Five builds. Three written up.

The problem

Outbound was running out of a spreadsheet and a shared inbox. Leads were researched by hand, enrichment happened in a second tool, and the CRM was updated days later, if at all. Nobody could answer which message produced which meeting, so the team kept sending the messages that felt good.

What we built

• A deduplicated prospect database with the source and date behind every record.
• An enrichment step that fills only the fields the sequence uses, so cost tracks value.
• Separate sending domains with warmup, and reply handling that routes to the right person.
• A single write path into the CRM: one record, entered once, updated everywhere.

Stack

• Postgres as the system of record for prospect data.
• Python jobs for enrichment, scoring and list hygiene.
• CRM API for every write, scheduled jobs for sync.
• Sending infrastructure on its own domains, monitored for reputation.

What we measure

• Reply rate by sequence and by segment.
• Meetings booked per hundred contacts, traced to the source that produced them.
• Hours of manual research removed per week.
• Records with an owner and a next action, as a share of the whole list.

The problem

The site had accumulated pages over three years without an owner. Twenty-two URLs competed for the same three queries, the important pages were the slowest, and the ones that did rank pointed at a contact form nobody monitored. In assistants, the company was invisible.

What we built

• A page architecture organised around buyer questions, with overlapping URLs merged.
• Structured data and a machine-readable summary so assistants can parse and cite the site.
• A page template with a real next step on every page that earns traffic.
• A publishing rule set that stops the sprawl from coming back.

Stack

• Static build with a single generator, so structure is enforced rather than maintained.
• Structured data validated in the build, not after deploy.
• Search Console and analytics read together, one report.

What we measure

• Indexed pages that earn at least one click a month.
• Answer-engine citations per tracked query.
• Conversion rate on pages that receive organic traffic.
• Core Web Vitals on the pages that matter commercially.

The problem

The founder had a real point of view and no system for getting it out. Posts were written from scratch when there was time, which meant irregular output, no reuse, and a voice that changed depending on who was writing.

What we built

• A weekly capture session that produces raw material for everything else.
• A production pipeline with templates per channel, so a long-form idea becomes a short post, a carousel and a script.
• A documented voice guide, with a review step that catches drift before publishing.
• A scheduling and measurement layer that ties content to replies and qualified visits.

Stack

• Transcription and clip extraction from one recorded session.
• Asset templates so design is a fill-in, not a decision.
• Publishing queue and an analytics pull into the same reporting table as the pipeline.

What we measure

• Published pieces per week, against a committed cadence.
• Qualified visits and replies, not impressions.
• Inbound conversations attributable to content, per month.
• Time from capture to publish.