Predicting exhibition performance is a tricky business. We usually talk about forecasting in general for visitor attractions being a matter of guess or grind – either pick a number, or spend months pouring over spreadsheets. For exhibitions specifically, it’s usually the former: up until now, picking how an exhibition might turn out has really been anyone’s guess. That’s because, hidden under the surface of a relatively simple number, there’s a whole pile of things going on. All sorts of factors influence exhibition performance – the nature of the show, the curatorial angle, the intended audience, budgets, location, what’s showing elsewhere and more. Unfortunately for museums, getting the numbers wrong – which the industry admits to doing most of the time – has a significant impact. Over estimation usually means missing revenue targets on admission and merchandising fronts, which compound to kill the exhibitions bottom line. Underestimate and risk underselling the exhibition budget itself or associated promotional efforts, with the missed opportunity of valuable visitors you could have seen walk through the door.
A year ago, Dexibit took up the challenge of predicting exhibitions for London’s National Gallery. Since then, we’ve analyzed millions of exhibition attendees from all over the world at different museums, to develop a machine learning model capable of simulating exhibition performance. Over the past six months, we’ve been in beta testing mode. Exhibition predictions, it turns out, aren’t the easiest forecasts to test – the team often left biting their nails for months to see their predictions unfold in the real world, once the exhibition doors opened.
But this journey led to a few interesting discoveries. We worked out that each exhibition has its own behavior curve, that these usually relate to classification, indicative of audience type. We talked to lots of museums about how they evaluate exhibition performance and determined a common practice on how to answer the question of what makes a good exhibition. And we realized that the beauty of predicting exhibitions was actually simulation.
Because how these museums were using these forecasts was not to ask for just one number. They didn’t just want to know how the exhibition would do at say, $25 adult admission, but how it would perform at various prices. Or how that picture would change if those prices were variable. Or what we would recommend, if total visitorship were the biggest priority. The same was true for all sorts of other factors – budgets, schedules – even the styling of the exhibition itself. Sometimes these questions were asked before the exhibition took place, at other times they were simulated as a part of the exhibition’s retrospective, to analyze what could have been as part of lessons learned reflection.
So after predicting visitor behavior on everything from French impressionists to ancient sculptures, this week we launch our initial exhibition simulator in Dexibit. At your request, it delivers a variety of scenarios and advisories based on your chosen metadata. Given the delicate nature of exhibition simulation, we’ll be doing this via a controlled rollout for the next few months, designed to keep a tight reign on quality forecast features.
To access exhibition simulation:
- Click on the ‘Forecasting’ module from the left hand menu
- Select ‘Exhibition Simulator’ from the choice of forecasting models
- If you have multiple exhibitions forecast, select the exhibition of your choice from the top right of the forecast ribbon
To use exhibition simulation, Dexibit will need several years of historic exhibition metadata. To find out more, turn to your data concierge.
Release date: October 10th 2018