Upinion — Internal product decision

Where do we go?
Social messaging vs. custom PWA

Two viable platform paths. Each solves a different version of the same problem. This page exists to make the trade-offs explicit so the choice is deliberate, not default.

WhatsApp reaches 60% of Algeria[1] Meta banned general-purpose bots Jan 2026[8] 7% of Sahel migrants visit websites for info[5]

Access & Reach

Penetration in Algeria (active monthly users)[1][2][3]
A
WA 60% · FB 33% · TG 36%
B
~7% effective visit rate
Cost to user (data cost to engage)[7]
A
Zero-rated on Mobilis/Ooredoo
B
Paid data required
Smartphone-only viability (works on low-end devices)[6]
A
Works on lower-end Android
B
Depends on browser support
Familiar surface (users already know how to use it)[5]
A
High — daily-use surface
B
Low — only 7% visit websites

Engagement UX

Form / multi-step flows (survey & onboarding screens)
A
Flows — capped 8 components/screen
B
Full control
Voice + text answers (audio-first input for low-literacy users)
A
Chat layer only — not in Flows
B
Full control
Persistent info hub / archive (tabs, history, search)
A
Not possible in platform
B
Yes — full app shell
Scroll-feed / TikTok-style pattern (infinite scroll content)
A
Impossible in platform
B
Yes — fully buildable
Push notifications (re-engagement after user closes app)
A
Native + topic-based
B
PWA push — requires opt-in install
Return path after closing (user finding us again)
A
Native — message arrives
B
Browser history unreliable

Cost Structure

Per-message fee (variable cost at scale)[9]
A
$0.02–0.05 marketing; free utility in window
B
Zero per-message cost
Approval lead time (time before we can launch)
A
2–4 weeks WABA review
B
Zero — we control deploy
Hosting cost (infra spend)
A
BSP fees on top of Meta
B
CF Pages free tier

Risk & Resilience

Vendor policy exposure (platform risk)[8]
A
High — Meta bot ban Jan 2026
B
Zero — open standards
Pseudonymous branding (operating under a non-legal entity name)
A
Viable with linked microsite
B
Trivial — we control domain
Surveillance surface (who sees user metadata)
A
Meta sees all metadata
B
We control what we collect
Long-term portability (ability to migrate off platform)[10]
A
Locked to Meta ecosystem
B
Open standards — portable

Synthesis — Balanced view

Option A wins reach by ~8.5×, but trades UX control and policy stability. Option B wins on control but only converts the 7% of users who visit websites. Hybrid is plausible — A for acquisition, B for engagement. If reach is the priority, Option A wins decisively — 60% Algeria penetration vs an effective ~7%. Messenger and Telegram add another layer of redundancy. The cost-to-user advantage (zero-rated data) further compounds the gap for low-income migrants. If UX control is the priority, Option B wins — Flows cap at 8 components/screen and forbid voice input mid-form. A persistent info hub, scroll-feed patterns, and custom voice-first flows are only possible in a PWA we control. If cost is the priority, Option B wins on margin (zero per-message), but Option A is cheaper to acquire users at scale (carriers zero-rate WhatsApp). The real cost question is: what's the value of a conversation you could never have started? If reliability matters most, Option B wins — Option A's policy surface (Jan 2026 bot ban) is the largest single risk. Meta can revoke WABA access faster than we can migrate an established user base.

✓ Decided — Round 3 · 26 May 2026 (afternoon session)

Option A — WhatsApp wins. We commit.

Gerben: "PWA is great but if it's not used, that's the most important thing." The team agreed to drop the standalone PWA bet and design the full experience inside WhatsApp using Flows. Round 4 explores how the conversation patterns work in detail — see the new interactive prototype.

Revisit if...

  • Meta enforcement tightens further — a second policy shift that affects structured-survey bots would force a re-evaluation. Quarterly check.
  • Trust-at-first-contact fails in pilot — if "Anwar" can't get past the phishing-suspicion barrier without partner referrals, the architecture holds but the go-to-market changes.
  • Compliance / security review demands zero Meta exposure — then we fall back to a Telegram-only stack (free BotAPI, no Meta) and accept the ~36% reach cap.[3]