Can I calculate total number of purchases using Purchase Rate or Revenue/AOV?

Algolia does not expose an exact count of total purchases. Instead, we provide metrics like Purchase Rate, which is designed to show how many searches led to at least one purchase as opposed to the total number of items purchased.

You may notice that when analyzing purchase data shown in your Analytics dashboard, there are discrepancies between the number of purchases estimated using Purchase Rate vs those calculated from Revenue ÷ Average Order Value (AOV). 

This is expected and is due to how each metric is defined and calculated.

Below, you’ll find an explanation of what each metric represents, what each calculation results in, and why they might not align.

Why don't they match?

When estimating the number of purchases, the following formulas are often used:

• Purchase Rate × Tracked Searches estimates how many searches led to at least one purchase
• Revenue ÷ AOV estimates how many distinct transactions (orders) occurred

If you find there are more purchases than transactions, it typically means that different searches contributed to the same cart or transaction. 

What each metric measures:

Purchase Rate

• Reflects percentage of searches that resulted in at least one purchase
• Does not reflect total number of purchases, only how many searches led to at least one
• If you calculate purchases using Purchase Rate × Tracked Searches, you're estimating how many searches led to a purchase, not how many purchases happened
• If one search triggers multiple purchase conversion events, only one purchase conversion event is counted toward Purchase Rate - this is known as deduplication

Average Order Value

• AOV is calculated using Revenue ÷ Transaction Count
• Transaction count is the count after grouping by transaction_id (which is assigned by Algolia)
• One transaction can include multiple items and/or come from multiple searches
• If you calculate purchases using Revenue ÷ AOV, you're estimating how many orders/transactions were completed, which is a different concept from the total number of purchases

Example

The example below shows five different searches

• two searches with the same transaction_id meaning the items were purchased together in a single transaction
• a third search with it's own transaction_id (separate transaction)
• two searches resulting in no purchase

• Revenue: (2 × $8) + (1 × $10) + (3 × $5) = $41
• Purchase rate: 3/5 = 0.6% --> back-calc: 0.6 x 5 = 3 searches led to a purchase
• AOV: 41 ÷ 2 = $20.5 --> back-calc: 41/20.5 = 2 orders completed

Here, as two different searches ("leggings" and "skirt") contributed to the same transaction (ab6y8), the AOV-based estimation shows fewer orders than the Purchase Rate estimation (searches that led to purchases).

Summary

Each method measures something fundamentally different:

• Purchase Rate (search-based): How often do searches lead to a purchase?
• Revenue ÷ AOV (order-based): How many unique transactions were completed?

As of now, there is no direct metric exposed in Algolia Analytics for total number of purchases, so understanding how these estimates work is key to interpreting your data accurately. 

Discrepancies will arise naturally when:

• multiple searches contribute to the same cart
• a single search leads to multiple purchases