
An interactive installation is software in every sense, with one difference that changes every design decision. It's used by people standing up, for a few minutes, without having read any instructions and free to walk away at any moment. No onboarding, no manual, no second chance. If it can't be understood in three seconds, nothing happens.
For anyone producing events it's a channel unlike any digital content. The person is already there, in a space you've built, at a moment when they have attention to give. The software just has to avoid wasting it.
For a real estate developer in Milan we built interactive art installations using QR codes, a web app and generative AI. Visitors scanned the code, interacted from their own phone and watched the result appear in the installation. The principle holds in general: the phone the audience already has in their pocket is, in most cases, the most reliable way to make a physical object interactive.
In an ordinary product you can afford a screen explaining how it works. Here the explanation has to be the object itself: where it stands, the light, the visual invitation. If you need a sign telling people what to do, the project already has a problem.
Behind the user there's no technical support. If something freezes, the person doesn't open a ticket, they leave. That's why the system has to return to its starting state on its own after a period of inactivity, without anyone having to step in.
Then there's the environment, which works against you. The most underestimated variable is the network in crowded venues, because five hundred people in the same space means congested wifi and patchy mobile signal. The light changes from morning to evening, with the noise any unamplified audio becomes useless, and surfaces get dirty.
The public does unpredictable things. Someone touches three things at once, someone gives up halfway, someone deliberately tries to break everything. You have to design assuming it will happen, because it always does.
And the deadline doesn't move. An event has a date that isn't up for negotiation, and that's the constraint that decides, more than any other, how the work gets organised.
The first is about what the person should take away when they leave. An image to take home, something to share, contact details left behind, or just a moment to remember. They're different goals and they lead to different projects. If you decide after picking the technology, you end up building something elegant and useless.
Then you have to settle who controls the screen. If the interaction goes through the visitor's phone, hardware costs collapse and maintenance becomes simple, because there are no touch screens to clean or devices to keep an eye on. That's why the QR code, unglamorous and very reliable, remains the best tool in the vast majority of cases.
You also need a precise answer to what happens with no network, and "let's hope" doesn't count. The system must behave in a defined way when the connection drops, with content already loaded locally, a queue for requests and a dignified message. Of all the choices, designing for a missing network is the one that saves the most installations.
The last is who's in the room. An installation staffed by a trained person can afford a bit more complexity. One left on its own has to be simple and able to get itself back on its feet.
Test early and test on site. A technical rehearsal in the real venue, with the real network and the real light, is worth more than weeks of testing in the office. The power socket that isn't there, the glare on the screen at four in the afternoon, the signal that doesn't reach that corner: you only find these things out there.
Always plan a reduced plan B. What does the installation show if the external service doesn't respond, if the generative part takes too long, if the network drops? A poorer version that's still alive is infinitely better than a screen with an error message in the middle of an event.
During the event you also need someone on call, with remote access to the systems. Not because a failure is expected. Being unable to intervene for three hours on the one day that matters has a cost that nothing else compares to.
Part of an installation's value is decided once it has already been taken down, and it's the part people think about least.
If the experience produced content, like generated images, messages or things the audience made, that material lives much longer than the event, on the client's channels and on the participants'. Deciding at the start what format it comes out in and how it gets collected costs little. Recovering it afterwards, when it's scattered everywhere, costs a lot or simply can't be done.
If instead the installation has to tour several stops, the technical choices change. Content has to be kept separate from code, configuration has to be editable without development, and branding and copy need to be easy to adapt. That gets decided on day one, because making something built for a single evening reusable in practice means building it again.
Installations are often judged on gut feeling, and you can do better with little effort. How many people interacted compared with attendance, how many completed the experience, how many shared or left contact details, how much content was generated. The system can collect this data by itself, in aggregate form and without taking personal data it doesn't need. These are the numbers that let you do better next time and explain the investment to whoever approved it.
If the experience produces images of people, you also have to decide beforehand what happens to those images, how long they're kept and who can see them. That's not bureaucracy, it's a design choice, and it has to be communicated clearly to the public at the moment they interact.
In the end a successful installation is one that any passer-by understands instantly and still remembers the next day, even if technically it isn't the most ambitious.
It's software used in a physical space, standing up and for a few minutes, by people who haven't read any instructions. It has to make itself understood in a few seconds, return to its starting state on its own and work in an environment with an unreliable network, changing light and an unpredictable audience.
Because they move the interaction onto the phone the visitor already has in their pocket. Hardware costs collapse and maintenance becomes simple, with no touch screens to clean and no devices to watch over. It's unglamorous but very reliable, and in the vast majority of cases it remains the best tool.
You need to have decided that beforehand, with content already loaded locally, a queue for requests and a dignified message. Designing for a missing network is the choice that saves the most installations, together with a reduced but working plan B and someone on call with remote access.
