Data scientists, business intelligence professionals, and marketers have learned to expect a status quo when it comes to using artificial intelligence in their marketing:
It’s time-consuming and requires handoffs between teams, marketing systems for data collection, and activation.
As reported in Harvard Business Review, “It has been a common trope that 80% of a data scientist’s valuable time is spent simply finding, cleaning, and organizing data, leaving only 20% to actually perform analysis.”
That’s a lot of time spent simply organizing datasets.
When it comes to building models on customer data for marketing, data scientists are handcuffed by fragmented systems that don’t provide a centralized, single source of truth.
In turn, they’re not able to act on the output of models due to system disparity.
Enter AI Workbench, an integral component of our leading customer data platform (CDP), which was crafted with both the modern marketer and expert data scientist in mind.
AI Workbench: Built to facilitate better AI marketing for BlueConic customers
We’ve worked with over a dozen customers to find the right solution to these problems — and that’s why we’re excited to officially launch BlueConic’s AI Workbench.
With direct access to a unified profile database and connections to marketing activation channels, AI Workbench enables data science and marketing teams to work more efficiently.
Data scientists, analytics, and business intelligence teams have the ability to import notebooks and customize predictive models within the customer data platform.
Marketers have built-in models they can tweak and test, without having to hard code them or rely solely on data science teams to pull customer scores when they need them.
AI Workbench embeds Jupyter notebooks within a CDP to give you:
- Access to unified customer data for advanced machine learning modeling (both for anonymous and known prospects and customers)
- The ability to immediately deploy models across multiple external systems
- A marketer-friendly UI for machine learning modeling (i.e. customer lifetime value, propensity to buy and churn, look-alike modeling)
- Shared efficiencies across and between data science, BI, and marketing teams
What’s this look like in action? Let’s take a look at a few examples.
Access to unified customer data for modeling
As mentioned, data scientists spend 80% cleaning data, pulling lists, de-duping data, and merging contacts — and only 20% of their doing higher level work that they were hired to do: helping the business solve a challenge and make decisions that impact the bottom line.
With AI Workbench, data scientists can pull data from a single, unified database.
Being built on a profile database, customer data is persistent, and profiles update in real time at an individual level. Data scientists can import their own models or tune pre-existing models in BlueConic by selecting the specific profile properties (i.e. customer attributes) they want to use or pull in data via an API.
For instance, a data scientist could hypothesize which customer attributes influence a customer’s risk for churn, then pull those attributes into a model.
Relying on a pure-play CDP as a data source allows them to look at a broader range of attributes that update in real time, which makes the model smarter.
Getting models into production with ease
After a data scientist builds a model for a business user (i.e., a marketer), it is often difficult to get the model into production across various external systems.
AI workbench makes it easy to apply a model against certain profile properties and store the result as a part of the profiles. Data scientists benefit from this in two ways:
- 1) They can quickly experiment with different model approaches (e.g., if the click-through rate of an email based on a predicted next best action is not actually better than the control group, the data scientist can easily change the model and try again)
- 2) Sending the result/prediction/customer score to multiple platforms is just as easy as sending it to a single platform. For example, you could use a model to calculate customer lifetime value (CLV), create segments based on high or low CLV scores, then send these segments back out to your email service provider, CRM system, and to personalize on-site experiences all in one place.
Improve KPIs with real-time inputs for better models
AI Workbench brings your models closer to the data and reduces data latency.
Machine learning models are pulling from profile data that is constantly being updated, then profiles are updated with data from notebooks as often you’ve scheduled.
Let’s say you’ve created an uplift model to understand which customers are likely to purchase, regardless of whether or not you spend ad dollars on them, and which customer will not buy regardless of the number of ad dollars you spend on them.
You can save on your ad spend and focus specifically on those who need to be persuaded — driving up your click-through rate and lowering cost per acquisition.
Shared efficiencies across and between teams
The proximity of models to customer data and marketing systems allows marketing and data science teams to limit the number of hand-offs between teams and iterate on ideas at an accelerated rate.
For example, with parameters, data scientists can flag different variables in code to serve as the “plug-and-play” variable. A marketer can go in and select one of the available variables to then create models on their own.
Watch our webinar on AI marketing to learn how our pure-play CDP enables business users across organizations to build, train, and deploy models with ease.