In order to extract insights from data collaboration, data modelling is a crucial process that decides how data will be stored, accessed, shared, updated, and leveraged. A data model is generally categorized in three ways: conceptual, logical, and physical.

Conceptual data models: The high-level view

A conceptual data model helps convey high-level business concepts and requirements, identifying key entities and relationships without technical details. Its goal is to establish a shared understanding of the organization's data needs among stakeholders and lay the foundation for more detailed data models.

This conceptual model is already established when working with Amazon Ads and defines some of the most common structures and language we use today—like the relationship between entities and advertisers or campaigns (orders) and line items (ad groups).

Logical data model: Structuring for insights

A logical data model translates these high-level concepts into a more detailed manner, translating concepts from the conceptual model into a structured format. This includes defining entities, attributes, relationships, and constraints in a technology-independent manner, aiming to create a comprehensive and normalized data view.

At Gigi, our data modelling occurs at a logical data model level. By standardizing how data is ingested and organized, we can ensure data consistency and quality. This allows marketers to extract insights from their Amazon Ads campaigns and collaborate with other first-party and third-party datasets. Without a data model, this would not be possible.

Physical data model: The database blueprint

A physical data model represents data in a technology-specific structure tailored to a database management system (DBMS). This involves defining physical storage structures, indexing mechanisms, partitioning schemes, and other optimization techniques to meet performance and scalability requirements. While this data model type is irrelevant for us at Gigi (as we don’t use this type of data model) to extract insights from your Amazon Ads campaigns, it's worth noting.

Data Modelling at Gigi

As mentioned, we use logical data modelling at Gigi to organize data so marketers can extract insights from their campaigns. This helps simplify and abstract away some of the friction of building Streaming TV campaigns within the Amazon DSP. This includes creating proprietary Amazon Marketing Cloud (AMC) SQL queries, connecting and measuring insights from off-Amazon channels into your campaigns, and more.

Original datasets from Amazon, first-party channels (like Shopify), and third-party datasets are all processed through Gigi's data pipelines and automatically converted into the Gigi data model. This allows us to structure data across multiple sources in a standardized manner and ensure insights are extracted without discrepancies or inconsistencies.

An (over-simplified) example of how we convert data through the Gigi data model would be:

• We have one customer data table and sub-level tables of billing and shipping addresses.

  

• We create a 'one to many' association between the customer and billing/shipping information tables.

• This results in an order table with each row in 1 order, then a sub-level table that captures the line items in each order.

  

• Then, we would create an association (one to many) between order and line item tables.

With the above data converted into the Gigi data model, marketers can extract customer and purchase insights, like creating a seed audience of top-spending customers across Amazon and Shopify with the highest lifetime value, and use this for lookalike audience building.

Why data modelling at Gigi is important

Data quality and consistency

When dealing with datasets across multiple sources, data modelling allows you to standardize how the data is organized and helps alleviate some data consistency issues that might arise with subjective classifications (like what qualifies as a "new-to-brand" user). To avoid data quality and consistency issues, we enforce our Gigi data model during the data transformation process to ensure we do not import, ingest, or transform any incorrect or unexpected data. Additionally, we monitor data quality daily to ensure data from the source database matches the data in the destination database.

Complexity and usability

Complicated data models can confuse dashboards and reports, especially when they aren't intuitive for end users. At Gigi, we emphasize user-friendly interfaces that allow marketers to customize their reports and extract insights. Our data model is designed to be straightforward and is comprised of 6 core tables that cover the shopper journey, capturing customer demographic, browsing, and shopping/purchasing behaviour.

Integration capabilities

Data collaboration is vital to accessing omnichannel insights for measurement and audience building. At Gigi, our data model is designed to support interoperability, seamlessly integrating with first-party and third-party data sources, making data collaboration easy.

At its core, data modelling enables businesses to transform complex, multidimensional datasets into actionable insights. They can then create a robust framework for analytics, reporting, and operational decision-making. At Gigi, we use data modelling as the backbone of our data solutions, ensuring marketers can extract meaningful, reliable insights from their advertising data.