What is Semantic BI?

Semantic BI means that key metrics are defined once, in a “semantic layer”, and made available to everyone. This allows folks who aren’t data experts to self-serve without needing additional help or context. This is in contrast to most traditional BI tools, where data experts are required to create most new queries.

In this blog post, we'll explain why semantic BI isn't just a good option—it's the best one for forward-thinking companies aiming to democratize data access.

‍

Why semantic BI is better...

‍

• ‍Unmatched governance and standardization: The semantic layer enshrines metric definitions as code, which brings governance and standardization to metrics. Everyone is reading from the same page! Changes to metric definitions go through code reviews and when they’re live the new definition is automatically updated everywhere.
• Radical reduction in data requests: With semantic BI, the typical flood of data requests becomes a trickle. A surprisingly small number of fields goes a long ways: five metrics and ten dimensions can answer over 30,000 distinct questions! This frees up the data team for more complex and strategic data projects.
• Empowerment and insight for all: Empowering non-data staff to explore data independently doesn't just enhance their productivity—it transforms them into data advocates. The use of data by more employees revolutionizes the decision-making processes by keeping everyone data informed.
• Precision in data handling: Semantic BI enforces correct aggregations since queries are generated at run time for each combination of metrics and dimensions. Common mistakes—like taking an average of averages or rolling up daily uniques to weekly or monthly—are much less likely.

‍

‍

When is semantic BI the wrong choice?

‍

At Lightdash we believe semantic BI is a powerful tool, and for most companies it empowers people to make faster, more data-informed decisions. However, there are some situations where semantic BI is not the right choice:

• No data warehouse: semantic BI relies on a central database that contains all (or most) of the data you need for reporting. If your data is still siloed in tools like a CRM you’re probably not ready for semantic BI.
• No data pros: you’ll also need someone on the team to set up and maintain the semantic layer. That usually means at least one full-time data practitioner, or retaining a data consultant to fill the gap.
• No data culture: if your team is reluctant to engage directly with data to help inform decisions, semantic BI may not be the best choice.

‍

‍

Semantic BI vs. other approaches

‍

There are three main ways companies get data and insights to the people who need it. None of these approaches are inherently bad. The right approach depends on the stage of your company, the people and teams you have, your company culture, and the state of your data infrastructure.

‍

1. Data free-for-all

‍

‍

• Approach: Everyone in the company is given access to raw datasets. This could mean direct database access, a drag-and-drop query builder, or even the reports section of tools like Salesforce, Zendesk, or HubSpot.
• Pros: Data free-for-all works well for smaller companies where everyone understand the core datasets, or companies where data is very simple.
• Cons: Since metrics across different reports rarely match exactly, people often mistrust the data or get bogged down trying to make numbers match. This approach leads to more incorrect insights since everyone has to learn data intricacies through trial and error. Losing trust in data means more people abandon it altogether and make decisions based on their gut.

‍

2. Traditional BI

‍

‍

• Approach: Once a company hires a data team, it’s common for them to handle all data requests directly. If someone outside the data team needs data to make a decision, they submit a request and the data team builds something in the BI tool.
• Pros: If the data team is all on the same page, this approach results in the fewest mistakes.
• Cons: Data requests tend to grow exponentially. Each answer to a data question results in two new questions. The data team spends all their time on handling inbound requests and trying to get on the same page about how to define key metrics.

‍

3. Semantic BI—the Lightdash way

‍

‍

• Approach: Give people meaningful building blocks to answer their own data questions. If done correctly, most common questions can be answered without the data team, and people can drill into results or pivot/regroup data themselves.
• Pros: Simple data questions can be answered without help from the data team. This means the data team has more time to dig into complex analyses, and business users get the results they need to make decisions much faster.
• Cons: A portion of the data team’s time needs to be spent on maintaining clean data models and the semantic layer, but the efficiency and consistency gains should leave them with time to work on more complex data projects as well.

‍

‍

How business intelligence evolves as a company grows

‍

‍

Let’s use a fictional product, Flubber, to illustrate the evolution of BI as a company grows.

Initially, Flubber sells through Shopify and relies on its built-in reporting. As the team is small, a Data free-for-all approach works well, with easy access to reports and daily standup meetings to resolve inconsistencies.

As Flubber expands to retail stores, channel partners, and launches a mobile app, data becomes more complex and siloed. A data team is hired to build a data warehouse, enabling a Traditional BI approach. Initially, this works well, with the data team balancing dashboard creation and proactive analysis. However, ad hoc data requests soon overwhelm the team and they don’t have time for high impact data projects. The data team also becomes a bottleneck for making business decisions.

To address this, Flubber adopts Semantic BI. This approach standardizes key metrics and empowers employees to self-serve, reducing ad hoc requests and allowing the data team to focus on strategic projects.

‍

Conclusion

‍

Semantic BI represents a shift in how organizations approach data accessibility and governance. By defining key metrics once and making them universally available, semantic BI empowers everyone to self-serve, reducing reliance on the data team for simple queries. This approach also leads to better governance and standardization of metrics since each key metric is defined once in code. In contrast to traditional BI methods, which often bottleneck data requests through a centralized team, semantic BI encourages a more decentralized approach to data. Ultimately, semantic BI's emphasis on self-service and metric definitions not only supports data-driven decision-making but frees up data teams to focus on higher-value projects, driving innovation and efficiency for the company.

If you’d like to learn more about semantic BI, check out this article about BI for the semantic layer, by Oliver, Lightdash CTO.