When you have 1000 product teams, each would have some mechanics particular to the product.
You normally have some efficient general approaches that make leads for everyone with a variability in parameters, but other than that you also have techniques that work for some products and not for others.
If you’re nuts enough to make 1000 different funnels you will end up understanding that:
- There are many similar funnels
- Seeing similar examples brings up an idea that other mechanics can be used hypothetically successfully in similar situations
- Similar situations often doesn’t mean that the product is similar
When it comes to clusterization, the normal way of doing it is to set up some machine learning over all ever created parameters to predict some target variable. What is the variable we are to predict?
Deal with a client.
Now, the list of potential clients is 8 billion people narrowed down further to some few millions which are relevant in some way and addressable.
If you compare 2 humans you see they both have 2 legs, 2 eyes, different height, different color hair, etc. How good are physical features to predict behavioral outcomes, like who’s gonna get married? Not so much.
Organizations have similar physical parameters that are easy to see. How many people work there? In what industry? What products are being sold? When you start clustering organizations, you come up with clusters, accounting to every feature you can see, but losing something intangible.
A deal predictability. What is the main concept behind deal predictability?
Two companies. One does car tire replacement, another sells software. They both purchased marketing services, electronic document management systems and insurance.
Machine learning says: it's a very different product cluster. Sales says: there’s some upselling potential.
The main concept behind deal predictability is a client problem.
When you think about a problem, it’s not just one thing. It’s a chain of problems that often have non trivial connections. And there are no client / problem dictionaries available to enrich datasets.
So
- what you actually try to predict is some weird interaction of problems that the organization has
- This can’t be studied physically as it is not visible
- But we can analyze done deals because it’s sort of a physical experiment verification of a problem being present in a client organization
To prioritize the deal predictability parameter, you can try studying the intersection of deals done. This will
- unwrap you the hard to see mixture of client problems,
- in a verified way,
- will help define new upsales strategies,
- and will be a good ground for unification of leadgeneration funnels to a set of products with an intersecting set of clients.
#marketing #sales #machinelearning #ai #leadgeneration