Defining 'data' for cultural organisation: data horizons
In the past couple of weeks, I’ve read and listened to a couple of good things about ‘data’ in the cultural sector – the challenges, opportunities, and what to do about it all.
In doing so, it occurred to me that a term as broad as ‘data’ is only a useful shorthand up to a certain point. You can only get so far before you have to put the brakes on and say ‘wait, what do you mean by that?’
That’s because we use it to cover a huge list of sources, uses, and processes that each have very different considerations. And as soon as you start using absolute terms in reference to abstract nouns it's hard not to think of all the exceptions.
And if that opening sounds familiar then that's because I said the same thing six years ago when talking about 'digital' and I applaud your memory.
What were you reading / listening to, Chris?
I enjoyed David Reece's recent series of articles for Arts Professionals. The first one, What do we mean by being data-driven? contained this quote:
Being truly data-driven requires acknowledging the limitations of data. It’s in these gaps that innovation often occurs. Bold decisions go beyond anything the data can tell us.
I also listened to episode 2 of the Digital Culture Network podcast where James Akers interviewed Ollie Couling, who said:
I think data is the enabler to something. It's not the answer to something. Exploring that idea a little bit, we have all this historic data. It's great. It helps us make decisions. But it shouldn't be the thing that makes the decision. It should be the thing that starts the conversation or creates a hypothesis or that you use to make some assumptions. The assumption making - that kind of hypothesis making - that's the strategy bit.
The data is there to help you shape your decisions and not to tell you what to do. Because, if you think about that logically, if you're only working on data from the past, that's creating its own kind of feedback loop in itself. You need to go out there. You need to think of new things and explore.
(Lightly edited to remove conversational bits. Full transcript on the link above).
None of which is particularly wrong. It's just that the use of 'data' here is way too broad to be able to talk in specifics and I thought I might be able to add some specificity.
Data horizons
So how about this? A simple categorisation that can be applied to different types of data which, looked at over different time periods, lend themselves to different uses.
I've jotted down a few use cases that are specific to the arts and culture sector, but I'm sure you can think of more.
| Horizon | Description | Purpose | Data Velocity | Example Use Case |
|---|---|---|---|---|
| Immediate | Real-time views and automations | Instant action | High | Personalised website content and timely comms |
| Short | Reactive analytics | Rapid response | High | Monitoring and automatic optimisation of live campaigns |
| Medium | Operational optimisation | Process improvement | Medium | Post-campaign reviews, content planning, audience segmentation |
| Mid-to-long | Predictive analytics | Anticipate trends | Low-Medium | Forecasting attendance and allocating marketing resources |
| Long | Strategic insights | Guiding longer-term strategies | Low | Multi-year audience engagement/programming strategy |
| Furthest | Historical analysis | Context and evaluation | Static | Evaluation of arts programs |
(Apologies if that table doesn't work too well for you on a mobile device).
In case you're thinking about it, I did consider putting examples of data sources against each horizon but realised that's not feasible. The same data can be viewed differently on different timescales and combined in various ways for various outcomes.
Applying the data horizons
With everything laid out, perhaps you can see why I'm uncomfortable with absolute statements to the effect that 'data shouldn't make decisions for you'.
Sometimes it can and should. Sometimes it can't and shouldn't. Some types of data lend themselves to that kind of use and some don't.
There's a spectrum that runs through various points including:
- automated by data
- data-driven
- data-informed
- sod the data, we're doing it anyway
To take a very specific example, at the 'Immediate' horizon, you would want to set up proper data collection (of page views and conversions) so that it can drive the automatic optimisation of your Meta Ads campaigns.
Other data might automate whether you try to upsell a ticket buyer to a membership or send someone a post-show email.
Because there are specific use cases, this is where you'll find off-the-shelf tools to do the job for you. As the horizons get more distant data can become less specifically actionable and is more likely to be one of many inputs needed to make decision or form a hypothesis.
In this context, the points made by David and Ollie make the most sense when looked at through the lens of the 'Long' horizon where the aim is to identify strategic insights.
Even then, I'd stick up for data somewhat. Dismissing data as secondary risks reducing strategy to guesswork disguised as creativity.
Data (in the form of market trends, performance metrics, benchmarking, and customer preferences) can be just as valuable an input to a strategic vision as organisational capabilities, macro-environmental factors, competitive intelligence, and gut feel. Data vs creativity is a false binary.
One last thing for now. I like Ollie's point that:
if you're only working on data from the past, that's creating its own kind of feedback loop in itself.
…and I've certainly seen 'strategies' that amount to little more than post-rationalising existing activity, with targets that are barely incremental rather than (as you'd usually hope) potentially transformational.
However, at other data horizons, feedback loops are brilliant and necessary. When you're getting into the execution of a strategy, that sort of iterative learning is needed so that the approach, and the strategy itself, can be refined over time.
I think there's more to say on this topic, and maybe spelling out the various types of data would be a useful companion piece to this. But I've not been writing much on this blog recently so one step at a time.