial.
2. Copy mechanics, never copy. Find the equivalent in your own world, a contract becomes an oath.
3. Two files: observe, then translate. Mixing them guarantees plagiarism by convenience.
4. Long funnels convert when value needs framing, and every question must visibly feed something.
5. Personalise with the user's own number. Theirs beats any statistic.
6. Know which manipulations you're refusing, and write down why. Fake countdowns, invented reviews, premature rating prompts.
7. Never ask for data you won't act on.
8. Check what your dependencies ship, an SDK you didn't invite can change your privacy label.
9. Request permissions in context, never on launch.
10. A beautiful funnel with no analytics is a guess. Instrument before you optimise.

Next: Chapter 11: UX Fundamentals →

This completes Part II, Defining the product.