IHC attribution model

With the IHC model, we want to combine and evaluate two perspectives:

  • interaction between consecutive sessions within the customer journey,
  • evaluation of performance and engagement within each session in the context of the overall customer journey.
  1.  

The sessions within the customer journey (CJ) are related to each other by their IHC values, time sequencing and rate of engagement.

IHC is fully data-driven and at its heart, IHC uses the mathematical concept of partial truth, also called fuzzy logic.

Please have a look at the following videos, where we explain IHC and also the usage of the results.

IHC Attribution - A three Phase Interaction model

IHC stands for the interaction phases: InitializerHolder and Closer.

IHC model graph

The IHC model projects/ maps each individual session into the three-phase concept of customer interactions.

  • Initializer is the stage of awareness and exploration for a potential customer,
  • Holder is a period of continued customer attention and interest being kept, and
  • Closer is the homestretch, the checkout phase where the action is taken.

The three phase concept, under various names, of interaction has been a pillar of marketing science for 30+ years. 

What is IHC?

Here we explain the basic concepts of the IHC attribution model. (Note that the membership curve explanation in the video is a simplification of our quantile-based membership approach.)

Basic Insights

We analyze the IHC attribution results and which information we can obtain with the IHC view marketing campaigns.

Combined IHC insights

In this video, we focus on how to evaluate marketing campaigns based on IHC results: along with some examples, we use different perspectives for campaign performance evaluations.

IHC Parameter Insights

When training your customer journey (CJ) data, you will obtain some insights. One of the main insights are the IHC phase weights, see below chart. The phase weights represent the importance of each phase in the overall buying process. The sum of the three weights is 1.0 (=100%). A high weight can be interpreted such that this phase tends to spread of multiple touch points and a longer period. Whereas a low weight represents that this phase is usually covered with less touch points and needs therefore less attention, hence a lower weight.

Attribution service model IHC phase weights
Attribution service model results initializer

The initializer function shows the Initializer phase membership of a subsequent session, based on the hour difference to its prior sessions = hours since last session. This function is based on the inverse quantile function of the inter-session times in your customer journeys. You see that fast, or normal, session follow up times do not receive initializer memberships. But the larger the inter session gap becomes, the higher the initializer membership in this session.

The closer membership function is in reverse time direction and shows the closer membership of a closer valid engagement session in relation of the hours to the conversion, 0 is the time of conversions. This function follows the quantile function of your customer journeys, to represent the usual closer phase engagement behavior in time to conversion moments.

Attribution service model results closer
IHC attribution model results graphic

Last but not least, the IHC training algorithms is also estimating channel specific IHC weights or impacts. A value of 1.0 represents that the channel shows an average strength in respect of the IHC phase compared to the other channels in your customer journey data. A value above 1 represents a comparably high phase impact strength when present in CJs. Correspondingly, a value lower then 1 stands for comparably low impacts. The learning algorithm takes the channel sample size into account during the estimation.

The IHC parameter training and hence the strength of the IHC results depends heavily on the training data set of your customer journeys. We recommend to

  • Train on 10-100k customer journeys, test different samples sizes, the more homogeneous your customers, the less training data you need to obtain stable results
  • Use different conversion types for e.g. new conversions and return conversions, but also for different markets
  • Regularly re-train your IHC parameter on at least bi-weekly or monthly basis

 

In case of questions, please contact us. We are happy to support you in the IHC setup.