Agent vs Human Funnel

DIFFERENTIATOR· sample data · production shape

Side-by-side comparison of how agents vs humans move through the booking funnel. The shape, not the numbers, is what matters: agents shop less, convert faster, and need different rate-shopping affordances.

Agent sessionsi

180k

Human sessionsi

820k

Agent end-to-end conv.i

5.25%

Agent lift vs humani

4.77×

Human funnel · search → fulfilled

step-to-step conversion in %

Searches820,000

100.0% of top

Look377,200 (46.0%)

46.0% of top

Shop120,704 (32.0%)

14.7% of top

Book attempt9,897 (8.2%)

1.2% of top

Confirmed9,303 (94.0%)

1.1% of top

Fulfilled9,023 (97.0%)

1.1% of top

Agent funnel · search → fulfilled

same steps, different shape — agents shop less, decide faster

Searches180,000

100.0% of top

Look127,800 (71.0%)

71.0% of top

Shop74,124 (58.0%)

41.2% of top

Book attempt10,822 (14.6%)

6.0% of top

Confirmed9,848 (91.0%)

5.5% of top

Fulfilled9,454 (96.0%)

5.3% of top

Step-by-step conversion · agent vs human

% from previous step · sample data

Agent traffic by source · last 30 days

%-share of MCP sessions tagged by agent-id

Anthropic Claude38%

OpenAI ChatGPT32%

Google Gemini14%

Custom MCP integrations11%

Other / local LLMs5%

Why this is the moat metricDIFFERENTIATOR

Agents convert differently from humans: lower window-shopping, faster decisions, retry-on-error patterns. Knowing this lets the platform price agent traffic differently and ship API features (deterministic responses, structured errors) that lean into agent strengths. The chart above is the single highest-leverage thing a Head of Data here could ship in year one.

Sample data. Agent/human conversion numbers are synthetic. Production version needs a session-source tag at booking time + MCP middleware to attribute sessions to agent vs human.