Metrics

A metric is a value that describes or summarizes features from a collection of data points. For example: count of total number of user IDs, or sum of revenue.

Metrics are used to summarize dimensions or, sometimes, other metrics.

​

Adding metrics to your model

name: my_model
columns:
  - name: user_id # dimension name of your metric
    metrics:
      num_unique_user_ids: # name of your metric
        type: count_distinct # metric type
      num_user_ids:
        type: count

Once you’ve got the hang of what these metrics look like, read more about the metric types you can use below.

​

Metric Categories

Each metric type falls into one of these categories. The metric categories tell you whether the metric type is an aggregation and what type of fields the metric can reference:

​

Aggregate metrics

Aggregate metric types perform (surprise, surprise) aggregations. Sums and averages are examples of aggregate metrics: they are measurements summarizing a collection of data points.

Aggregate metrics can only reference dimensions, not other metrics.

​

Non-aggregate metrics

Non-aggregate metrics are metric types that, you guessed it, do not perform aggregations.

Numbers and booleans are examples of non-aggregate metrics. These metric types perform a calculation on a single data point, so they can only reference aggregate metrics. They cannot reference dimensions.

​

Metric configuration

name: sales_stats

joins:
  - join: web_sessions
    sql_on: ${web_sessions.date} = ${sales_stats.date}

columns:
  - name: revenue
    description: 'Total estimated revenue in GBP based on forecasting done by the finance team.'
    metrics:
      total_revenue:
        label: 'Total revenue GBP'
        type: SUM
        description: 'Total revenue in GBP'
        sql: 'IF(${revenue} IS NULL, 10, ${revenue})'
        groups: ['product_details', 'item_details'] # this would add the metric to a nested group: `product details` --> `item details`
        hidden: false
        format: '[$£]#,##0.00' # GBP rounded to two decimal points
        show_underlying_values:
          - revenue
          - forecast_date
          - web_sessions.session_id # field from joined table
        filters:
          - is_adjusted: true

Here are all of the properties you can customize:

​

Metric types

​

percentile

Takes the percentile of the values in the given field. Like SQL’s PERCENTILE_CONT function.

The percentile metric can be used on any numeric dimension or, for custom SQL, any valid SQL expression that gives a numeric table column.

For example, this creates a metric median_price by taking the 50% percentile of the item_price dimension:

columns:
  - name: item_price
    metrics:
      median_price:
        type: percentile
        percentile: 50

​

median

Takes the 50th percentile of the values in the given field. Like SQL’s PERCENTILE_CONT(0.5) function.

The median metric can be used on any numeric dimension or, for custom SQL, any valid SQL expression that gives a numeric table column.

For example, this creates a metric median_price by taking the 50% percentile of the item_price dimension:

columns:
  - name: item_price
    metrics:
      median_price:
        type: median

​

average

Takes the average (mean) of the values in the given field. Like SQL’s AVG function.

The average metric can be used on any numeric dimension or, for custom SQL, any valid SQL expression that gives a numeric table column.

For example, this creates a metric avg_price by taking the average of the item_price dimension:

columns:
  - name: item_price
    metrics:
      avg_price:
        type: average

​

boolean

Tells you whether something is True or False.

The boolean metric can be used on any valid SQL expression that gives you a TRUE or FALSE value. It can only be used on aggregations, which means either aggregate metrics or custom SQL that references other metrics. You cannot build a boolean metric by referencing other unaggregated dimensions from your model.

boolean metrics don’t do any aggregations; they just reference other aggregations.

For example, the avg_price metric below is an average of all of the item_price values in our product table. A second metric called is_avg_price_above_20 is a boolean type metric. The is_avg_price_above_20 metric has a custom SQL expression that tells us whether the avg_price value is greater than 20.

columns:
  - name: item_price
    metrics:
      avg_price:
        type: average
      is_avg_price_above_20:
        type: boolean
        sql: 'IF(${avg_price} > 20, TRUE, FALSE)'

​

count

Does a table count, like SQL’s COUNT function.

