The TIBCO Platform is a real-time, composable data platform that will bring together an evolving set of your TIBCO solutions - and it's available now!
A chart showing the TIBCO Platform vision
Jump to content

About This File

The Next Best Action Accelerator provides a reference architecture and code assets for building an event-driven marketing platform for customer engagement. It is configuration-driven through a custom web interface based on TIBCO Spotfire which allows marketing personnel to define target audiences and offers for customer engagement.

 

converted-file.thumb.png.a69ed63817751de3c7ea9dd3c6432a9e.png converted-file.thumb.png.c904153d30473c4a615d088c26dfa99d.png converted-file.thumb.png.72ec0809de72aff774b2d8653c57d415.png converted-file.thumb.png.888461794a87f26993d3ad11f4ef8977.png converted-file.thumb.png.57c79aa2969c59f5fdd4d3868fa8e58b.png converted-file.thumb.png.6dbb06f46bce1bae89b450527e070b07.png


Here's a video showing the Accelerator in action:

Business Scenario

Event-driven marketing is based on capturing customer events and then determining audience membership in real-time. Linked offers are then issued to customers who can then take some action to receive the benefits of those offers. Various types of real-time events can be handled, such as purchases, entering a store, or completing a qualifying context for an offer. The event then triggers an audience selection based on who the customer is and what they have done in the past. This may include the use of data science models for segmentation and/or propensity. A customer may match several audiences, which in turn may match several offers. A series of best action rules then apply to ensure the optimal offer for the customer is selected and issued. Again, this may involve the use of data science propensity models, as well as business rules.
 

Concepts

The Accelerator configuration is driven by two primary static data objects:
 

  • Audience ? a targeted customer group defined by selection attributes and triggering events
  • Engagement ? any interaction between the platform and customers


Audiences are composed of a series of filters that narrow selection down to specific criteria. Currently this includes customer demographics and customer segmentation using a PMML model. Audiences also include a triggering event that cause them to be evaluated. This could be a Purchase, Position, Campaign Trigger, or other arbitrary Generic Event.
 

Engagements are used to make offers to Audiences. When a triggering event occurs, it may trigger off zero to many Engagements based on audience membership. Default functionality is to examine all these offers and determine which is the best based on business rules, and then issue only that offer to the customer.
 

An offer follows this lifecycle:
 

  • Matched ? the customer and event matches one of the included Audiences, but none of the excluded Audiences
  • Issued ? the offer is determined to be the best for a given set of offers triggered by an event, or the Engagement is marked as Engage All Matched
  • Qualified ? the customer has completed one of the qualifying contexts for a given offer and they are awarded the next action


The business logic to determine the best offer from the set can be customized. In the Accelerator the following decision path is taken:
 

converted-file.thumb.png.122aee1190d447448fe6e96a44fe911f.png
 

The steps are:
 

  1. For each matched Engagement, are there one to many qualifying contexts with a propensity model?
  2. If the Engagement has qualifying contexts with propensity models:
    1. Calculate the propensity for each qualifying context that has a model
    2. The propensity for the offer is the maximum propensity of all those qualifying contexts
  3. Calculate the value of the engagement to the customer using:
    1. For Coupon type, the monetary value of the coupon
    2. For Points type, using a nominal 0.02 currency units per point
  4. Sort the Engagement list first by descending propensity for those that have propensity values, then by offer value for those that do not
  5. The Engagement at the top of the list is the best action and is issued to the customer


For example, this set of offers has a mix of propensity and offer values:
 

EngagementPropensityOffer Value
Bonus Points for buying Chalk0.4300.05
Bonus Points for buying Home or Clothing Accessories0.1200.25
Bonus Points for Toy Wagon for Male Customers 2.50


Those offers with propensity are ranked higher than those with only offer value. In this case Bonus Points for buying Chalk is the offer the customer is most likely to engage with, therefore it is issued as best offer.


The dynamic data model for the Accelerator takes the form of report events:
 

  • Order ? customer purchases one to many order lines of products
  • Position ? customer position is recorded through consent via a mobile app
  • Generic ? this is a catch-all event that can be used for arbitrary named events with an optional single value
  • Offer ? an offer has been made to a customer; this is generated by the Accelerator or externally
  • Campaign ? trigger for a campaign


Any of these event types may trigger off an Audience selection and customer Engagement cycle.
 

Benefits and Business Value

The accelerator provides immersive visual analytics and machine learning to magnify the power of the marketer resulting in dramatically improved effectiveness of customer interactions.
 

Technical Scenario

The accelerator demonstrates several scenarios for event-driven marketing along with some sample audience and offer configurations. There are two different contexts provided: Retail and Telco.
 

In the Retail case, the first scenario demonstrates order processing with offers being generated in response to customer purchase events. Audiences are identified and offers made to customers based on rules and model-driven propensity analysis. Customers then complete qualifying contexts to receive awards like points and discount coupons.
 

The second scenario shows how location events and geofences can be used to detect when customers are within certain defined areas, or within a specified radius of store. Offers are generated based on location and customer attributes.
 

In the Telco case, these two scenarios are also implemented. There is also a scenario to show how a campaign trigger can be used to target a broader audience, and examples of how arbitrary events like churn risk and dropped calls can be used to personalize offers for customers to improve engagement.
 

Components

converted-file.png.757e8039e37d06c704bbf49a921e0abf.png

×
×
  • Create New...