Deciding Data

Use AI to tell the right story.


The Challenge

You can tune many aspects of your digital marketing, including your paid bids, value propositions, images, text copy, and target audience.

Of these, your value propositions are the most important. They express the reasons customers should work with you. If you get those reasons right, it improves all your interactions with potential customers. This improvement applies to both paid and unpaid channels, customers that are just learning about you or making a final decision.

Beyond that, knowing what an audience wants is more than just a marketing optimization tool. It helps you discover what to do next. With this information, you can learn:

  • What to emphasize in future marketing images and copy.
  • Which product features should be the highest priority to build or maintain.
  • What products you should offer next.

We use AI to help you see what value propositions work for each audience.


Measure The Message and The Mechanics Separately

  Ads that express value proposition 1 have been resonating the most with marketing professionals. These ads may warrant a bit higher proportion of the budget (particularly ad 3). But, it might be worth further testing different ways of expressing it. There are indications that this audience may be relatively sensitive to how this value proposition is framed.

All the ad variants for value proposition 3 have seen similar performance. Even though we have only been running them for a few days, there is a good chance (78%) that they have the highest conversion rate with marketing professions. But, their click-through-rate is below average. These might be good candidates for ads in a different medium, such as email or paid search.


Better Together

AI is great at testing and optimizing. Humans are much better at telling meaningful stories and creating emotionally compelling ads that connect with customers.

This software helps humans and AIs use each of their talents most effectively. Humans create ads and suggest possible audiences. The AI estimates which aspects of an ad show promise, sharing what it learns from automatic statistical tests. This allows us to quickly focus on the best performing ads and discover what each audience really cares about.


Open Source

Behind the curtain, this software uses a combination of Probabilistic Models, Reinforcement Machine Learning (beyond black box models), and useful visualizations and interfaces.

We are working on releasing much of it under the Mozilla Public License 2.0. However, due to privacy concerns we won't be releasing some of the data used to augment certain models.