The count metric can be used on any dimension or, for custom SQL, any valid SQL expression that gives a set of values.

For example, this creates a metric number_of_users by counting the number of user_id values in the table:

columns:
  - name: user_id
    metrics:
      number_of_users:
        type: count

​

count_distinct

Counts the number of distinct values in a given field. It’s like SQL’s COUNT DISTINCT function.

The count_distinct metric can be used on any dimension or, for custom SQL, any valid SQL expression that gives a set of values.

For example, this creates a metric number_of_unique_users by counting the number of unique user_id values in the table:

columns:
  - name: user_id
    metrics:
      number_of_unique_users:
        type: count_distinct

​

date

Gives you a date value from an expression.

The date metric can be used on any valid SQL expression that gives you a date value. It can only be used on aggregations, which means either aggregate metrics or custom SQL that references other metrics. You cannot build a date metric by referencing other unaggregated dimensions from your model.

Creating a max or min date metric with type: date

If you want to create a metric of a maximum or minimum date, you can’t use type: max or of type: min metrics because these are only compatible with numeric type fields. Instead, you can calculate a maximum or minimum date by defining a metric of type: date and using some custom sql, like this:

- name: created_at_date
    dimension:
      type: date
    metrics:
      max_created_at_date:
        type: date
        sql: MAX(${TABLE}.created_at_date)

​

max

Max gives you the largest value in a given numeric field. It’s like SQL’s MAX function.

The max metric can be used on any numeric dimension or, for custom SQL, any valid SQL expression that gives a numeric value.

Because type: max metrics only work with numerical fields, you can’t use them to find a maximum date. Instead, you can use the MAX() function in the sql parameter of a metric of type: date to get a maximum date (you can see an example of this in the date section.

For example, this creates a metric max_delivery_cost by looking at the delivery_cost dimension and taking the largest value it finds:

columns:
  - name: delivery_cost
    metrics:
      max_delivery_cost:
          type: max

​

min

Min gives you the smallest value in a given numeric field. It’s like SQL’s MIN function.

The min metric can be used on any numeric dimension or, for custom SQL, any valid SQL expression that gives a numeric value.

Because type: min metrics only work with numerical fields, you can’t use them to find a minimum date. Instead, you can use the MIN() function in the sql parameter of a metric of type: date to get a minimum date (you can see an example of this in the date section.

For example, this creates a metric min_delivery_cost by looking at the delivery_cost dimension and taking the smallest value it finds:

columns:
  - name: delivery_cost
    metrics:
      min_delivery_cost:
        type: min

​

number

Used with numbers or integers. A number metric doesn’t perform any aggregation but can be used to perform simple transformations on other metrics.

The number metric can be used on any valid SQL expression that gives you a numeric or integer value. It can only be used on aggregations, which means either aggregate metrics or custom SQL that references other metrics. You cannot build a number metric by referencing other unaggregated dimensions from your model.

For example, this creates a metric called total_gross_profit_margin_percentage based on the total_sale_price and total_gross_profit_margin aggregate metrics:

columns:
  - name: sale_price
    metrics:
      total_sale_price:
        type: sum
  - name: gross_profit_margin
    metrics:
      total_gross_profit_margin:
        type: sum
        total_gross_profit_margin_percentage:
           type: number
          sql: '(${total_gross_profit_margin}/ NULLIF(${total_sale_price},0))'

The example above also uses the NULLIF() SQL function to avoid division-by-zero errors.

​

sum

Adds up the values in a given field. Like SQL’s SUM function.

The sum metric can be used on any numeric dimension or, for custom SQL, any valid SQL expression that gives a numeric table column.

For example, this creates a metric total_revenue by adding up the values in the revenue dimension:

columns:
  - name: revenue
    metrics:
      total_revenue:
        type: sum

​

string

Used with fields that include letters or special characters.

