Here’s a question that keeps artistic directors awake at night: what should we program next season?

It sounds simple. It isn’t. A major opera company’s season represents millions of dollars in investment, years of planning, and a complex balancing act between artistic ambition, audience expectations, singer availability, budget constraints, and the lurking fear that you’ll program a bold new work and nobody will come.

Traditionally, these decisions have been made through a combination of experience, instinct, and educated guesswork. Artistic directors draw on their knowledge of the repertoire, their sense of what audiences want, and their personal aesthetic vision. It’s part art, part science, part gambling.

But increasingly, there’s a new element in the mix: data.

What Predictive Analytics Actually Means Here

Let’s be clear about what we’re talking about. This isn’t an algorithm choosing the season. No computer is spitting out a printout that says “Program Tosca in March and Wozzeck in July.” The decisions remain human, creative, and ultimately subjective.

What predictive analytics does is provide better information for those human decisions. By analysing historical ticket sales, subscriber behaviour, audience demographics, economic conditions, and competitive programming, models can estimate how different repertoire choices are likely to perform at the box office.

Think of it as weather forecasting for opera seasons. The forecast doesn’t decide whether you take an umbrella. You do. But it’s useful to know there’s a 70% chance of rain.

Who’s Doing This

Several international companies have been quietly building analytical capabilities over the past few years. The Met has long used data to inform its programming, and European houses including the Bavarian State Opera and the Dutch National Opera have invested in audience analytics platforms.

In Australia, the picture is more nascent. Opera Australia has been developing its data capabilities, particularly around subscriber behaviour and ticket purchasing patterns. Smaller state companies have less data to work with but are finding creative ways to use what they have.

Some AI consultants in Melbourne have been working with performing arts organisations on predictive models that go beyond simple historical analysis. These models can incorporate external factors — economic indicators, tourism patterns, even weather forecasts for outdoor performances — to build more nuanced projections.

The technology isn’t prohibitively expensive. Cloud-based analytics tools have brought the cost down significantly, and even mid-sized companies can build basic predictive models using commercially available software.

What the Data Can Tell You

I spoke to several people involved in arts analytics, and the insights they described were fascinating.

Repertoire risk assessment. Data can estimate the box office impact of programming a lesser-known work versus a proven favourite. This doesn’t mean companies should only stage crowd-pleasers — but knowing that a challenging new work is likely to sell 65% of capacity rather than 90% allows you to budget accordingly, plan your marketing, and perhaps pair it with a high-selling title in the same season.

Optimal scheduling. When you stage a production matters almost as much as what you stage. Analytics can identify which nights of the week perform best for which types of repertoire, how far in advance different audience segments book, and when to launch marketing campaigns for maximum impact.

Subscriber churn prediction. For companies that rely heavily on subscription revenue, understanding which subscribers are likely to lapse — and why — is enormously valuable. Models can identify at-risk subscribers before they leave, allowing targeted retention efforts.

New audience identification. By analysing the characteristics of current attenders, models can identify potential new audience segments and the programming or pricing strategies most likely to attract them.

The Artistic Tension

Now, here’s where it gets complicated, and where I have mixed feelings.

Opera is an art form. Programming an opera season is a creative act. The best seasons tell a story — they juxtapose works in interesting ways, introduce audiences to unfamiliar repertoire, take risks that expand the company’s identity. An artistic director programming purely by numbers would produce safe, repetitive seasons that maximise revenue and minimise artistic significance.

I’ve seen this happen in commercial theatre, where data-driven programming has led to an endless parade of revivals, jukebox musicals, and adaptations of known properties. The result is financial stability and creative stagnation.

Opera cannot go down that path. The art form’s vitality depends on new works, bold interpretations, and occasional glorious failures. Data should inform programming, not determine it.

The artistic directors I’ve spoken to understand this. The good ones use data as one input among many — alongside their artistic vision, their relationship with audiences, and their responsibility to the art form. Data tells you the likely consequences of a decision. It doesn’t tell you whether the decision is worth making.

A Practical Example

Let me illustrate with a hypothetical that’s based on real conversations.

An artistic director wants to program a new Australian opera in the 2027 season. The data model suggests it will sell around 60% of capacity — below the breakeven point for the company’s main stage. Without data, the decision is binary: program it or don’t.

With data, the decision becomes more nuanced. The model might show that pairing the new work with a popular Italian opera in the same week could cross-subsidise the programming. It might identify a subscriber segment that’s most likely to attend new works and suggest targeted marketing. It might recommend a smaller venue where 60% capacity is actually profitable. It might project that the new work will attract media attention that increases overall brand awareness, benefiting ticket sales across the season.

The data doesn’t say “don’t program the new work.” It says “here’s how to program it smartly.”

The Australian Context

Australia’s opera ecosystem is particularly well-suited to analytics-informed programming, for a few reasons.

Our companies are small enough that changes can be implemented relatively quickly. We have a concentrated market — most opera-going happens in a few major cities — which makes data patterns clearer. And our geographic isolation means that touring international productions is expensive, making local programming decisions even more consequential.

The Australia Council for the Arts has been encouraging arts organisations to build their data capabilities, and several universities have established research programs at the intersection of arts management and analytics.

Looking Ahead

Predictive analytics won’t replace the creative judgment of artistic directors. It shouldn’t. But it will make those judgments better informed, better supported, and more likely to produce seasons that balance artistic ambition with financial sustainability.

The opera companies that figure out this balance — that use data to enable risk-taking rather than prevent it — will be the ones that thrive. And as someone who cares deeply about the future of this art form in Australia, that prospect makes me cautiously optimistic.

Just don’t let the algorithm pick the cast. I draw the line there.