One Platform

The problem

Snagajob carried more than twenty years of innovation, market shifts and acquisitions. That left us with competing products, and each one got the same new features every year. Keeping track of how they differed was hard, and it slowed every part of the company. We needed a better path.

How I thought about it

Consolidation was the answer. The question was how. The last attempt had been to build a new platform. When I mapped out that path, I saw it would end with two platforms, not one. That is the XKCD “standards” comic in real life.

That split was also the biggest cost. Running two platforms side by side would have dragged on innovation for the next three years. That is three years of lost time. So I went the other way: build on the oldest platform, not the newest, and move customers onto it as it caught up.

What I did

Data first. The first step was to reconcile the data, so the company had one language for its customers no matter which product they used. That took months of mapping every product’s data to the CRM and ERP.

The first win. With the data in one place, we could switch the business model from subscriptions to pay-for-performance. We built the new system fast, and any of our products could use it. The previous pay-for-performance attempt had been locked to a single product.

Assimilation. With pay-for-performance working, we kept building features into the oldest platform. When it matched a product’s features, we moved that product’s customers over. By October 2023, every product was consolidated onto it.

Results

Revenue under the new model reached 2× in year one and 4× in year two.

• Engineering. New feature development went down by as much as 83%. We retired older technology (PHP, AWS EC2, Silverlight, .NET Framework, SQL Server and more), which saved money, cut operational overhead and lowered our exposure to risk.
• Sales. A simpler sales cycle, and an easier time finding the right fit for each customer. New sales staff onboarded faster: they no longer had to learn six separate products that did much the same thing.
• Support. Shorter average handle time and a smaller ticket backlog. Nobody had to work out first which product, or products, a customer needed help with.
• Finance. Simpler month-end close and forecasting, with less variation to account for.
• The whole company. Reports agreed with each other, because Kafka became the source of truth and the product and reporting read from the same place. Conversations got simpler too. “Customer”, “applicant” and “location” each had one definition and one source of data.

What I’d do today with AI

The slow part was months of mapping data across products, and that is work I would now hand to agents: one goal per product, a plan for each, and a verify step that checks every mapped field against the CRM before anything is called done. I would still make the call on which platform to build on myself. What would change is how fast I could test it, and every surprise from the mapping would go into memory, so the next product’s plan would start from it.