Skip to content
Admetriva

Product attribution methodology

How Admetriva splits each product sale between channels, with the Weighted Multi-Touch v1 model, the attribution window and worked examples.

On this page

Admetriva attribution answers the question: which channel did a product sale come from? It uses only the data measured by the tracker on your site and a deterministic, transparent model, Admetriva Weighted Multi-Touch v1. It is not a “data-driven” model and it does not use AI.

The channels

Google, Meta, TikTok, Organic, Referral, Direct and Unknown. A new click from a channel opens a segment, and the events that follow (views, cart, checkout) are attached to that segment until the next external click.

TikTok appears here only as the source of a click that reached your site. Importing data from TikTok Ads is not available yet.

The attribution window

The window is 3, 7 or 30 days, 7 by default, and you choose it in Tracking settings. Each channel has its own timer:

  • only a new click from the channel renews the timer; views, the cart, a direct return or a page reload do not extend it;
  • reloading a page that still has gclid in its address is not a new click;
  • a Meta click does not erase Google: Google stays eligible until its own window expires;
  • a channel is eligible if its last click was no more than one window before the order;
  • an expired channel gets nothing;
  • with no eligible channel, the sale becomes “Direct” or “Unknown”. The source is never made up.

The session (30 minutes of inactivity) and the attribution window are different things.

The weights

Interaction in the channel's segment Score
eligible click 1.00
another product viewed (up to 5 per segment) 0.20
the purchased product, viewed 1.00
the product added to cart 4.00
checkout started with the product in the cart 2.00

Each score is multiplied by an age factor: 0.5^(elapsed time / half the window). At the edge of the window the factor is 0.25, never zero for a click that is still eligible.

channel credit = channel score ÷ sum of eligible channel scores

What gets split

A product line in the order is one product conversion, whatever the quantity. The credits of a line add up to exactly 1, and the attributed revenue and quantity add up to exactly the line's revenue and quantity. The rounding remainder goes to the channel with the highest score.

Examples

A single channel. A Google click 2 days ago, a view and an add to cart, an order of RON 499.98 for 2 units. Google gets credit 1, 2 units and RON 499.98.

Two channels, 7-day window, an order of RON 100, ignoring age for clarity. Google: click + view = 2. Meta: click + view + cart = 6. Meta gets 6/8 = 0.75 and RON 75, Google 2/8 = 0.25 and RON 25.

History is not rewritten

Each attribution is saved with the model and window used. A later change to the model does not alter sales that have already been attributed. Raw browsing data is kept for 30 days; orders, attributions and daily totals remain.

What Admetriva attribution is not

  • It does not replace Google Ads or Meta attribution. Each platform claims its own conversions, so their totals can overlap.
  • It does not measure leads: the tracker measures products and orders.
  • Browser orders are gross revenue; exact revenue comes from the store's server-to-server integration.

Didn’t find your answer? Write to us and we’ll help.

Contact