AIDM case study

• Context: aidm.me lets businesses send WhatsApp campaigns, reply to customers and manage contacts through Meta’s official API. It sits inside AppX’s ecosystem, which already serves thousands of educators and creators.
• The hook: Only 49% of messages were being delivered. Businesses signed up, got stuck connecting WhatsApp, and left. Contacts with unconfirmed opt-in were going into campaigns unwarned, which is how WhatsApp numbers get banned.
• Core problem: The product had been built around WhatsApp’s API, not around a small business owner. It assumed users already understood Meta’s rules, templates and pricing, and it gave them no way to see whether any of it was working.

Who it’s for

What I prioritised

I triaged with MoSCoW, not RICE. Without analytics there was no honest Reach number to score, so I ranked by user pain and business risk instead.

Layer 1: Fix the foundation

Pain point:

New accounts opened on an empty Live Chat screen with nothing to click, and Setup read “Pending” before the user had done anything.

Instead of:

Adding a hint to the empty screen, which would explain the dead end without removing it.

Solution:

Setup is now the first screen until WhatsApp is verified, with honest status and visible progress. While Meta reviews, users are prompted to import contacts, so the wait isn’t wasted.

Pain point:

Free-plan users could build an entire 6-step campaign, about 10 minutes of work, before finding out they couldn’t send it.

Instead of:

A louder warning inside the flow. People would still lose the work they’d already done.

Solution:

The upgrade or connect prompt now appears the moment someone clicks Launch. I also gave the feature one name everywhere, because it had been called four different things.

Pain point:

Names had to be typed in snake_case. Category, which decides Meta approval and price, had no explanation, and Language looked broken because nothing on screen changed.

Instead of:

A rebuild. The live preview, presets and 3-step flow were already strong.

Solution:

Names format themselves as you type, and Category and Language each get one plain line saying what they do and cost.

Pain point:

Opted In, Opted Out and Unknown appeared on every contact with no explanation, even though WhatsApp only allows template messages to opted-in contacts.

Solution:

Each status explains itself right where users see it, so the ban risk is visible before anyone hits Send.

Layer 2: Build what was missing

Once the basics worked, I went back through the product and wrote a spec for the features businesses needed to run WhatsApp day to day. Engineering shipped it in 4 days.

Pain point:

Every number was trapped inside one campaign. “Is this working?” meant adding up campaigns by hand, and there was nothing to send to a manager or client.

Solution:

An account-level Analytics page with date ranges, messages over time and one-click reports. This was the biggest gap in my review, so it led the spec.

Pain point:

Most owners aren’t copywriters, and a blank template box is where campaigns stall.

Solution:

“Generate with AI” writes a first draft in WhatsApp’s format, so users start from something that works and has a better chance of Meta approval.

Pain point:

After every campaign, replies cluster into the same few questions, and agents typed each answer fresh.

Solution:

Saved replies the team drops into a chat in one click.

Pain point:

Everyone shared one login, so nobody knew who replied to whom, and every agent could see every customer’s number.

Solution:

Separate logins with Admin, Manager and Agent roles. Agents only see the chats assigned to them, and customer phone numbers can be masked from agents.

North Star: messages delivered to customers

That’s the value a business pays for. Delivery rate went from 49% to 77%, and reached 218,000+ delivered messages and 133,000+ people.

Secondary metrics

• Activation: 64% connect WhatsApp the day they sign up
• Engagement: 19K+ customer replies to businesses
• Retention: 44 campaigns re-run by returning businesses

Guardrails

• 0 WhatsApp numbers disabled by Meta
• 77% delivery held at 3× the volume
• 3 in 4 delivered messages read, steady throughout

Per day, before vs after the new features

Risks & mitigations

• Meta bans for messaging unconfirmed contacts: opt-in status explained in place. 0 numbers banned.
• Agents misusing customer data: role-based access and phone number masking.
• Meta quietly reclassifying a template mid-campaign: auto-pause specced for the next phase.

Trade-offs

• Ads manager and in-chat payments: pushed out. Each was a multi-quarter build, and neither helps a business that can’t yet see its own results.
• Template builder: kept, not rebuilt. Fixing the copy gave most of the value for a fraction of the engineering effort.
• Using the product instead of analytics. Slower to prove, but it found a ban risk no metric would have flagged.

Next phases

• Phase 2: readable variables (“Insert: Customer Name” instead of {{1}}), festival templates for Diwali and Holi, auto-pause on reclassification, and auto-resolving old chats.
• Phase 3: click-to-WhatsApp ads and in-chat payments, once businesses are running steady campaigns.