We suggest upgrading to IE 7 or downloading Firefox for a more enjoyable web experience.
The strengths of Bayesia’s technology for marketing in 18 points
With BayesiaLab:
The manual modelling of expert knowledge by the Bayesian network will allow you to model your complete risk, fraud, or even customer behaviour chain. The graphical representation of Bayesian networks and BayesiaLab’s ergonomic design make it an invaluable brainstorming and communication tool.
You can use the power of non-supervised learning to extract the set of significant probabilistic relations contained in your databases (base conceptualisation). Apart from significant time savings made by revealing direct probabilistic relations compared with a standard analysis of the table of correlations, this type of analysis is a real knowledge finding tool helping one understand phenomena.
You can characterize your target variable through supervised learning.
For example, this target variable can represent fraud, or the propensity to buy
a product or even satisfaction. The Bayesian assessment model obtained from a set of
independent tests (case not used for learning) will provide the overall precision of
the network, a confusion matrix (instances, reliability, precision) through which network
behaviours can be recognized as a predictor, and an interactive lift curve as a decision
aid for choosing the threshold at which point it is economically interesting to carry out
a marketing action.
You can also use the power of the Markov blanket search algorithm to select the
minimal sub-set of variables really significant in relation to the target variable.
The segmentation of your databases (clustering) may be carried out in two manners.
The detection of variable homogeneous groups will allow you to define latent variables
that you can use either to reduce the size of your analysis or to integrate these new
variables with manifest variables. As for the detection of groups of homogeneous
individuals, it will help your experts to adapt to marketing campaigns.
If expert knowledge is available, BayesiaLab can rigorously integrate it
as well as your databases (update of Bayesian expert knowledge according to feedback data)
You can test the lever effect (e.g. improvement of the image, deployment of
training actions) by enriching your automatically learned networks with expert knowledge.
The association of costs with these levers will allow you evaluate different policies.
These adaptive questionnaires will return the most relevant questions in terms of
target information contribution and in terms of costs associated with the questions
(cost of acquiring the information). When you have made your choice, the answer will
be taken into account to determine new more relevant questions.
You can use your Bayesian networks offline to classify your databases
(add 2 fields to the entry database: the predicted variable and the probability
associated with this predicted value).
By using the BayesiaLab analysis toolbox, you will have a really good understanding
of your data: analysis of the strength of relations, analysis of relations between
target and other variables, analysis of relations between a specific target value and
other variables, analysis of observations to be able to know whether all the observations
are fully concordant or if some are contradictory, causal analysis cancelling the
direction of the arcs when these can be inverted without modification of the probabilistic
significance of the network.
You can «play» with your networks to test hypotheses (What-if scenarios)
Initially, you can use the power of the BayesiaLab algorithms
to automatically learn a Bayesian network, modeling the choice
and with Bayesia Market Simulator:
Then, from a specification of your market (file containing a set of individuals
described by their characteristics), you can use your Bayesian network to determine
the market shares hoped for in relation to the characteristics of the products
placed on the market (scenario).
Finally you can rigorously carry out your missing value imputation tasks using the set
of values entered. The distribution of probabilities on the possible values of missing
values is in fact inferred by using the Bayesian network.
You can very easily test all the scenarios you want on niches, i.e.
by selecting a subset of market characteristics
You can save, for each individual, the probabilities corresponding to each product
tested, as well as the rejection probability for the set of products.
You can also define rules governing similarity between products.
The simulator will consider that in the event of an individual rejecting
one offer from a set of similar offers, he will then choose none of the offers
of this set. In other words, the rejection will be the result of the common
characteristics of the offers and not the specific characteristics of each of the offers.
Finally, the generation of log files (text or html) will allow you to keep invaluable
records of all your simulations.
Now see more details about our marketing software: BayesiaLab (handling and analysis of Bayesian networks for simulation, analysis and decision making and Bayesia Market Simulator.