Why does an Adobe WebSDK migration usually take twice as long as it needs to?
Time to value dropped from roughly 1,200 hours to under 200
One Adobe SME replaced a typical six-person team
The web-side technical work that traditionally takes weeks
A healthcare subsidiary of a Fortune 500 company had a privacy incident, and their plan of action was the right one: move from Adobe Analytics to Customer Journey Analytics, adopt Healthcare Shield for PII andPHI governance, and once complete, turn their analytics back on.
In addition to the privacy incident, an initial migration had already moved their historical data into CJA, but it stopped short of a full enablement. There was no unified semantic schema, most of the Healthcare Shield capabilities sat unimplemented, and PII and PHI access controls existed on paper, but were never switched on. The client didn't need the old data preserved so much as they needed the whole thing done properly and built to support what was coming next: Real-Time CDP, Journey Optimizer, and the rest of Adobe Experience Platform.
None of this was optional. Every day without functioning PII and PHI controls extended the client's compliance exposure, and with analytics paused across every property, they had no visibility into how their own experience was performing. Getting it right mattered. Getting it done fast mattered just as much.
With a kickoff date of November, Moonbird's estimate for a complete and compliant implementation was three months. The client optimistically chose February as their earliest acceptable delivery window, but made clear they doubted it was realistic. They were still telling their own business internally it wouldn't land until May.
It landed in February, just as Moonbird promised.
The unified semantic schema came first, built from the historical data which had to be found across three properties, and where the same datapoint had been mapped three different ways. Reconciling that into one coherent structure took about twelve hours of client meetings, spread across six weeks and roughly twenty iterations. The schema generation itself moved quickly with AI assistance from Starling. The remainder of the three months went into the part that can’t be compressed; getting a client's team to understand and trust a new structure well enough to run it themselves.
This is where Moonbird's speed shows up beyond the headline numbers. Partway through, the client asked for an event-driven data layer outside the original project scope, the kind of addition that would normally mean a new round of estimating and a pushed-back timeline elsewhere. Here, the team folded it into the existing workstreams and generated the specs with Starling, and specs for all three properties came back correct on the first pass, no revisions needed.
The web-side technical work — traditionally the heaviest lift, with a technical resource manually rewriting every rule against the new taxonomy — took the Moonbird practitioner using Starling about thirty minutes across three properties, with no revisions needed.
Ivory, Moonbird's AI co-pilot for tag governance, assisted the team on the tech debt side: converting a custom function library into rules driven by the data layer and manageable through Adobe Tags rather than custom code.
CJA configuration across development, UAT, and production followed, with DULE labels applied to every field for PII, PHI, identifiability, and exportability, and the access controls and enforcement rules that had been left off the original implementation were finally turned on. Healthcare Shield, in other words, went from “purchased” to “fully activated.” Reporting was built out afterward, alongside stakeholder interviews with the business users who would end up running it daily.
The actual migration of the properties using Starling took about thirty minutes. Schema generation (that typically takes four to eight weeks and well over a hundred hours) took about a day. Coordination and approvals (that usually run around a hundred fifty hours) clocked in around fifty. Technical debt work (that often runs eighty to a hundred fifty hours) came in around twenty. The rollout itself, ordinarily weeks of sprints, took two days. Altogether, a project that would typically require six people and around twelve hundred hours compressed to one Adobe SME and under two hundred hours, close to an 85% reduction in time to value.
Moonbird delivered the event-driven data layer, and the client's team spent about two and a half months bringing it to life, a normal runway for that kind of work. For once, the vendor wasn't the bottleneck: Moonbird finished and waited on the client, not the other way around.
What's left now is a fully governed, fully enforced CJA implementation across three web properties (and a mobile app) with reporting live and a client team that understands the schema well enough to use it.They're positioned for Real-Time CDP and Journey Optimizer whenever they're ready to make the move, and they finished so far ahead of their own internal expectations that they've had to walk back a deadline they'd already announced…which may be a first.
