Agent Based Modelling (ABM) is a new approach to modeling systems comprised of interacting autonomous agents, or, in the case of Atapa, people. It is one of the most exciting practical developments in modeling since the invention of relational databases.
The effectiveness of advertising is traditionally modelled using regression approaches. These approaches involve the statistical analysis of previous advertising campaigns. From this analysis the effectiveness of future campaigns can be estimated.
Regression approaches are useful because they are relatively simple to apply and are strongly based on empirical data. However, they're only really good at predicting the past as real advertising campaigns can exhibit significant non-linearity and time-dependence. These features cannot be adequately estimated without an accurate model of the underlying causes. Regression approaches don't provide for time-based analysis ("longitudinal effects") nor rapid what-if scenario testing.
Agent-based modelling involves a “bottom up” approach to estimating advertising effectiveness. Instead of estimating the behaviour of an entire population in a single step, the population is modelled as a group of independent agents. Every agent can interact with other agents via word of mouth, as well as maintain a memory of previous exposures to the advertising message. The overall response of the population is then simply a summation of the individual responses (as in the real world).
As with any model, there is a requirement for testing and validation of all assumptions and data. Even so, modelling the population as a collection of individuals allows a more intuitive depiction of behaviour, and can give more accurate estimates of the response to an advertising campaign.
