Most arts organizations grew last year. The sector average still fell 25%.

This is where most coverage of the 2025 report goes wrong, so it is worth being careful.

The headline is that average total revenue fell 25% from 2023 to 2024, below pre-pandemic levels for the first time since 2021, and down 36% across 2019 to 2024 once inflation is accounted for.

But two things in DataArts' own methodology complicate that.

The sample changes every year. Organizations reporting fell from 5,011 in 2019 to 2,914 in 2023 and 1,486 in 2024. DataArts deliberately include everyone who completed a profile in a given year rather than only those present across all six, to reduce selection bias. So the 25% figure compares one set of organizations against a different, much smaller set. It is not like for like, and the widely quoted "more than 6,500 organizations" is the six-year total rather than the 2024 sample.

In the matched comparison, most organizations grew. DataArts ran it themselves on the 1,288 organizations that reported in both 2023 and 2024, and found that 62% experienced an increase in revenue. Their conclusion is that losses were "concentrated among a subset of organizations and not a universal experience".

So the honest answer is that the sector average fell sharply, driven by concentrated losses most likely at larger organizations, while most individual organizations in the matched set did better than the year before. If your own numbers went up last year, you are not an outlier.

Two of those rows together tell you where development sat in the priority order. Overall personnel expenses fell 23%, while payments to artists fell only 11%. Organizations protected the art and cut around it, which is defensible and probably correct, but it means the people who maintain donor relationships are part of what got thinner.

And attendance rose while the number of programs nearly halved. Fewer things, better attended, run by fewer people.

Most of what these findings point to is not a software problem, and we would rather say so.

A 25% fall in foundation support and a 26% fall in government funding are not things a tool fixes. Neither is rebuilding a reserve, nor deciding whether to run fewer programs, nor the question of whether your earned revenue model still works. Those are board and leadership decisions, and anyone selling you software as the answer to them is selling you something.

Two things here are different, because they are retrieval problems rather than strategy problems.

Knowing which relationships are going quiet while there is still time. With a median surplus of 1% and four in ten organizations holding three months of working capital or less, the cost of noticing a lapsed supporter late is higher than it was. That is a question with a correct answer sitting in records you already own.

Keeping what your team knows when the team shrinks. Staff counts are at a six-year low, down two part-time and two full-time positions on average. Each departure takes relationship context with it, and there are fewer people left who remember.

If the choice is between a tool and keeping a development post, keep the post. If the people you have are spending their week working out who to contact rather than contacting them, that is the gap this closes.

Since this page is about a sector running on 4.2 months of working capital, it would be dishonest to overstate what a tool changes.

It does not raise money. It does not replace a foundation grant, and it will not tell you whether to cut a program.

What it catches is the thing that goes wrong quietly. On a Monday the list is already built: the supporter whose recurring gift stopped in September, the household that moved from $500 to $5,000 and got the same acknowledgement as everyone else, the giving anniversary falling on Thursday, the first-time donor from a fortnight ago who has heard nothing since, the board member who mentioned a connection two years ago that nobody followed up.

Each carries the reason it surfaced and the records behind it, so the morning goes on deciding what to say rather than working out who to say it to. For a team that lost two positions last year, that is the difference between a stewardship programme that runs and one that gets postponed.

A person still writes the notes. That part should not change.

Not directly, and this is the most likely reason we are the wrong fit for an arts organization today.

Our direct connectors are Salesforce Nonprofit Cloud and NPSP, Bloomerang and Little Green Light, plus Mailchimp. None of those is an arts patron system. If you run Tessitura, Spektrix, PatronManager or similar, you are working through CSV export and document upload rather than an automatic overnight sync.

That is a real limitation rather than a detail. It works, and it is how most organizations would start, but it is a manual step. Worth testing with a single export before you decide anything.

Patron and donor records carry giving histories, household relationships and staff notes about people who never agreed to any of it being typed into a chatbot. Consumer AI plans do not come with the agreements that kind of data needs.

Identifying details are removed before anything reaches a language model, your data never trains anyone's AI, and each organization sits in an isolated tenant. If you are writing rules for your own team first, our AI acceptable use policy template is free and ungated. Our PII redaction whitepaper sets out how the redaction actually works, if your board or your IT lead wants the mechanism rather than the assurance.