Through the years we at Comet have delivered a dizzying number of decisioning programs acrossmyriad industry verticals, business drivers and channels—engaging with many functions across typically large, complex organisations to deliver impressive commercial benefits and delightful customer experiences.

After a while, we started to notice clearly identifiable core functional areas of the decision logic architecture that would be repeatedly built again and again no matter the industry, use case, etc.

When we joined all this together into a single, cross industry foundational logic blueprint, we realized that we had really struck a note with clients who were looking to embark on this decision management journey. Clients told us they were looking for a best-practice template to work from.  That they don’t want to re-invent the wheel. And that they want to benefit from our many years of experience and learning to avoid the potential pitfalls and rabbit-holes that others can find themselves in.

Clients were eager to buy in to a repeatable architecture that was proven to perform at scale and deliver commercial benefits in some of the world’s largest organisations.  This OmniChannel architecture would start our clients “higher up the mountain”—significantly reducing risk, getting to ROI quicker and providing from day 1 a platform on which a truly OmniChannel implementation can be delivered over the length of a multi-phase, multi-project program of change.

Over time we have made many industry- and use-case-specific additions on top of our foundation logic blueprint that speak to some of the more specific challenges or opportunities your industry and organisation may face. At the core, however, always lies the beating heart of any of our decisioning engine implementations:  the OmniChannel Foundation Logic Architecture.  This architecture is made up of 5 functional areas that together deliver a decision management solution no matter the industry, use case etc.  Frankly, if you’re not doing at least these 5 things in your decisioning architecture; you’re not really doing decisioning, and that’s the truth!

Here are the 5 functional areas that all OmniChannel decision management architectures need to account for:

Each of these 5 functional areas can further be broken down to specific key capabilities delivered within them.  The sequence and configuration of these key capabilities are the secret sauce to our architecture that ensures performance at scale and ongoing agility that will allow your architecture to extend and grow across your organisation in to the many channels of interaction between you and your customer.

Broadly speaking each functional area covers:

Prepare Data:

There are a number of offline data preparation activities that need to be optimized for an efficient, maintainable system.  These can include(but are not limited to):

  • Customer Data Profile
  • Propositions (Offers & Treatments)
  • Customer Journeys
  • Champion Challenger definitions

Configure Propositions:

Once we have defined a customer data profile and sets of proposition data to import in to logic at runtime, the first thing we need to do is to configure the list of propositions down to a qualified subset of propositions that are valid, eligible and relevant.

In addition, we need to fetch all of the disparate model data (propensity scores from adaptive and/or offline predictive models) and set them on each of the qualified propositions, ready for subsequent use later in the logic flow.  Same goes for customer journey data.

Optimize Propositions:

Now that we have configured our subset of valid, eligible & relevant propositions we need to optimize them toward some commercial objective aligned to corporate strategy.  There are many commercial objectives to choose from depending on where the customer is in their lifecycle (e.g. onboarding when the customer has recently joined, x-sell / upsell,etc.

Our architecture caters for a wide set of commercial objectives and encourages the decision engine to decide in real time which commercial objective is the most important to pursue right now. For example,. if the customer just told us they want to leave, a retention optimization strategy focused on minimizing revenue dilution is critical – it’s not time to attempt to upsell!

Do not under-estimate how important it is to implement a tried-and-tested prioritization algorithm.  To a large extent your relevancy model (covered in the configuration function) and prioritization algorithm are where the money is to be made in these decisioning programs.  Best to count on a tried and tested model that has been proven to deliver 100’s of millions of dollars in organisations similar to your own.

Personalize Propositions:

Now that we have a qualified & optimized subset of propositions it’s time to personalize the delivery of those propositions in a way that makes sense for this customer or contact center agent.

The way personalization is applied depends greatly on who we are outputting these propositions to.  It could be that we are outputting them to an agent in the contact, therefore it is highly beneficial to outline why this particular proposition is being made to this customer today.  Highlighting relevance and connecting to data that the agent can see in their profile panel really helps the contact center agent engage and comply with the next best action recommendations we are asking them to communicate.

Similarly, if we are outputting an offer directly to a customer in a digital channel then some personalized, dynamic messaging that outlines the value to that customer can significantly increase the likelihood of that customer accepting this offer.

Construct Presentation:

Last, but certainly not least: while everything in the previous 4 functions has been omnichannel, we must of course be sympathetic to each channel’s unique content preparation requirements.

Each channel does require certain content, for example, id’s, contact history preparation and volume restrictions in order for it to render content appropriately.  Here we take our qualified, optimized and personalized subset of propositions and prepare them to be presented in the specific channel that is requesting a decision to be made right now.

In addition, a critical channel specific concept is that of managing the delivery of these propositions across multiple containers in a channel.  If you ask yourself “ok, I have these 15 offers that are highly relevant and optimized and I have 3 available containers on this screen to populate – which offers should I put in which containers” then this container management concept is critical for maximum acceptance of the offers we place in front of customers.  It is not as simple as and certainly not world-class to simply assign the top offer e.g. to the hero banner at the top of the screen.  Sophisticated machine learning has moved significantly beyond this, and so should you!

In summary, our OmniChannel foundation logic architecture contains all of the best-in-class thinking from across our wide range of decisioning implementations and makes that available in a performant, prebuilt solution that assures a successful, de-risked and commercially attractive decisioning program.

This blog is meant as a high level introduction to just the foundational piece to our suite of accelerator intellectual property.  Please reach out to to learn more about this, and about the rest of our full suite of accelerators..


Nicholas Gent
Practice Lead – Intellectual Property

Nicholas Gent

Practice Lead – Intellectual Property

Twitter | Linkedin

Nick leads the Intellectual Property practice for Comet in North America.

His role is to accelerate the delivery of our decisioning programs by providing prebuilt best-in-class decisioning architectures that are proven to be performant at scale and deliver millions of dollars in benefits for our clients. 

Nick has 8 years’ experience implementing & supporting decision management solutions in to some of the world’s largest and most complex organisations including Everything Everywhere, Vodafone, BSkyB, Royal Bank of Scotland, Elisa, Comcast, Verizon Wireless, DirecTV, Cox Communications, Fifth Third Bank & PNC Bank