← All work← Back to quest log

Case study · On AGCompleted quest · On AG

Sell-out data implementation

I took over On's long-stalled effort to bring in retail sell-out data. Over two and a half years I oversaw the project that delivered a standardised data model, grew global coverage to nearly 70%, and launched it as a secure internal database in late 2025.

CompanyFaction
On AG
RoleClass
Lead, Commercial Analytics
LaunchedEra
Late 2025
ToolsGear
SQL · Looker · Python · Google Cloud

The problemThe call

A large share of On's business runs through wholesale partners. Without sell-out data, On only knew what it sold to retailers - not what was happening in the market: how much stock was on shelves, how products were selling against expectations, and where to act. Efforts to fix this had run for years with little progress. When I took it on, regions were siloed with different levels of access, and the data arrived in different formats, for different periods, with different metrics and granularity.

ApproachThe run

Step 1

Take stock

Consolidated everything that already existed - every data stream, its format, granularity and frequency - and validated it with colleagues and retail partners.

Step 2

One model for everyone

Helped shape the design of a central, standardised model anyone in the business could use - hiding the differences in granularity and frequency so people could get what they needed without knowing the quirks underneath.

Step 3

Grow coverage

Worked with sales teams, retail partners and third-party vendors to expand coverage in every region.

Step 4

Lock it down

Sell-out data is highly sensitive - volume data, and above all value data - so access runs on a permissions structure where people only see what they need.

ResultsRewards

Nearly 70%global sell-out coverage, up more than 10 percentage points
Up to 30%more coverage in individual regions
One modelreplacing siloed regional data, launched with role-based permissions in late 2025

Looking backDebrief

The project came with years of history, and much of the work was bringing order to what already existed - validating data streams, checking with colleagues and going back to retail partners. Moving it forward with so many stakeholders also took constant coordination, with frequent touchpoints and feedback sessions to keep everyone aligned. I hadn't run a project of this kind before, and it taught me a great deal about project management.

ContextLore notes

Partner names and sales figures are kept private; coverage figures are approximate.