Analytics, Q&A audit,
and reports.

Every Convo tour ships with the metrics you’d expect from an audio platform. How many visitors started the tour. How many made it through each stop. Which language they picked. How many opened a chat. Average session length. The shape of the day — busy morning, quiet afternoon, the spike after a school group came through. If your director wants a single chart that shows the tour is alive, this is the chart.

We surface these because procurement asks for them and because they’re useful. But I don’t think they’re the interesting part. Broadcast metrics tell you a visitor pressed play. They don’t tell you what the visitor was thinking when they pressed play, or what they wanted that the tour didn’t give them. That’s a different layer — and it’s the one the legacy audio platforms can’t produce, because their tours don’t have a back-and-forth in them.

Convo has a back-and-forth in it. So we can show you the layer underneath. For the broader pattern on why the broadcast metric set has aged out, see measuring museum audio-guide engagement.

When a visitor asks the guide a question — by voice, by text, or by pointing their camera at an object — the exchange is written into the admin portal. You see the question, the answer the guide gave, the language it was in, the stop it happened at, and the source passages the model pulled from to ground the response. If the guide declined to answer because it couldn’t ground the question in your materials, that’s in the log too.

Citations matter here. The visitor guide answers only from materials you’ve uploaded — the wall cards, the catalog entries, the exhibition essays. It isn’t reaching out to the open internet. The audit log shows which of your passages the model leaned on for each answer, so when a curator reads something that feels off, they can trace it back to the source and either adjust the source or tune how the model is using it. That feedback loop is the difference between a guide you trust enough to put in front of a public and one you don’t.

In a procurement conversation, this is the page I open. Not the analytics dashboard, not the script editor — the audit log. It is the most boring and most reassuring view in the product. You can read what your visitors asked last Tuesday. You can read what the guide said back. You can read where it got the answer from. If you want the underlying authoring side of this, see authoring; if you want the visitor side, see visitor Q&A.

Reading the audit log line-by-line is useful for one stop. It doesn’t scale across a thousand visitor conversations. So the admin also runs an anonymized aggregation over the questions for a tour, or across all the tours in your org, and produces three views: themes, gaps, and a takeaway memo. This is the Q&A Insights surface, and it ships as part of every tour above the pilot tier.

Themes are the topics visitors keep returning to — clustered by meaning, not by keyword. If half the questions at a portrait stop are versions of “who is she,” that surfaces as a single theme with a question count and a representative sample. Gaps are the inverse: the clusters where the guide kept declining or hedging because your source materials didn’t cover the ground visitors were asking about. Gaps tell you where the tour script and the visitor’s curiosity are out of phase.

The takeaway memo is two or three paragraphs of plain prose that summarizes the pattern. Not a chart, not a dashboard — a memo. The version you’d paste into the slide your director carries to the board, or into the quarterly report that justifies the next exhibition cycle. It’s the artifact that makes the rest of the data legible to people who don’t live in the admin portal.

Interpretation has always been one-way. A curator writes a wall text or a tour script, the visitor reads or listens, and the curator finds out whether it worked by — what, exactly? Anecdotes from the floor. A once-a-year visitor survey. A graduate student parked in the gallery with a clipboard. The loop, in other words, takes a year and is mostly vibes.

When the guide is a conversation, the loop becomes weeks. The wall text and the audio script are now hypotheses about what visitors want to know. The Q&A pattern is the data. If the takeaway memo for the spring exhibition shows three weeks running that visitors keep asking about provenance and the tour keeps not having the answer, the next exhibition gets a provenance section. If a curator’s favorite stop is the one nobody finishes and nobody asks about, that’s a signal too — maybe the framing is wrong, maybe the object is in the wrong place.

I think this is the most underrated thing the product does. The audio is the thing that gets demoed, but the feedback loop is the thing that changes how the institution works over a year. The longer argument is in the audio guide is not the product and, on the question side specifically, in the questions visitors ask.

A few things to name plainly, because procurement asks and because I’d rather you find out from me than from the trial.

We don’t track dwell time at the gallery level. If your question is how long the average visitor stood in front of object number four, that’s answered by gate sensors and computer-vision systems, not by a phone in a pocket. We can tell you how much of a stop a visitor listened to and whether they finished it. We can’t tell you whether their eyes were on the work while it played.

We don’t roll Q&A themes up across multiple Convo admin orgs. The clustering is per-org by design. If you’re a network of three independent museums on three separate Convo accounts, you won’t get a unified theme view across all three. Within an org — across many tours, sites, or exhibitions — the rollup works.

And Insights doesn’t replace your visitor-survey program. A survey gives you what visitors will say to a clipboard; Insights gives you what they already asked the guide. Those are different data sets and they answer different questions. Both are useful. If you’re choosing between them, choose both.

Pricing for tours that include Insights is on the pricing page.