A museum director in Lyon asked us last year which audio guide platform she should buy. She'd been given four answers by four people. The procurement officer sent her a quote for a hardware fleet with a five-year maintenance contract. The head of IT forwarded a SaaS link. The curator wanted to keep working with the voice-over studio they'd used for the previous tour. A board member had heard about "an AI thing" at a conference.
All four answers were technically audio guides. None of them were the same category of purchase.
The word "platform" is the clarifying term here. The old buying decision was which vendor would record your tour. The modern one is which software your museum will operate for the next decade. They are not the same question, and treating them as if they are is how institutions end up locked into three-year contracts with systems that can't do what their curators now need.
A platform is software the museum runs, not a product a vendor hands over. That distinction sounds pedantic until you try to change a stop description on a Friday afternoon.
With a project-style vendor, the content is theirs. You commissioned the work, they produced files, those files live on their servers or their devices. When you need a change, you open a ticket. Maybe the change costs money, maybe it's included, but either way it's their turnaround time, not yours. We've seen museums wait six weeks for a two-sentence correction because the vendor had to schedule a voice actor and a sound engineer.
With a platform, the content is yours. You log into a CMS, edit the text, and the change is live before you finish your coffee. The vendor provides the infrastructure — rendering, delivery, analytics, synthesis — but the editorial layer sits with the museum. That shift changes how the guide behaves across a year. Temporary exhibitions get their own content without a procurement cycle. A factual error caught by a docent gets fixed the same day. The interpretive voice evolves as the curators' thinking evolves.
The easiest test: ask the vendor to walk you through exactly how you'd add a new object to an existing tour. If the answer involves their team at any step, it's not a platform.
The capabilities the category should have
The platform category has matured to the point where certain capabilities are table stakes. If a vendor doesn't have them in 2026, they're selling an older product in platform packaging.
Content management. A working CMS with role-based access, version history, and preview. Not a spreadsheet you email back and forth. If you want the deep dive on this, our piece on content management for museum audio guides covers it.
Multilingual delivery without per-language production costs. Traditional guides charged per language because each one required a new recording session. Modern platforms generate speech from a single source. Adding Japanese should not be a procurement exercise. If it is, you're paying for a production model that the underlying technology no longer requires.
Analytics beyond checkout counts. Which stops hold attention. Which ones visitors skip. What people ask about when they go off-script. Which languages are actually used. Where visitors drop off the tour. The data should feed into curatorial decisions, not sit in a PDF that nobody opens.
Integrations. Ticketing, membership, CRM, website, CMS, analytics. The guide is not an island. If the platform can't connect to your existing stack, you'll end up with parallel systems and reconciliation work. We wrote more about this in museum system integrations for audio guides.
Branding and white-labeling. The visitor experience should feel like the museum, not the vendor. Logo, colors, tone, domain. If a visitor opens the guide and sees the vendor's brand front and center, the platform is marketing itself on your visitors' attention. That's backwards.
Accessibility as a first-class concern. Transcripts, audio descriptions, screen reader compatibility, adjustable playback. Built in, not a compliance bolt-on after a complaint.
Pricing flexibility. The museum should decide whether visitors pay, how much, or whether access is free. That decision should be a setting, not a contract amendment.
Most vendors will claim all seven. Fewer will show you the CMS in a live demo with a realistic dataset. That's the point at which marketing turns into product.
How the economics actually changed
The platform category is inseparable from a pricing shift that most procurement teams haven't fully absorbed.
The old model required capital. A vendor quoted you a hundred thousand euros for hardware, content production, and setup, plus annual maintenance. You signed, paid, amortized the asset over five years, and hoped the vendor was still around at renewal. Usage had no bearing on what you paid. If ten thousand visitors used the guide or a hundred thousand, the invoice was identical.
The modern model doesn't work that way. SaaS-style subscriptions, per-interaction billing, and revenue share arrangements all tie what the museum pays to what actually happens on the floor. A platform with usage-based pricing has no incentive to collect a large check and disappear — the vendor's revenue only grows if visitor adoption grows. That aligns the vendor's roadmap with the museum's outcomes in a way that a hardware contract never did.
