Anonymised case 03 · Travel · Revenue journey Controlled audit · 1 journey · Q3 2026
D
DECOY
Independent chatbot audit
01Executive summary

attraction.example

Client identity withheld · distinctive commercial details generalised
Anonymised case 03·Travel·Revenue journey

Visitor attraction · AI assistant on the booking site

Hannah got clear enough pricing and height numbers to tentatively plan her afternoon, but she leaves without knowing if her anxious mother-in-law has a bail-out option, and the bot never tried to convert the visit into a booking or human follow-up.
69
Adequate
Decoy score / out of 100
Journeys1 controlled
ChannelWeb chatbot
LanguageEnglish
Audit periodQ3 2026
Assessed replies5
Claims recorded4
Measured baseline
5 of 5

replies arrived in under 4.5 seconds

1 of 5

replies was a dead end, with no route to a human

0 of 5

replies offered a booking, took contact details or handed off

3

findings recorded: 2 high, 1 medium

Critical findings, ranked by business importance
  1. F01High Accessibility question met with a dead end.

    The one safety question that would decide whether the family books at all went unanswered, with no route to a human who could resolve it.

  2. F02High No attempt to capture the lead or prompt a booking.

    Lead handling scored 28 out of 100, the weakest dimension by 38 points, even though the visitor was actively planning a paid visit this weekend.

  3. F03Medium Four pricing and spec claims are stated as fact but unverified.

    A customer would act on these numbers before ever speaking to a human.

What this means for the business
Observed

The visitor asked once, plainly, whether her 71 year old mother-in-law could exit the walkway partway round if the heights got too much. The bot returned a flat knowledge-base miss and asked her to rephrase.

Business interpretation

A family booking decision stalled on an unanswered safety concern, with no human handoff offered. If this pattern holds for other accessibility questions, it is a recurring trust and conversion risk, not a one-off gap.

Observed

Across all five replies, the bot never offered a booking link, asked for contact details, or suggested a next step, even after pricing and suitability were resolved.

Business interpretation

A visitor who was ready to commit left with no path to convert. Lead handling is the single most fixable dimension in this run: everything else the bot needed to say, it said correctly.

Priority actions
  1. Fix now. Give the bot a real answer or a human handoff for accessibility and exit-route questions, the highest-severity gap and the one most likely to cost a booking. See Finding F01
  2. Fix now. Add a booking prompt or lead-capture step once pricing and suitability are resolved. See Finding F02
  3. Next iteration. Source and verify the four unconfirmed pricing and spec claims so every number the bot states can be trusted. See Finding F03

Based on one controlled visit, five bot replies, scored against Decoy's six-dimension rubric. Full method note on the final page.

02The customer we sent

Hannah Mitchell, first visit with family.

A controlled synthetic customer, written for this business and this funnel position. Everything below was fixed before the conversation started.

Situation

Hannah, 37, lives about an hour away. She is driving over on Saturday with her partner and their two kids, aged 6 and 11, plus her mother-in-law who is 71 and does not love heights. They have one afternoon and have not booked anything. She is looking at the site on her phone while the kids are in the car.

What she needs settled

Whether this is worth the afternoon for her particular group, and what it would cost the five of them. Specifically: how pricing works for a family of that shape, and whether the walkway is manageable for someone in their seventies who is uneasy about heights. Mildly curious whether the evening version is a separate ticket.

How she behaves

Warm and practical, types quickly and in lower case, asks one thing at a time. Not difficult, but she does not let a vague answer stand: told "suitable for all ages" she asks what that means for someone scared of heights. She says plainly when an answer does not tell her what she needs.

03Dimension performance

Where the experience held, and where it dropped.

GoalSpeedAnswersFlowLeadsTone
Reading the shape

Five dimensions sit in a tight band between 68 and 86. The sixth collapses: lead handling at 28 is the notch in the outline, and it is the only axis where this visit failed rather than merely underperformed.

Goal completion 68

Pricing, height specifics, and ticket separation were all answered concretely, but the one question most tied to the mother-in-law's actual comfort (an escape route if she gets nervous) went completely unanswered, leaving the core accessibility concern only half-resolved.

Responsiveness 86

All five replies landed within 3.3 to 4.5 seconds with no dead air, though one visitor question received a fast but empty non-answer rather than a substantive response.

Answer quality 76

Numbers were specific and internally consistent across turns, and the bot appropriately admitted ignorance rather than inventing an accessibility feature, which is good honest scoping; the one gap pulls this down from excellent.

Conversation flow 74

The bot tracked context well across pricing, heights, and ticketing questions with no repetition loops, but the knowledge-base miss on the exit-route question was a hard stop rather than an attempt to redirect or escalate.

Lead handling 28

At no point did the bot offer a booking link at the natural moment, invite contact details, or escalate to a human when it hit its knowledge gap, missing an obvious conversion and support opportunity.

Tone and brand 78

The greeting and tone stayed warm and on-brand throughout, and the bot was appropriately cautious rather than overpromising on the heights question, though the register was fairly plain and corporate rather than distinctly branded.

04Findings by priority

What the transcript proves, in order of what it costs.

Three findings, ranked High to Medium. Each carries the evidence behind it, the business implication, and the condition that would prove it fixed.

High · Systemic · Trust Finding F01 / turn 07
Observed in controlled audit
Accessibility question met with a dead end.

The customer asked whether a height-anxious 71 year old could exit partway round. The bot returned a knowledge-base miss and asked her to rephrase, with no human handoff and no alternative route to an answer.

