There is an interesting interview from TRG Arts regarding the benefits to arts organizations from hiring someone to manage data. Heather Kitchen, Managing Director of Dallas Theater Center* says:
When I began as the chief administrative person at a theater 25 year ago, even a large regional theater did not have a computer driven ticketing package. As time evolved, and we moved past Lotus 123 spreadsheets for tracking ticket sales, I observed and appreciated the power of accurate data and how it could be a powerful tool – not a decision maker, but a tool.
True. Notice that this is not what is sometimes wrongly described as ‘big data’, but instead ordinary tracking of sales, donations, responses to marketing and to the events themselves, data for which there is better and better software for managing and displaying the numbers. Ms Kitchen is quite correct that data analysis is a tool, but only that – human beings still have to make decisions about where to target new marketing, or whether to change subscription prices.
My worry is that sometimes this tool is misapplied, to the point of leading to poor decision-making. The misapplication comes from a conflation of average and marginal – it is important for arts managers to understand the difference. Here are two examples:
- Suppose your museum has a very good data analyst, who has been looking at the characteristics of people who buy memberships. The analyst has found that on a per capita basis, people living in postal districts W and X tend to be more likely to obtain a membership that people living in postal districts Y and Z, even though all four districts are equidistant from the museum. You have the opportunity to expand localized marketing. Should you focus on W and X, since you seem to be having success there, or should you focus on Y and Z, perhaps reaching an untapped market? The thing to realize is that at this stage, at least, we don’t know the answer to that question. We know that on average we are having more success in W and X, but that does not tell us whether an increase in resources at the margin would have more impact there, or in districts Y and Z. Historical data, even thorough, accurate historical data, can’t tell us what to do.
- You have been running a series of concerts, with price differentials for orchestra and balcony seats. On average, you sell about 80% of your orchestra seats and 65% of your balcony seats. Should you raise, or lower, either of your two prices? Again, with the information given, we don’t know. That’s because we don’t know (yet) about the response at the margin – do consumers react strongly or weakly to price changes from the status quo (i.e. is demand elastic or inelastic at current prices), and how do they respond to the differential between the two prices. Even very accurate data on sales rates at current prices won’t tell you that. In economists’ terms, the data are giving you a point on the demand curve, but not telling you about what the whole curve looks like.
But all is not lost. There are ways data can help you make decisions. What is needed is a willingness to experiment. In the first case, suppose you increased marketing efforts in W and Y, but not in X or Z. Suppose you found there was a bigger effect on memberships in Y than in W. Then I might suggest that in future, targeting what are currently weak areas is a better strategy than targeting those where you are already strong (assuming there are not vast demographic differences between these four areas). In the second example, try small changes in one price or the other, and track what happens: for what prices and ranges are customers price-sensitive, and where are they not? This can guide you in finding whether a price cut will yield an increase or a decrease in revenue.
In each case, you need to find out something about the effects of marginal changes. And you cannot do that simply by looking at historical data (unless your data is already full of various experiments in adjusting prices or marketing or fund raising strategies). Simply collecting numbers isn’t enough – you have to be willing to be something of a scientist, devising experiments.
A final note: some of these experiments will yield negative results – an increase in marketing targeted at a population segment that ended up having no effect on demand; a price decrease that ended up costing the organization revenue, as only a few extra tickets were sold. That is a consequence of being experimental. A current buzz circulating is about the ‘benefits of failure’ – this is surely an application. And the benefit in all of these cases, if you are paying attention, is that you learn something.
* In my country this would be a Theatre Centre.
Trevor O'Donnell says
Thank you for this concise description of data usage, Michael. I wonder, though, if you can conduct such an analysis without considering content. In order for these data to be accurate predictors, the content of the communications would have to be rigorously consistent.
But in every new cycle arts administrators use different messaging, which is typically developed by amateur staffers based on creative impulses and personal opinions about the marketplace’s motives. Can we really make rational, data-based choices when such an important variable is so uncontrolled?
In both the goegraphic and price examples I think we need to have more information. Who are the people in postal districts W and X and why do they respond so well? And what is it that motivates certain people to choose balcony over orchestra? What data should we be collecting about these customers’ needs and desires in order to be able to develop persuasive messaging?
The arts are wise to begin using data, but if we don’t take an equally sophisticated approach to the content of our persuasive strategies, the data choices we make will be of limited value.