For smaller museums, this matters more than the feature list. A guide that requires no upfront spend and scales with visitor volume is a line item the finance office can actually justify. For larger institutions, it shifts the guide from a capital project requiring board approval to an operational tool that can be turned on, tested, and adjusted without the weight of a five-year commitment. The full picture on pricing structures is in our audio guide pricing models piece.
One implication that often gets missed: when vendors are paid on usage, they compete on adoption. That changes what they build. A usage-aligned vendor has a structural reason to make the guide easy to find, easy to start, and good enough that visitors complete it. A flat-fee vendor doesn't — once the contract is signed, their incentive is to minimize support costs. You can feel the difference in the product.
Most platform evaluations go wrong the same way. The museum fills out a feature matrix, vendors claim yes to every cell, the procurement team picks the one with the lowest price or the most familiar logo, and two years later the curators are writing workarounds.
The questions that actually separate vendors are narrower than the feature matrix suggests. We covered the full version in our audio guide RFP guide, but the short list:
- Show me the CMS with a realistic dataset loaded.
- Walk me through adding a new stop end to end.
- Show me analytics from a museum similar in size to ours.
- What happens when a visitor asks a question your content doesn't anticipate?
- What does a pilot look like, and what does it cost?
Any vendor selling a genuine platform can do all five in a working session. Vendors selling a repackaged service will handle two or three well and get vague on the others. The vagueness is the signal.
One more filter worth applying: look at how the vendor treats content ownership in the contract. If the terms are clear that the museum owns its content and can export it, that's a platform. If there's language about "licensed content" or "vendor-maintained materials" that the museum can't take elsewhere, the software is a rental dressed up as an asset.
The buyer archetypes
Not every museum should buy the same platform. Four patterns cover most situations we see.
The major institution with an aging hardware fleet. You have a hundred thousand plus visitors a year, a recognizable brand, and devices from 2014 that nobody wants to touch anymore. Your constraint is continuity — you can't turn off the existing guide while you switch. You need a platform with strong content migration support, a serious CMS your curators will actually use, and pricing that handles your volume without blowing up. Revenue share or a tiered subscription usually fits. The decision drivers are analytics depth and curatorial voice control, not price.
The mid-sized museum replacing or launching a guide. Twenty to eighty thousand visitors, a small team, limited patience for procurement theater. You want something you can set up in weeks, not months, and change without calling a vendor. Usage-based or low-tier subscription pricing works well here because it removes the upfront capex question. The important evaluation criterion is whether your staff can run the platform with the resources you actually have, not the resources you wish you had.
The heritage site or outdoor attraction. The guide often is the experience. There are no wall labels to fall back on. You need route-based navigation, offline capability, and strong multilingual delivery. A platform built for indoor chronological tours may not map to your use case cleanly. Ask for references at sites structurally similar to yours before anything else.
The network, foundation, or multi-site institution. You're buying for a portfolio, not a single venue. The decision is almost entirely about operational fit: can a central team manage content across sites, can branding vary per location, can analytics roll up, can pricing differ per venue? Most platforms handle one or two of these well. Few handle all four. This is the archetype where the "platform vs service" distinction has the highest stakes — a service model simply doesn't scale across a portfolio without constant vendor overhead.
If you're picking between hardware and software, hardware vs software audio guides gets into the tradeoffs in more depth.
The three-year test
Here's the filter we'd apply to any platform decision. Imagine it's three years from now. Your collection has changed. You've added two temporary exhibitions a year. Visitor expectations have moved. A new accessibility requirement has shown up. A language you didn't serve is suddenly a quarter of your visitors.
Can the platform you're about to sign for handle all of that without you calling the vendor?
If yes, it's a platform. If no — if any of those scenarios requires a new contract, a new production cycle, a new capital request — you're buying a product. Products age. Platforms compound. The institutions we've watched succeed over multi-year horizons are the ones that treated this decision as an operational software choice, not a content procurement. The ones still in pain are the ones that bought a project and expected it to keep up.
Most of the platforms in this category, including Musa, publish their CMS demos and will run a working-session evaluation against your actual objects. It's worth taking them up on it before the RFP even goes out. You learn more in an hour with a live system than in a month of written proposals.
Buy the thing you'll still be happy operating in 2029. Everything else follows.