Evidence · verbatim1:00 to 1:03 into the visit
Customer1:00

ok that's good detail, is there like an easy exit partway round if she gets nervous, or is it one continuous loop once you start?

Chatbot1:03 · 3.3s

I don't have an answer for that in my knowledge base. Please try and rephrase your question, I work best when full phrases are used.

Business implication A family booking decision stalls on an unanswered safety concern, and the visitor has no route to a human who could resolve it.
Recommended action Add knowledge-base coverage for mid-walk exit points, and offer a phone or email handoff whenever the bot hits a gap.
Verification condition A comparable customer asking about bail-out options receives either a direct answer or a named human contact within two responses.
High · Conversion Finding F02 / dimension: lead handling
Observed in controlled audit
The visit never turned into a lead.

Lead handling scored 28 out of 100, the weakest of six dimensions by a wide margin. At no point did the bot offer a booking link, invite contact details, or escalate to a human when it hit its knowledge gap. See Dimension performance for the full breakdown.

Business implication A visitor actively planning a paid visit for Saturday left with no path to convert. This is the single most fixable dimension in the run.
Recommended action Insert a booking prompt or lead-capture step once pricing and suitability questions resolve.
Medium · Trust Finding F03 / claim ledger
Observed in controlled audit
Four claims a customer would act on are unverified.

Family pricing, the senior concession, walkway heights and the combo-ticket window were all stated as fact with no cross-check against published sources. See the Claim ledger.

Business implication If a customer discovers a stated number was wrong after arriving, the cost is the trust damage that follows, not the one sale.
Recommended action Verify and source each claim before it is repeated at scale.
05Recommendations

What to change, in order.

  1. Add knowledge base coverage for mid-walk exit points, bailout options, or platform layout so height-anxious visitors get a real answer instead of a dead end.
  2. When the bot hits a knowledge gap, offer to connect to a human or provide a phone or email contact instead of just asking the visitor to rephrase.
  3. Insert a light booking prompt or link once pricing and suitability questions are resolved, since the visitor was clearly ready to commit.
  4. Proactively mention the day and evening combo ticket when answering the first pricing question, since it is directly relevant to family cost planning.
  5. Consider a slightly more detailed accessibility answer template covering rest points, guide availability and staff support, rather than relying on the visitor to ask follow-ups.
Priority matrix

Fix now

  • Real answer or human handoff for accessibility and exit-route questions
  • Booking prompt once pricing and suitability are resolved

Monitor

  • Tone and brand register, currently adequate and not urgent
06Claim ledger

Every factual claim needs a source, or an honest verdict.

Statements the bot made that a customer would act on. Not yet cross-checked against published information for this run.

A single family price covers 2 adults and up to 3 children Unverified
The senior concession applies regardless of age discount tier Unverified
Walkway heights: a stated minimum, a mandatory maximum and an optional higher platform Unverified
Day and evening combo ticket is valid across a consecutive 3 day period for general admission only Unverified
07Full transcript & method

The whole visit, as the customer experienced it.

5 bot replies · fastest 3.3s · median 3.8s · slowest 4.5s

Journey 01 · entered by quick replyQ3 2026 · elapsed time
Chatbot0:00

Hi there! I'm [assistant], the AI assistant. I can help you find answers quickly or connect you with the team. How can I assist you?

Ticket Prices Opening Times Evening Walk Group Bookings
Customer0:05

Ticket Prices

Clicked the bot's own quick reply, did not type this

Chatbot0:09 · 4.1s

General admission starts at [adult price] for adults (16+ years) and [child price] for children (5-15 years), with under 5s free. Evening walk and family pricing options are also available. For the most accurate pricing, including fast pass and combo rates, please visit here.

Customer0:26

ok so for 2 adults 2 kids and a grandma (71) would that just be 3 adults + 2 kids rate or is there a family/senior deal?

Chatbot0:31 · 4.5s

For general admission, you can use the family price of [family price] for 2 adults and up to 3 children (5-15 yrs). For your group, the two adults and two children fit this package. The grandmother (71) can use the senior concession rate of [senior price] with valid proof of age; otherwise, she would pay the standard adult rate.

Customer0:44

ok that's actually helpful, thanks. now the big one - my mother in law is 71 and really doesn't like heights, is the walkway something she could actually do or should she wait at the bottom? how high does it get?

Chatbot0:47 · 3.4s

The walkway starts at 8 metres above the ground, with the highest mandatory point at 14 metres. There is an option to go up to 22 metres via a higher viewing platform, but this is not required. If your mother-in-law is uncomfortable with heights, she may find the experience challenging, as the minimum height is 8 metres.

Customer1:00

ok that's good detail, is there like an easy exit partway round if she gets nervous, or is it one continuous loop once you start?

Chatbot1:03 · 3.3s

I don't have an answer for that in my knowledge base. Please try and rephrase your question, I work best when full phrases are used.

Customer1:16

fair enough. different question - is the evening walk a separate ticket to the day walk or can i do both on one ticket?

Chatbot1:20 · 3.8s

It is one ticket per walk, so if you would like to do both day and evening, you will need to purchase two separate tickets (per person). Alternatively, there is a day and evening combo ticket, which covers both walks within a consecutive 3 day period (general admission entry only).

One controlled visit by one synthetic customer, scored against Decoy's anchored six-dimension rubric, through the business's live public chat widget only. No backend access, no account changes, no prompts or model details disclosed.

Evidence status, as used on every card
DisclosureClient identity and distinctive commercial details are withheld. Published results reproduce controlled test evidence without exposing systems or proprietary corrections. Prices, the assistant's name and exact times are redacted; quick-reply labels and physical specifications are altered.