Safi’s team were highly numerate, but they hadn’t yet built a map that defined and related key metrics in its business model.

Beginning in our preparatory Sprint 0, we created a domain model schematic of Safi’s business model to map out the universe of potential metrics and business interactions. This domain model became the central conceptual guide for data engineering, which our team used during the engagement and Safi’s team benefits from for future engineering work. The model serves as a blueprint that the team still uses when extending and enriching its data platform.

The Safi engineering team had already setup a data pipeline to push data into BigQuery, which worked perfectly well for running the app. However, the raw formatting and fragmented structure of the data made it difficult for business users to work with.

Armed with the domain model, we designed a set of data models that could transform the data into a more useable form while making minimal changes to their existing data stack with BigQuery and Google Cloud Platform.

The data modelling consisted of three distinct layers, which follow our tried-and-tested approach:

1. The staging layer, which received and cleaned the raw data;
2. The domain layer, which surfaced core data objects and implemented Safi’s business logic for necessary metrics; and
3. The presentation layer, which made data aggregations and remaining business logic so the data is most usable in the BI tool, Holistics.

The data model was the engine of Safi’s stack, and the 3-layer framework enabled us to build the most flexible, reliable, and scalable platform. Structuring the data in this form allowed us to simplify their infrastructure; for example, we were able to collapse 20+ source tables from their previous model into a single domain table in the new model, reducing the future maintenance load for the engineers without sacrificing analytical robustness.

To take the data through its ‘last mile’ and expose it to business users, Safi needed a cost-efficient visualisation tool that the entire team could use for self-service analytics. After reviewing their requirements, we helped them select Holistics.

The core goal of the engagement was to set the business up with working examples of best-in-class dashboarding. We created a preliminary set of dashboards for critical functions, training the team in both dashboard generation and extension. The initial batch of visualisations gave the team enough runway to spark interest in self-service analytics. Since the engagement, each function has created its own dashboards and used engineering resources to enrich the models we delivered.

Previously, Safi had been tracking their essential KPIs in a Google Sheet that posed several limitations. Leaders could not analyse data by different dimensions or view metrics at a more granular timescale than months, limiting their ability to understand performance drivers.

To kickstart the data transformation, we created the team’s first commercial dashboard in Holistics. The dashboard is consumed by the leadership team on a weekly basis and reports on foundational commercial health metrics, including key areas such as: GMV, revenue, customer segmentations, and unit economics. Not only does it provide greater analytical granularity, but also enabled the team to answer questions such as:

• Is revenue concentrated in a very small number of customers?
• Which materials have higher take rates and profit margins?
• Do we have any opportunity to grow revenue in new geographies?

With this dashboard in place as a ‘single pane’ view of the entire business, the broader organisation became familiar with Holistics and its self-service capabilities.