Who they are
Stella & Dot is a San Francisco social-selling company — jewelry and accessories sold through a field of independent stylists — that expanded into new business lines as it grew, adding Keep Collective and EVER Skincare.
It is also where the team's direct sales record begins: the first platform they built inside the channel, on the road that runs through Norwex and Juice Plus+ to EMMA.
The challenge
As Stella & Dot expanded, its e-commerce and commissions systems struggled to keep pace. The commerce platform was Magento, built by a third-party agency; commissions ran on the Thatcher engine. A development team had been established in San Francisco and Seattle, but it was spending most of its time fixing bugs from the original agency implementation and fighting stability issues — little of its capacity ever reached new work.
How serious that had become showed in the new CIO’s first week — which happened to be Black Friday week. The platform went down repeatedly, for hours at a time, and could not even be restarted: the moment it came back up, the waiting traffic crashed it again. The team wrote emergency code live, siphoning traffic away so that a volume the servers could actually survive was allowed through.
That week settled the diagnosis. The company had grown past its platform — and this was not a capacity problem more hardware could fix, because the architecture no longer scaled uniformly. Promotions and launches would keep taking it down, marketing was already delayed by platform inflexibility, and the differentiated brand experiences planned for Keep Collective and EVER Skincare were impossible on it.
To scale further and launch new verticals, Stella & Dot needed a modern, resilient and extensible platform — one that could power multi-brand operations and support rapid iteration without compromising reliability.
What we did
Stella & Dot's transformation was led from the inside. Neil Markey, now Managing Director of AHM Technology, served as CIO; Craig Harris was VP Engineering, owning the back office and the commissions platform. Several more of today's AHM Technology team were part of it as well — including its Head of Product, its Chief Architect and a number of its engineers.
It was this team that took the case to the executive staff and the board: build a new, scalable platform designed not only for Stella & Dot’s volume but for the brands the company planned to expand into. The approach that followed combined platform stabilization first, team building, and a parallel rebuild of the company’s core technology infrastructure.
Platform Stabilization and Continuity
Before anything new could be built, the existing platform had to stop failing. A dedicated team fixed the most critical scalability issues and put safety controls in place — so an outage like that first Black Friday week could never happen again. The stabilized stack then ran for over 18 months, buying time for a clean-slate rebuild while reducing incidents and preserving business continuity.
Eighteen months of stability bought for the rebuild
Internal Engineering Org and Modern Architecture
The team expanded and retrained the existing engineering organization in San Francisco and Seattle, and opened a new development center in Barcelona, Spain — while leading the design of a modern microservices based architecture.
Key components included:
The architecture prioritized modularity, rapid deployment, and scalability.
- Headless commerce engine
- Admin and back-office systems
- Custom commissions and genealogy engine
- Mobile-first stylist tools
An expanded, retrained organization across three sites — and an architecture built for more than one brand
From Thatcher to an In-House Commissions Engine
The Thatcher commission engine was designed to run every fifteen minutes — and toward month end, when volume peaked, those runs consistently failed. Stylists could go hours without their earnings and reports updating, at exactly the moment they were watching them most closely. Fixes sat with the vendor, and the vendor did not deliver them.
So the team built its own commissions platform: one designed to scale, and to run three brands with three different fields — with sponsorship allowed across those fields, a real complication of the compensation plan that the engine had to handle natively. The new platform processed the company’s entire order volume — even a full month-end close — in around two minutes.
Month-end commissions in minutes, on an engine the company owned
Parallel Build and Brand Launch Strategy
The team executed a dual track plan, launching new brands on the modern platform before migrating the core business.
This phased rollout allowed for real world feedback, system optimization, and validation of multi-brand capabilities.
- Keep Collective launched on the new stack within 7 months
- EVER Skincare followed 4 months later
Two brands live on the new stack inside eleven months
Core Migration Planning
With Keep Collective live and the new stack proven, planning turned to the core business. A comprehensive gap analysis between the legacy and new systems produced an eleven-month roadmap aligned across engineering, product and business stakeholders — anchored to a launch date picked deliberately to land just before the company’s annual conference.
The team hit that date exactly. The flagship launched on the day chosen eleven months earlier — with zero issues, and completely transparently to the field.
Launched on the exact date picked eleven months earlier
Flawless Core Business Migration
A big-bang migration was the only honest option: a brand-new platform meant moving every piece of history — customers, stylists, orders, commissions — and at that volume the full migration run took several weeks. So it ran ahead of cutover, in complete passes, while the old platform kept transacting.
Months of work went into the cutover itself: detailed migration and cutover plans, rehearsals, fail-safes, and rollback plans for every step — so that when the night came it was execution, not improvisation. The platform came down, in-flight activity was drained, and a final pass migrated everything that had changed since the last complete run. Eight hours later the new platform was live — on schedule.
What mattered most was what happened next: nothing. Stylists and customers came back to faster performance and a modern responsive UI — and not a single customer, stylist or data-related issue was raised after go-live.
Months of cutover planning, fail-safes and rollback plans — and the rollbacks were never needed
The commissions runs that used to fail at month end now processed the entire order volume in about two minutes.