Marketing automation case study: SMS campaigns and opt-in workflows for financial advisors

The challenge: one client conversation across three messaging providers

Before the build, advisors worked across separate tools and manual processes. They wanted one messaging system made for financial services teams, showing how campaigns performed, how clients engaged and what had been said, while helping teams meet compliance requirements and deliver messages reliably.

The platform needed to meet four requirements:

• Different provider formats. Each messaging provider had its own API, behaviour and webhook format, and message data had to stay in step across all of them without inconsistencies or duplicate records.
• Real-time updates across devices. Advisors wanted message updates to reach team members and their devices as they happened, because delays or sync problems affect whether a conversation is shown correctly.
• Metrics at volume. Response rates, link clicks, appointments booked and conversation grouping had to be counted promptly across high volumes of messages.
• Reliable bulk sends. Campaigns brought queues, deliverability and throttling to manage, along with slow GraphQL queries, large media attachments and checks to ensure each message was sent once.

What we built: SMS marketing automation on one layer over three providers

We built both the front end and the back end: messaging workflows, live tracking and campaign management for advisory teams, at scale. Six pieces carry it:

• One message model for every provider. A provider abstraction layer gives the three providers one message model, with an adapter for each. Provider actions are standardised, and the next provider is easier to add. During multi-provider testing, this layer removed an estimated 25 to 35% of messaging inconsistencies between providers.
• Opt-in and keyword templates. Ready-made workflow templates for opt-ins and keywords, which each business can customise, and workflows that teams adjust with rules of their own.
• No duplicate messages. Tracking and validating each provider's own message ID stops duplicates.
• Live updates across devices. GraphQL subscriptions and WebSocket event triggers push message activity across users and devices as it happens.
• Background jobs and caching. Sidekiq jobs process bulk campaigns, emails, media and provider sync. Redis caches phone number lookups, temporary message data and tokens, and batch-loading in the GraphQL resolvers cuts N+1 queries. Together, the caching and batch-loading saved about 30 to 40% of unnecessary backend load, so dashboards respond faster when many messages go out.
• Reporting. Daily batch pipelines calculate the metrics, group conversations and track engagement, and click tracking shows how each campaign performs.

The stack also includes React for the interface, PostgreSQL, Heroku and CircleCI.

Where the automation stops, and who takes over

There's no AI model in this one: every automated step is code following a rule someone set, from the opt-in templates to the provider adapters. The advisors run the conversations and the campaigns, and the platform handles the repeated work around them.

This is the split our workflow automation starts from: code for the steps that never change, and people for the decisions. A model goes only on a step that needs reading or judgement, and none went into this platform.

Results

Link tracking and engagement reporting show advisors quickly how clients respond, and the adapter design makes the next provider easier to add.

Technologies and tools

Timeline

The build took six months.

If your team's work runs through several systems and the same steps repeat every week, our business process automation services start with a written plan for what happens when a step fails. When the hard part is connecting the systems themselves, our AI integration services list the connectors we build and what each one reads and writes.