A cultural guide should make a visit easier to understand, but ease is not the same as the absence of every deliberate action. Product teams often treat a tap, selection or confirmation as a cost to remove. Sometimes it is. Sometimes that action establishes a fact the system otherwise does not know. When the guide begins speaking about an object, the relevant fact is not merely where the device appears to be. It is what the visitor is considering and whether the explanation belongs to it. I learned that distinction through a feature I approved too quickly.
In our fictional 2017 pilot, proximity-triggered playback appeared to offer an effortless experience. Twenty-four walkthroughs produced fifteen starts for the intended object and nine for another room. The wrong start was not necessarily a visibly broken interaction. It could be fluent and informative while being attached to the wrong thing. That made the defect harder to recognise. A visitor might try to reconcile the commentary with the object, assume they had misunderstood the display or follow the guide's attention elsewhere. The product had introduced uncertainty while presenting itself as the source of clarity.
### The wrong explanation can sound perfectly convincing
A conventional demonstration favoured the automatic feature. A presenter knew where to stand, which sequence to follow and what was supposed to happen. A visitor could stop differently, look across a room, return to an earlier object or accompany someone with another interest. Those variations were not misuse. They were part of the activity the guide was intended to support. Product judgement required observing the relationship between the physical visit and the digital explanation, rather than treating the demonstration's smoothness as a reliable account of everyday use.
The engineering sponsor reasonably wanted to improve the triggering behaviour. That work might have been useful, but it did not remove the immediate product question. What should the guide do while the available evidence remained insufficient? We could wait, ask the visitor to confirm, provide a way to choose an object or continue making an automatic assumption. Each choice had a cost. The important point was to make the choice explicitly. Calling the current behaviour seamless would not establish that it understood the visitor's intention or make the nine mismatches disappear.
Arun's field notes shifted my attention from the number of starts to the experience of being misdirected. If people promptly recognised a mismatch and recovered, the consequence would differ from a plausible explanation that they quietly accepted. We needed to understand both. That did not mean treating every observation as a statistical conclusion about all visitors. It meant using the observed problem to question the design's underlying claim. The feature was not only starting audio. It was asserting a relationship between a person, an object and a body of interpretation.
### Ask for the fact the system cannot establish
We spent ₹31 lakh on a visitor-confirmed identification path. The venue's visible object identifiers supported a deliberate selection before playback, with a route back when the chosen object did not match. That was not a statement that proximity technology could never be useful. It was a decision about the product we could responsibly offer with the evidence available. We accepted an additional action because it helped establish the object being considered. The interaction had a purpose a visitor could understand, rather than serving as a generic precaution inserted everywhere.
Confirmation also has design risks. An identifier may be hard to find, the object list may use unfamiliar language or the selection may be difficult to reverse. Moving responsibility onto the visitor is not automatically a better experience. The supplier still needs to make the action intelligible and workable. Venue colleagues must check the relationship between identifiers and objects, while product and research colleagues examine how visitors use it. A useful confirmation asks for something a person can reasonably establish. It should not demand technical knowledge that the system itself has failed to handle.
The distinction matters commercially because the most striking feature in a sales demonstration may not be the most trustworthy feature during a visit. I had to explain why we were removing an apparent advantage and spending money to do so. The case rested on the specific mismatches and their consequences, not on a broad preference for manual interaction. A board or executive team should be able to understand what the automatic behaviour assumes, how that assumption was tested and what the alternative changes. That is a more useful discussion than choosing between fashionable convenience and supposed caution.
### Measure recovery as part of the visit
The original trial remained a mixed result: fifteen intended starts and nine mismatches. We did not relabel it as proof of improved positioning once the product changed. A different interaction needs its own evidence. I would want subsequent research to examine whether visitors can identify an object, begin the relevant explanation, notice a mistaken choice and recover without abandoning the visit. Those are related but different questions. A single completion measure could hide the difference between a visitor who confidently chose the right object and one who reached an endpoint after considerable confusion.
There is also a limit to how much a guide should ask a visitor to do. A cultural visit is not a sequence of tasks set by the product team. People may look without listening, leave an explanation unfinished or return to something that interests them. The guide should support those choices rather than interpret every departure from its intended sequence as failure. Product research needs to remain close to the purpose of the visit. Otherwise, a team may optimise compliance with its interface while making the surrounding experience less enjoyable and less independent.
My own mistake was to equate removing an action with removing a difficulty. In this case the action had been doing useful work: it connected the explanation to the object. The automatic replacement concealed that work and the uncertainty it carried. I now ask which relationship a feature must establish before it can make its promise. If the system cannot establish it reliably, a small, meaningful pause may be the most respectful design choice. The measure of a good guide is not how quickly it takes control of attention. It is how well it helps a visitor decide where their attention belongs.