How does your institution forecast visitation? For many museums, the answer is either quick and rough, or months of time consuming and complex analysis. Either way, accuracy is often a challenge. Whether you need to know how many tours to schedule or muffins to bake, knowing visitor numbers in advance is imperative for balancing admissions, experience and efficiency.
Now with Dexibit, you can forecast visitation at the click of a button, using our patent pending technology and the latest in machine learning. No more guesswork. No more number crunching.
General availability release date: 29th November 2017
Thank you to Te Papa Tongarewa, our research partner museum for this project, and those that participated in beta.
Long range outlook
In Dexibit’s new forecasting module, you’ll be able to produce all sorts of predictions, starting with a long range outlook, which creates daily visitation statistics up to 3 years in advance. This is incredibly useful intelligence for museum functions such as financial planning, advertisement scheduling, front of house rostering and inventory ordering.
Dexibit’s forecasting works by feeding your historical visitation data into our predictive model, which determines the level of impact that multiple factors have on your venue’s visitation. The model then makes a prediction for future visitors, based on what it’s learnt. You’ll need a minimum of 12 months historical data loaded in – the more, the merrier.
Multi feature analysis
No two museums are the same – that’s why Dexibit’s model uses machine learning to automatically learn and adapt to the unique factors that influence visitation at your institution. If you’re in Philadelphia, the weather might be important. In Sydney, the local cruise ship schedule will have a part to play. For London, regional events could have strong impacts, both positive and negative. In Paris, the school term may be a big consideration.
Dexibit’s forecasts analyze for features such as:
- Day of week, including weekend versus weekday impact
- Public holiday
- School term
- Time of day (for museums managing granular, timed visitation)
- Exhibitions and events at your museum
- Regional events in your town or city
Most importantly, our machine learning model works out how each of these factors influences visitation at your museum all at the same time, because it’s often never one thing or the other.
While Dexibit already knows the seasonality, month, day of week and time of day of your museum’s rhythm, using the calendar you can also feed your forecast additional insight, such as exhibitions and events happening in the museum, or regional events, public holidays, school terms and cruise ship dockings happening in your town or city.
Remember to add both what has happened historically, plus what you are planning or expecting to happen in future. The more context you provide your forecast, the more accurate it becomes. When scheduling regional events into the future, you can load multiple events into a day to have a cumulative effect on your forecast and even choose which events typically have a positive versus negative impact for your venue, and how strong of an impact you expect that to be.
An important part of using a forecast is knowing how much you can rely upon it and the degree of variance to expect – whether it should only be used as a general guide, or whether you can have a high degree of confidence in the numbers. Dexibit’s forecasts provide transparency into how the model came to a result, in terms of the various influence the top features had on the numbers; alongside a predicted accuracy rating. This additional insight provides important information to help understand what is behind the numbers. To raise the accuracy of your forecast, remember to make sure you’ve recorded what’s happening in and around your museum in the calendar, both historically and into the future.
Coming up next…
We’re currently refining the forecasting model to push performance ratings even higher and provide additional data visualizations and contextual enrichment from your calendar to help you and your users digest its results. Next, we’ll be putting out our more granular visitation forecast, which provides down to the hour predictions up to 14 days in advance (currently in beta). Plus, we’re currently working on some exciting new forecasting research projects to provide even more predictive models for the machine learning enabled museum.