The situation
A high-value data consultancy competing with large advisory firms for predictive analytics engagements across finance, retail, and CPG. The work was commercially attractive and structurally difficult to scale.
Projects were heavily services-led. Contracts were under pressure from master-service-agreement and discounting dynamics. There was no dedicated data engineering capability, which meant pipelines were rebuilt close to from scratch for every new client.
What the real problem was
Not delivery efficiency. A business-model transition. The leadership team wanted a SaaS product at the end of it, and the honest sequencing question was what had to become repeatable before a product was even possible.
What I changed
- Refocused engineering on reusable pipelines rather than one-off delivery — the single change with the largest downstream effect.
- Narrowed the market to CPG and retail, where data structures and use cases repeat, and deliberately exited less repeatable finance work.
- Introduced Continuum, a modular predictive analytics framework and demand-planning product vision that gave the reusable components a commercial shape.
- Defined a customer portal for analytics, monitoring, and model drift detection — creating a technical basis for ARR-style contracts rather than a pricing fiction.
- Used drift detection and adjacent use cases to open expansion revenue inside existing accounts.
- Introduced Jira, agile rituals, and operating discipline across delivery, and strengthened leadership execution across funnel management, contracts, marketing, and PR.
What happened
Revenue grew by double digits while exiting a category of work — the combination that matters, because growing by adding scope is not the same as growing by getting better. New client implementation fell from months to weeks. The business ended with a reusable delivery model, a narrower and more defensible market focus, and a credible route from project revenue to recurring contracts.
Why this one matters
Services-to-product is one of the hardest transitions in software and one of the most commonly attempted badly. The failure mode is naming a product before the delivery model can support one. This is what the sequencing looks like when it works.