From data to insights

- Now United Fans Engagement Analytics Platform

Now United is a global pop music group featuring 14 artists. The group’s performances and activities are primarily through digital platforms like Instagram, Facebook, YouTube, etc. Now United label manager looks for fan engagement insights to help strategic decisions.

Roles

Lead designer of a team of two, collaborated with data scientists, PM, developers and customer stakeholders.

Responsibilities

Research, design workshop, UX and visual design, customer engagement.

Deliverables

Responsive analytics dashboard.

“How might we help Now United label manager gain fan engagement insights by SAP ML technologies”?

- Problem Statement

Final Solution

We interviewed the Now United management team on Sunset Blvd in Los Angeles to define the problem. We discovered the main persona Christina and her frustrations. Christina is often puzzled by some events that have no explanations.

“Every day I need to guess the sentiment. The problem isn’t about lacking information but the lacking of an efficient way to guide me to the conclusion.”

-Christina, Now United Label Manager

For example, the most intriguing mystery at the time was the sudden jump in subscriber numbers to the Youtube official channel in Saudi Arabia.

Now United wanted us to find the reason behind it so they can replicate it in other international markets.

We clustered unresolved questions into different user needs buckets and came up with four prominent use cases. Then, we turned them into design ideas.

Pitching to the stakeholders

The iPad demo based on customers’ requirements was presented to Juergen Mueller (SAP’s CTO) and Simon Fuller the founder of Now United.

Artist fan engagement was measured by mentions, subscriptions, comments’ sentiment scores, followers, profile views, etc.

There are massive data from four social channels, e.g., data categories, sentiment comments, and ML results.

We need to turn data complexity into simple insights so that the label manager Christina can quickly digest and take action.

First, Simplified the Information Structure

I like simplifying problems based on the fundamental structure. I sorted the user needs by priority and designed a flat structure with good discoverability and scenario coverages.

This flat structure is designed to reflect who, where, what, and when. Christina can quickly drill in from different angles with various exploration combinations.

For example, who > where >what, or where > what> who> when.

Artist related. - Who

Geo-location related. - Where

Social Channels. - What

Video related. - What

Anomaly Alert. - When

The combinations cover the questions we collected from the user interviews as the following:

Second, designed four layers of the user journey.

Anomaly Detection.

Quick Benchmark.

Locate Issue.

Discover Root Cause.

1. Proactive alert of the Anomalies.

I designed the anomaly alert with quick actions to help Christina get to the critical insights that might have been neglected. The proactive alerts are surfaced by the system automatically, which puts Christina on auto-pilot without the need to dive into the overwhelming metrics all the time.

2. Quick benchmark - Simplify massive social channel data to one score.

Now United Artist score (NU Score) was an invented term for benchmarking the artist's performance.

With the design simplification, the Label manager can quickly get a sense from just a single number.

The NU score summarizes the artist’s performance over a period of time. It gives a snapshot without providing the trend and changes over time.

I presented the metrics with more dimensions for the NU score variance with the same data set.

Here is an example to gain insights into the fluctuation and consistency of each artist’s performance.

At first, the data scientists categorize the NU score with a linear grade system from a score of 0 to 100. Without putting thought into how the end-users would perceive the grade and the coloring, this is obviously discouraging because everyone would look like a low performer.

I suggested adjusting the threshold to improve the NU Score distribution. A tiny change makes the insights more humane, positive, and friendly.

3. Drill into a particular metric to locate the issue.

To help with the cause-effect analysis, I iterated the initial NU versions and provided actionable next steps by adding a NU score breakdown by the data source.

It provided a straightforward visualization for Christina to grab any exact metric number from her weekly reportings.

An extra benefit of the design - a simulation feature can be easily added that can simulate how different scenarios affect the NU score.

Christina can have an As-If analysis by giving different weights to each metric based on their definitions.

One of the visualization challenges was data on different scales. For example, metrics like Impressions were in Million/weekly, but subscriber gain was in thousands/weekly.

I compactly visualized data by a double-ended chart to compare those data.

4. Discover the root cause.

There are various metrics behind “fan engagement”, and an interesting one is the social channel comments. The comments are helpful for investigating a root cause of an incident, however, it’s a heavy workload for label managers to read thousands of Youtube or Instagram comments.

To simplify the process, the data scientist applied sentiment analysis. I organized the sentiment results and grouped the comments in an easily consumable way.

When exploring data visualization design options, one seemingly straightforward proposal is to add depth by grouping comments and mentions into a single table chart. This Swiss army knife solution is space-saving because more detailed insights are expanded. However, the nested structure can cause users to lose track of the data hierarchy.

According to the persona needs, I landed on a solution with a series of simple correlated charts in a master/details structure that the user can see the entire insights story journey, reducing their mental workload.

This Design decision is driven by persona,

A flat structure allows Christina faster deliver the analytics report by easily switching between elements, doing comparisons, and jumping from one scenario to another.

Christina is not tech-savvy: She can see the information and any actions relevant to that detailed item by a master-detail pattern. It’s easy to understand and doesn't need a lot of explanation.

Save page space isn’t a priority to Christina.

UI Library Choice

In the initial stage of the project, we started with more explorative rich & sophisticated UI elements to better pitch the solution to the customer. As the project went on with more specific deliverables and a more concrete timeline, we decided to use a UI kit library SAC directly.

We were able to cut the efforts in engineering implementations without any compromise on the user's needs. I successfully managed the expectations from all stakeholders.

Solution Test & Impact

“The mystery of Saudi Arabia and our explosive success in the Middle East has been explained. The code was cracked by SAP and their incredible analytics.”

- Simon Fuller, Founder of Now United