Numeric

Numeric metrics are calculated per user, meaning each individual user's events are calculated and added to a single data point, ensuring equal weighting.

They are aggregated using the eventValue or custom property that was submitted via the track function.

When creating a numeric metric, we must choose how to aggregate the data:

• Count

• Sum

• Average

		Example
Count of events per user	Count the number of times an event was triggered by users Shows the average count. This value will increase as users return to your site, while new users will bring down the average.	The average number of video plays per user. The average number of help requests per user. The average number of purchases per user.
Sum of event values per user	Adds up the event values for events triggered by users. Shows the average sum. This value will increase as users return to your site, while new users will bring down the average.	The average total time each user spent watching videos on your site. The average total number of pages each user visited on your site. The average total amount spent by each user over the course of the experiment.
Average of event values per user	The average of event values for events triggered by users. Shows the average of averages. The	The average time spent by users on your site during each session. The average number of items added to the cart by users. The average value of purchases when users place an order.

---

Example: average / sum purchase price

When creating a metric, we've decided to select the purchase-event and use the price value that is sent with the event.

Node / JavaScript

const item: Product = { 
  productId: '0x123', 
  price: 49.99,
  discount: 0.20,
  ...
}

client.track('purchase_event', null, item);

Depending on the method chose, we could aggregate the average purchase price over time, or sum of purchase values over the course of the experiment.

Example: count of discount values

Suppose we're running an experiment with different discount values. We don't yet know which discount users will response to most, so let's create a metric that uses count.

Similar to the above example, we'll instead choose discount as the value field on the metric.

Node / JavaScript

const item: Product = { 
  productId: '0x123', 
  price: 49.99,
  discount: 0.20,
  ...
}

client.track('purchase_event', null, item);

Now, we'll choose count as the aggregation type, which will count the average number times a particular discount value was used.

To ensure fair representation across all users in the experiment and prevent skewed results, the aggregation method used averages the data, rather than allowing a single user's data to carry more weight than others.