AUDIENCE & BUSINESS
Houston Grand Opera
Understanding when opera audiences become donors.
A Rice D2K collaboration with Houston Grand Opera: connecting audience profiles, donation timing, and fundraising decisions through survival analysis.
View presentation poster- Customers before cleaning & filtering
- ~13,000
- First-donation RSF · Evaluation C-index
- 0.733
The source data covered approximately 13,000 customers across survey, ticketing, donation, and marketing records. Unusable or unsuitable records were excluded during preparation; the experiments below use their eligible subsets, not all 13,000 customers. The analysis used 46 features, including age and income.
THE DECISION
Who should the team understand and engage next?
I worked across data integration, EDA, survival modeling, evaluation, and recommendations in the Rice–HGO team, and presented the findings through a poster and live Q&A.
Data Sources
Survey · Ticketing · Donations · Marketing
Features
Age · Income · Attendance · Subscription
Analysis Methods
Kaplan–Meier · Cox · Random Survival Forest
From Customer Profile to Outreach Plan
A Profile to Explore
- Age55+
- EducationBachelor’s or higher
- Household Income$125k+ / year
- LocationHouston
- RelationshipSubscriber
- EngagementHigh satisfaction & recommendation intent
A group-level profile to guide outreach exploration, not a description of every donor or an individual eligibility rule.
Recommended Actions
- Frequent Attendance
Give repeat visitors relevant follow-up and consider annual visit frequency when planning outreach.
- Subscription Relationships
Build on existing subscriber relationships and strengthen subscriber benefits.
- Strong Recommendation Intent
Support engaged customers with a good experience and opportunities to recommend HGO to others.
Recommendation intent asks “Would you recommend HGO?”; satisfaction asks “How was your experience?” The notebook’s nps field is an individual response. Group NPS is the share of promoters minus the share of detractors.
These are proposed actions informed by the analysis; resulting fundraising gains were not measured in this project.
Where Are the Customers?
Mapped customer ZIP codes to explore their geographic distribution across the United States. The heatmap helps describe audience reach; warmer areas indicate stronger concentrations in the displayed data, not higher donation rates.
When Might a Customer Donate?
The question is not only whether someone donates, but how long it might take. Cox links customer features to donation timing; Random Survival Forest (RSF) combines decision trees to capture more complex patterns.
| Task and Model | Training C-index | Evaluation C-index |
|---|---|---|
| First donation · Cox, full feature set | 0.731 | 0.717 |
| First donation · Random Survival Forest | 0.800 | 0.733 |
| Second donation · Cox, before feature selection | 0.603 | 0.581 |
Saved notebook outputs, rounded to three decimals, from 80/20 splits. These are exploratory results: evaluation data were also consulted during feature and parameter selection. They are not a fresh, untouched final test. The poster’s second-donation score of 0.61 is from a full-data fit; feature-removal experiments reached about 0.625 on the reused evaluation split.
Reading the Results
C-index: Who Donates First?
If A donates before B and the model puts A first, that pair is correctly ordered. A score of 0.733 is roughly 73 correct out of 100 comparable pairs; 0.717 is about 72, and 0.581 about 58. A score of 0.5 is near random ordering—not a customer’s donation probability.
Cox: Which Features Relate to Earlier Giving?
The full first-donation model gives annual attendance a hazard ratio (HR) of 1.11. For 4 versus 3 annual visits, holding other features equal and before either person donates, the instantaneous donation rate is about 11% higher—not an 11-percentage-point probability increase.
Kaplan–Meier: Who Is Still Waiting?
It estimates a waiting curve from observed records. If 10 people are followed for a full year and 3 donate, the curve ends at 70%. Someone observed for only six months still contributes those six months; they are not treated as someone who will never donate.
Reading example (illustrative): if a curve is at 0.70 in year 2, the chance of still waiting for the target donation is 70%; the chance it has happened by then is 30%. “Survival” here means the donation has not happened yet.
INTERACTIVE DEMO · PROFILE TO ACTION
Understand One Customer. Plan the Next Step.
Build a prospective first-time donor’s profile. Explore timing, a gift scenario, and a possible next action.
University–industry collaboration: source data and models are private. This demo visualizes the project’s output format using illustrative rules and example values.
Fields reflect the notebook and audience profile. Age increases in small steps in this demo only; the Cox results did not establish a positive age effect. Region bands are illustrative: ≤50, >50–150, and >150 km.
How the Time Window Changes the View
More time creates more opportunity for a first donation. The example keeps this profile unchanged.
What Drives This Example?
A Possible Next Step
Five illustrative propensity levels: very low <15%, low 15–<30%, medium 30–<50%, high 50–<70%, very high ≥70%. Gift size is an input assumption; the project’s survival models estimate donation timing, not gift value.
