Metrics vs. Dimensions

In data analytics, a metric and a dimension are the two types of field that any report is built from, and mixing them up is one of the most common causes of a poorly designed dashboard. A metric is a quantitative value that can be summed, averaged or counted (monthly sales, number of orders, average order value, conversion rate). A dimension is a qualitative or categorical attribute that describes and classifies those numbers (channel, product, region, sales rep, date).

The practical rule for telling them apart: a metric answers how much or how many, while a dimension answers what, who, where or when. In a report they are read together, "sales (metric) by region (dimension)", and that combination is what turns an isolated data point into an answer. In most business intelligence tools, dimensions go in the rows or the chart axes and metrics go in the values.

Every analytics layer, like the one Metrix builds on top of a company's data, relies on this separation: it defines a catalog of consistent metrics and a set of dimensions to slice them by, so anyone can answer a business question without recalculating anything.

The confusion between metrics and dimensions seems technical, but it determines whether a dashboard is understood at a glance or whether nobody knows what they are looking at. Every well-formed business question combines the two: a number that matters and a context to view it through. Separating them clearly is the first step toward designing reports that answer questions instead of raising them.

How to tell one from the other

The fastest test is to ask which operation makes sense on the field. If you can sum, average or count it and the result means something, it is a metric: the sum of sales is total sales, and averaging order values gives you the average order value. If summing makes no sense but grouping or filtering by that field does, it is a dimension: summing province codes tells you nothing, but grouping sales by province does. Another guide: a metric is almost always a continuous number, while a dimension takes a limited set of values (a list of channels, months, categories).

The dimensional model underneath

This separation is not just a reporting convention: it is the foundation of the dimensional model used to organize data for analysis. In a star schema, metrics live in a fact table (each row is a measurable event, such as a sale) and dimensions live in dimension tables that surround that central table (products, customers, time). The fact table stores the numbers; the dimension tables store the attributes you slice them by. Thinking of a metric as a fact and a dimension as context helps you model correctly from the start.

Why it matters to a company in Latin America

For a midsize company in the region, the problem is rarely a lack of data: it is that every department calculates "sales" its own way and nobody agrees on how to slice them. Defining a single catalog of metrics (exactly what "net sales" is, how an "order" is counted) and a clear set of shared dimensions (the same channel, region and category names across every system) is what makes a dashboard trustworthy. Without that foundation, any KPI stays up for debate and meetings are spent fighting over the number instead of making the decision.

A concrete example

A distributor in Córdoba, Argentina, wants to understand its revenue. "Total sales" is a metric, but on its own it says little. When it breaks it down by the channel dimension, it discovers that 70% comes from WhatsApp; by month, it sees a seasonal drop in January; and by sales rep, it finds that two people account for half the volume. The same metric, viewed through three different dimensions, leads to three different decisions. That is the value of keeping both pieces cleanly separated and being able to combine them freely.

Edge cases and common mistakes

The most common mistake is treating a numeric identifier as a metric: a branch number or a zip code is a number, but summing them means nothing, so they work as dimensions. Some fields can also be one or the other depending on use: a customer's age is a metric if you calculate the average age, but a dimension if you group by "age bracket." Date is the most frequent case of a dimension that looks like a number: it is almost always used to group (by day, month or quarter), not to sum. And a calculated metric, such as conversion rate, cannot be averaged any which way: you have to rebuild it from its components (conversions over sessions), a detail that breaks many poorly built dashboards.