MKTATTRIBUTION Procedure
Heuristic Attribution Models
In addition to providing the Markov attribution model, the PROC MKTATTRIBUTION provides several heuristic attribution models: the first-touch, last-touch, linear, position-based, and time-decay attribution models. The first-touch attribution model attributes all credit for the conversion to the first-visited channel in the customer journey. The last-touch attribution model attributes all credit for the conversion to the last-visited channel. The linear attribution model distributes the credit for the conversion equally across all visited channels in the path. The position-based attribution model attributes 40% of credit to the first-visited channel, 40% to the last-visited channel, and the remaining 20% credit equally to the channels in between, if there are more than two visits in the journey. Otherwise, the position-based attribution model is the same as the linear attribution model. The time-decay attribution model attributes the credit increasingly to the channel that is closer to the conversion.
Assuming that there are n customer journeys, the number of conversions is m, where 
. 
is the pth channel in customer j’s journey 
with length 
, and 
. Let 
, where K is the number of channels, if 
and conversion occurs at the end of this journey; otherwise it equals 0. The contribution to channel i by the first-touch attribution model is calculated by
The bar chart in Figure 8 shows how the first-touch attribution model distributes the contributions for a journey that visits four channels.
Figure 8: Bar Chart for the First-Touch Attribution Model
The contribution to channel i by the last-touch attribution model is calculated similarly by
The bar chart in Figure 9 shows how the last-touch attribution model distributes the contributions for a journey that visits four channels.
Figure 9: Bar Chart for the Last-Touch Attribution Model
The contribution to channel i by the linear attribution model is calculated by
The bar chart in Figure 10 shows how the linear attribution model distributes the contributions for a journey that visits four channels.
Figure 10: Bar Chart for the Linear Attribution Model
If 
, the position-based attribution model is the same as the linear attribution model. If 
, then let 
and 
if and if conversion occurs at the end of this journey; otherwise it equals 0. The contribution to channel i by the position-based attribution model is calculated by
where 
, if , and otherwise it equals 0; and 
, if , and otherwise it equals 0.
The bar chart in Figure 11 shows how the position-based attribution model distributes the contributions for a journey that visits four channels.
Figure 11: Bar Chart for the Position-Based Attribution Model
Let f be the half-life parameter, 
if and if conversion occurs at the end of this journey; otherwise it equals 0. The contribution to channel i by the time-decay attribution model is calculated by
where 
is the normalizing coefficient for journey j.
The bar chart in Figure 12 shows how the time-decay attribution model distributes the contributions for a journey that visits four channels.
Figure 12: Bar Chart for Time-Decay Attribution Model (HALFLIFE=1)