The string metric can be used on any valid SQL expression that gives you a string value. It can only be used on aggregations, which means either aggregate metrics or custom SQL that references other metrics. You cannot build a string metric by referencing other unaggregated dimensions from your model.

string metrics are rarely used because most SQL aggregate functions don’t return strings. One common exception is MySQL’s GROUP_CONCAT function.

For example, this creates a metric product_name_group by combining the unique values of a dimension called product_name:

columns:
  - name: product_name
    metrics:
      product_name_group:
        type: string
          sql: 'GROUP_CONCAT(${TABLE}.product_name)'

​

Description

We add default descriptions to all of the metrics you include in your model. But, you can override these using the description parameter when you define your metric.

metrics:
  num_user_ids:
    type: count
    description: 'Total number of user IDs. NOTE: this is NOT counting unique user IDs'

​

Format

You can use the format parameter to have your metrics show in a particular format. Turntable supports spreadsheet-style format expressions for all metric types.

To help you build your format expression, we recommend using https://customformats.com/.

metrics:
  total_us_revenue:
    type: sum
    description: 'Total revenue in USD, with two decimal places, compacted to thousands'
    format: '$#,##0.00," K"' # 505,430 will appear as '$505.43 K'
  percent_of_total_global_revenue:
    type: number
    description: 'Percent of total global revenue coming from US revenue.'
    sql: ${total_us_revenue} / ${total_global_revenue}
    format: '0.00%' # 0.67895243 will appear as '67.89%

​

Example format expressions

​

Custom SQL in aggregate metrics

You can include custom SQL in your metric definition to build more advanced metrics using the sql parameter. Inside the sql parameter, you can reference any other dimension from the given model and any joined models. You can’t reference other metrics.

You can reference dimensions from the same model like this: sql: "${dimension_in_this_model}"

Or from joined models like this: sql: "${other_model.dimension_in_other_model}"

metrics:
  num_unique_7d_web_active_user_ids:
    type: count_distinct # metric type
    sql: 'IF(${is_7d_web_active}, ${user_id}, NULL)'
  num_unique_paid_user_ids:
    type: count_distinct
    sql: 'IF(${subscriptions.is_active}, ${user_id}, NULL)'

​

Using custom SQL in non-aggregate metrics

In non-aggregate metrics, you can reference any other metric from the given model and any joined models. You can’t reference other dimensions.

You can reference metrics from the same model like this: sql: "${metric_in_this_model}"Or from joined models like this: sql: "${other_model.metric_in_other_model}"

metrics:
  num_unique_users:
    type: count_distinct
  is_num_unique_users_above_100:
    type: boolean
    sql: 'IF(${num_unique_users} > 100, TRUE, FALSE)'
  percentage_user_growth_daily:
    type: number
    sql: '(${num_unique_users} - ${growth_model.num_unique_users_lag_1d}) / NULLIF(${growth_model.num_unique_users_lag_1d}, 0)'

​

Show underlying values

By default, we show all of the dimensions from the Table when you click View underlying data. If you have fields from a joined table included in your results table, then we’ll also show you all of the fields from the joined Table.

You can limit which dimensions are shown for a field when a user clicks View underlying data by adding the list of dimensions to your .yml files:

name: sales_stats
joins:
  - join: web_sessions
    sql_on: ${web_sessions.date} = ${sales_stats.date}

columns:
  - name: user_id
    description: 'Unique ID for users.'
    meta:
      dimension:
        type: string
      metrics:
        count_users:
          type: count_distinct
          show_underlying_values:
            - revenue_gbp_total_est
            - actual_date
            - web_sessions.session_id # field from joined table

The list of fields must be made of dimension names (no metrics) from the base table or from any joined tables. To reference a field from a joined table, you just need to prefix the dimension name with the joined table name, like this: my_joined_table_name.my_dimension.

The order that the fields are listed in show_underlying_values is the order that they’ll appear in on the view underlying data table.

