Oh we are so close! And yet large gaps remain. Marketing can be almost completely intelligently automated away, leveraging large data sets and prior experiences as well as content creation paradigms to create, launch and execute marketing campaigns in an effective and entirely automated way. The promise is real, and piece by piece, this will happen – let us examine what lies ahead.

To claim successful automation of the Marketing function, driven by AI, one has to be able to complete all the key aspects of a successful Marketing Campaign. I focus on the Campaign here, as it is one of the most valuable parts of the Marketing Function that drives awareness, consideration and pipeline for Sales to then close, or in case of SaaS functions, generate revenue entirely through customer self-service. Also, I use the phrase AI in its broadest possible sense – covering any and all techniques that use data to make informed choices.

The lifecycle of a Marketing Campaign has multiple aspects, and each of them today has a strong human touch – varying only in degrees across companies. Briefly, one could identify Marketing Campaign as having these elements:

  1. Identify what campaign to launch
  2. Identify your target customers
  3. Identify how & where to find those customers
  4. Develop a multi-channel plan to engage your target customers
  5. Generate content/experience that is appropriate for the plan
  6. Set campaign launch on auto-execution
  7. Create reports that use standardized KPIs on campaign performance
  8. Find the ROI of the campaign
  9. Incorporate success/failure into your next launch

I will pick up each one of these in order in a series of posts. Let us begin by reviewing the first one – “Identify what Campaign to launch”.

Identifying Campaign to Launch

Identifying which campaign to launch is a heavily, human-centered, activity today. Often done in alignment with sales and/or business goals around specific products – the choice of a campaign is either mandated by business targets or else is a decision based on market research and the need to bolster certain product groups based on competitive situation, exogenous events (say security topics in the news), or other opportunities that present themselves.

How can AI help in this space? Irrespective of how human-centric the decision to select campaigns looks, the decision itself is often based on well-founded data and overarching business goals. And this begins by understanding the characteristics of the product portfolio.


Every product and product mix in a company’s portfolio will likely generate a certain combination of revenue growth and margin expansion (or contraction) based on forecast demand, competitive landscape, brand positioning and sales strength in that geography. Forecast models are used in data savvy companies that break-out probable product sales down to the monthly/weekly levels (usually as a cumulative probability distribution – which will tell you probability of selling at least x units in a given period of time), giving significant visibility to product managers and business unit leaders regarding future sales. Accompanying these models is also a forecast for revenue growth and margin impact from those sales.


Once this type of forecast data is available for each product in the portfolio, optimization algorithms can recommend an optimal product mix that should be sold to hit revenue targets while retaining margins and market share and positioning the company for future success (health metrics or power metrics as some call it).


Finally, the recommended product mix informs the model that recommends the series of campaign launches that must be undertaken to help meet the target for the business.

Each of the models indicated above has value in its own right, and should first become part of the very human process of selecting campaigns to run. Over time, the output of each these models can be an input into the next step in an automated cascade feed design that recommends the set of campaigns a business must launch in a given year to meet its target goals.

So how realistic are the above steps from an AI perspective?

The main limitation today in any corporation comes from not having enough product/family level forecast models that can then be used to understand optimal portfolio mix to meet business objectives. The technology is there, but the investment in the level of Analytics talent, or outsourcing of this capability, is not being done – and is one of those areas where the leaders in data (Amazon, Netflix, AirBnB et al) outdistance themselves from the rest of the industry. If the major revenue/profit centers would take advantage of forecast models, the subsequent steps of optimal mix and recommended campaigns will flow more easily.

Select Models:

a. Forecast product family sales for each family of products
b. Forecast revenue/margin growth with various product mixes
c. Given an overall company target, recommend product/family mixes to promote
d. Recommend calendar of launches

In the next post I will cover identifying target audience and the role AI is already playing in this space.



Sri Srikanth

Advanced Data Analytics & Strategy, Senior Data Scientist, Cisco Digial