Role: Product Design Lead
Company: Ometria
Year: 2020 (Project in discovery)

Background

Over the last 12 months, our team (Ometria labs) created a pilot for selected clients to prove Ometria’s AI capacity. The goal was to demonstrate to our clients that predictive segmentation algorithms can increase performance (defined as opens, clicks and revenue). Predictive Segmentation has successfully passed the pilot stage.

The next step is Productionisation.

Objectives

• Allow customers to create effortlessly predictive customer segments that will lead to high performing campaigns and drive more revenue.

• Show customers the efficiency of predictive segments and push them to use this feature coupled over manual segmentations

• Show prospects of our innovation and retail focus. Productionizing this feature will support our branding and perception in the industry.

Definition of success

• Customers regularly using predictive segmentations: 10% of the campaigns are predictive

• Customers using predictive segmentation see an uplift of 30% in revenue or more

Clients interviews  

I worked closely with the product owner, technical lead and UX content writer, they were present in the majority of the interviews

• 10 client interviews have been performed

• Customers joined with multiple stakeholders, mostly CRMS and Customer Experience managers

Discovery

• Clients who participated in the pilot were intrigued by the promise of the predictive segmentations (Feelunique seen 95% uplift in revenue per email)

• They wanted to avoid manual job and manual guessing

• They wanted to see an Improve in the revenue: “Show me the proof that this generates more revenue than the manual process”

• They wanted to keep the current user flow to create campaigns. They were familiar and happy with the process

Ideation: V1 standalone builder

Each ideation sessions are done parallel with technical design, so we followed the recommendations that our engineers provided us.

One of the ideas that we explored was the stand-alone version of a predictive segment builder.

• This would create predictive segments that could be used in other places in the platform

• We let them dig deeper into the data so they can see negative and positive prediction about the product attributes

• We focused on the AOV metric, showing the user the uplift they will have if they used this type of segmentation versus their historic data

• We used our design library for fast high fidelity mockups and prototypes so we can test with users that are already in the pilot stage and get early insights

V1 standalone builder issues and limitations

• While clients and stakeholders were excited about the promise, this flow didn’t accelerate the process.

• Clients didn’t have time to dig into this new data set, they told us that the promise was to avoid the manual work.

• Creating a data set, storing it then use it to create in the campaigns was serious friction for clients.

Ideation: V2 builder within the broadcast campaigns

• Use existing broadcast campaign builder flow to integrate the predictive segmentation, therefore, reducing the steps to create predictive segmentations

• Show fewer data and let the system make the decision and creating the best user segments that will drive the revenue. This will streamline the process and reduce friction.

• Successfully tested with 5 retailers that also tested the previous version of the journey. Users have easily accomplished user journeys. They voiced their opinion that version 2 fit much better into their daily usage

• Most of the questions were around the technical aspect of the project. They wanted to know the logic behind the prediction and how accurate it was.

Outcomes

• Version 2 of Predictive segmentation is on alpha test with 4 retailers

• Another 4 retailers will join the test in March 2021

What I learned

• Expectations of users for the word “predictive” varies a lot. The word predictive in my opinion while very valuable from the sales side created unrealistic expectations from the customers.

• The level of trust plays a big role in feature adoption.

• Big retailers prefer to streamlined and easy to use flows vs small profit margins.

• Clients were happy to try new features as long as it fits in their sales & business strategies.