​

Groups

You can group your dimensions and metrics in the sidebar using the groups parameter.

To do this, you need to set up group_details in the model’s configuration. Then, you can use these groups to organize metrics and dimensions. You can create nested groups up to 2 levels.

name: baskets
columns:
  - name: basket_item_id
    description: 'ID for the product item within the basket.'
    dimension:
      groups: ['product_details', 'item_details'] # this would add the dimension to a nested group: `product details` --> `item details`
    metrics:
      count_total_basket_items:
        type: count_distinct
        groups: ['product_details', 'item_details'] # this would add the metric to a nested group: `product details` --> `item details`
  - name: product_name
    description: 'Full name of the product.'
    dimension:
      label: 'Product name'
      groups: ['product_details'] # this would add the dimension under the group label: `product_details`
    metrics:
      count_total_product_types:
        type: count_distinct
        groups: ['product_details'] # this would add the metric under the group label: `product_details`

​

Filters

Filters are applied to metrics any time that metric is used in Turntable. Filters can only be used with aggregate metric types.

For example, we could add a filter to our users count to make sure it didn’t include user IDs with closed accounts, like this:

name: sales_stats
columns:
  - name: user_id
    description: 'Unique ID for users.'
    dimension:
      type: string
    metrics:
      count_users:
        type: count_distinct
        filters:
          - is_closed_account: false

These filters do not appear in the Filters tab in the Explore view, instead, they are applied automatically in the SQL query that fetches your results. That means filters added using the filter parameter can’t be removed in the UI and won’t be visible to users unless they look at the SQL query.

​

Available filter types

​

Special characters in filters

To use special characters such as %!_> in your filter value you can either put the value in quotes, or escape special characters with a backslash \. For example, if you wanted to filter for subscription status of is_subscribed you can write the metric in one of these ways:

filters:
  - subscription_status: 'is_subscribed'

filters:
  - subscription_status: is\_subscribed

​

Filtering using a list of values

To filter a field using a list of values you can supply them as an array for that field. For example, if you wanted to filter for orders with order status completed or shipped you should write the metric like:

columns:
  - name: order_id
    metrics:
      completed_or_shipped_order_count:
        type: count_distinct
          filters:
            - order_status:
                - completed
                - shipped

​

Filters are joined using AND

For example:

filters:
  - is_closed_account: false
  - is_7d_active: true

Would give you logic like is_closed_account = TRUE AND is_7d_active = FALSE.

​

Adding filters from joined models

To filter using a field from a joined model, just use the syntax model_name.field, like this:

name: sales_stats
joins:
  - join: web_sessions
    sql_on: ${web_sessions.date} = ${sales_stats.date}
columns:
  - name: user_id
    description: 'Unique ID for users.'
    dimension:
      type: string
    metrics:
      count_users:
        type: count_distinct
        filters:
          - is_closed_account: false
          - web_sessions.is_bot_user: false

​

Metric filters cannot be used with non-aggregate metrics

You can’t use filters with non-aggregate metric types. Instead, if your non-aggregate metrics are referencing aggregate metric types, you need to apply metric filters to the aggregate metrics.

Here’s an example: imagine you wanted to calculate the average cost per item that had the status shipped. You would need to do something like this in your .yml:

name: orders
metrics:
  average_cost_per_item_shipped:
    type: number
    sql: ${total_cost_of_shipped} / ${count_unique_items_shipped}
columns:
  - name: item_id
    description: 'Unique ID for items ordered.'
    dimension:
      type: string
    metrics:
        count_unique_items:
          type: count_distinct
        count_unique_items_shipped:
          type: count_distinct
          filters:
            - status: 'shipped'
  - name: item_cost
    description: 'Cost for each item ordered.'
    meta:
      dimension:
        type: number
      metrics:
        total_cost:
          type: sum
        total_cost_of_shipped:
          type: sum
          filters:
            - status: 'shipped'