Avatar

AI in Marketing is here and it’s real – and it is solidifying and revolutionizing the art and science of identifying target customers for driving marketing campaigns among leading, data-driven companies across multiple industries.

As mentioned in my previous post, the first task of any marketing campaign is to understand those parts of the portfolio that we wish to promote with business goals in mind, and once we know that mix, the natural next step is to begin identifying the right target list of customers to engage.

Current Customers

The first place to start of course is by understanding the current install base to:

  • Look for opportunities to renew directly
  • Look for opportunities to upgrade to premium versions
  • Look for adjacent product installations and opportunities to cross-sell
  • Subset the above list to companies that have money to spend on these products
  • Subset the list further based on degree of effort required versus probability of success

In other words, the primary focus has to start with those customers with whom the company has an existing relationship. If the company cannot succeed here, its long-term viability is in serious jeopardy.

So how does AI help here?

Almost immediately, three types of models show up here as invaluable:

  1. Predict the next likely product a customer may purchase/take interest in
  2. Forecast the money available for spending for a given technology at a customer site
  3. Indicate whether a customer primarily runs the competition’s products

Thus, if we know the portfolio of products that the company needs to promote (identified in the prior post), the above set of models can now identify and isolate customers who are good candidates to target by the marketing campaign. Some additional rules may need to be added to account for company specific nuances such as ability to deliver in those geographies, strength of partner network and more – but between the models and some simple rules, we can get a list of usable and valuable target customers intelligently and automatically.

If such models are being created, where do we fall short today?

While maintaining the installation base and product portfolio datasets are challenging, companies are continually moving towards a viable foundation in these spaces. The challenge often arises in nuances that exist on a per customer basis and ensuring that the right product mix is being presented to them. The first part of the challenge is on the modeling side, which can be addressed by bringing in features that address customer specific variations. The second part is in allowing dynamic offers to be created, rather than relying on static bundles. This has more to do with a company’s processes and internal alignment of the departments involved in creating offers rather than a technology challenge (to be sure, there are technical challenges, but dynamic bundle creation and pricing has been around for quite some time – one my early publications was in this space).

Net New Customers

Moving beyond the world of existing customers to those of net new customers – we need to first understand how and when do customers choose a new vendor, and what influences them to do so?

Boiling it down, customers are primarily influenced by:

  1. A pressing current need frequently exacerbated by unhappiness with current solution/solution provider
  2. Peer recommendations for solutions/solution providers
  3. Trusted Experts’ opinions in the space
  4. Brand reputation and proven track record of the vendor being recommended in the relevant space coupled with suitable perception of pricing, support and ability to deliver on the solution

On top of it, there is the steady attraction of building in-house solutions versus selecting a new vendor.

As may be becoming obvious, identifying net new customers that a company can reach out to is fraught with the limitations of the data available to the company regarding all of the above-mentioned factors, and reliance on first principle approaches of word-of-mouth, networking, and direct recommendations continue to dominate this space.

On the AI front then, companies such as Google (DoubleClick), Oracle (BlueKai) and Adobe (Audience Manager) offer marketing solutions that offer look-alike models, and once provided with the list of a company’s current customers, can suggest others who are behaving in a similar way across multiple websites and social media sites they track. This is a good place to start to identify previously unknown prospects, but that list needs to be further vetted to account for likely conversion to customer based on:

  • brand strength of the company in the region of the prospective customer
  • ability of the company to deliver solutions in that geography
  • strength of company’s partner network in that space
  • company’s reputation with key influencers in that area
  • existing install base of the company’s products in that industry

Of course, that does not solve for the factors mentioned earlier around pressing needs of the prospective customer, unhappiness with current solution provider and peer recommendations, but this is one of those situations where we have to leave the unknowables to remain as such and work with what we have.

From an AI perspective, in addition to look-alike models, a model that identifies the region/country of the prospective customer and leverages knowledge of company’s brand strength, ability to deliver solutions, strength of partner network and more to score the prospect on likelihood for conversion into customer would help deliver a prioritized list of net new customers to target.

Summary

To summarize, some of the models we need to identify target customers, with the input being the portfolio of products to promote, would be:

  1. Predict next set of products current customers may benefit from
  2. Forecast money available to spend at each customer site for company’s portfolio of products
  3. Indicate whether the company values competitor products versus company products
  4. Rule based model to cleanly identify renewal, upgrade, cross-sell opportunities
  5. Look-alike models to identify net new customers
  6. Scoring new prospects for likelihood for conversion to customers

All these models are in use at various levels of effectiveness in different companies in the industry, and companies that are putting it all together are beginning to distance themselves from the competition. A company does not have to produce all these models in-house – but a mix of help from external agencies, AI vendors and in-house expertise can make the work of automatically and intelligently identifying target customers a rapid reality!

 


Posts in this series:

  1. Can Marketing be completely AI-driven & Automated? A Genesis
  2. Using AI to identify your Target Customers – automatically! (this post)