Failed-Search and Helpdesk Query Analysis¶

Behavioral diagnostic — instantiates Latent Affordance Surfacing

Mines the searches that returned nothing and the support tickets asking for things that already exist, reading them as recorded evidence of capabilities users wanted but couldn't find.

Where an audit predicts where users will fail to find an affordance, Failed-Search and Helpdesk Query Analysis reads where they already did. It mines zero-result searches, null queries, and support contacts — the logged residue of users reaching for something — and matches those intents against the list of what the system can actually do, separating "they wanted a thing that doesn't exist" from "they wanted a thing that exists but couldn't find." Its defining move is that it is behavioral and backward-looking: the gap doesn't have to be guessed, because it announces itself in the user's own words every time a search comes back empty or a ticket asks how to do the already-possible. Unmet intent is data the system is collecting whether or not anyone reads it.

Example¶

An online retailer lets customers cancel an order themselves from account settings — the capability genuinely ships. Yet site-search for "cancel order" returns product listings, and the call center logs a steady stream of "how do I cancel" contacts. The analysis pulls the zero- and low-relevance search terms and clusters the support tickets by reason, then matches each cluster against the capability list. "Cancel an order" surfaces as a top failed intent for a capability that already exists — a clear effective-option-set gap, evidenced not by an inspector's judgment but by thousands of customers trying and failing in their own words.

It ranks that intent by volume and cost-to-serve, and — because it is a standing loop, not a one-off — it watches what happens after a fix ships. When search gains a "cancel order" synonym that routes to the self-service flow, the null-search count and the call volume for that intent both fall; the loop reads that decline as confirmation the gap actually closed, rather than assuming it did.^[1]

How it works¶

Ingest the traces of reaching. Pull zero-result and low-click searches, internal search logs, chat transcripts, and support-contact reasons — the places users leave evidence of an intent the surface didn't satisfy.
Cluster into intents. Group the raw queries and tickets into the underlying things users were trying to do, so volume accrues to intents rather than to phrasings.
Match against what exists, and rank. Classify each intent as exists-but-unfound (a surfacing gap) or doesn't-exist (a roadmap signal), and rank the surfacing gaps by volume and cost-to-serve.
Close the loop. Track each intent's failed-query and ticket volume after a fix ships, so a drop confirms the gap closed.

The diagnostic is distinguished by being passive, continuous, and in the user's own words: it finds gaps from behavior that already happened rather than from inspection or a designed test.

Tuning parameters¶

Source breadth — site search only, or internal search, chat, tickets, and call-reason codes too. Wider sourcing sees more intent but mixes channels with different noise.
Clustering granularity — how finely queries are grouped into intents. Too coarse merges distinct needs; too fine scatters volume and hides the pattern.
Exists-vs-missing classifier — how intents are matched to real capabilities. A strict match under-counts unfound features; a loose one mislabels genuine gaps as surfacing problems.
Ranking weights — volume versus support cost versus downstream revenue. The weighting decides which gaps rise to the top of the worklist.
Loop cadence — a one-off study versus a standing dashboard. Continuous cadence catches new gaps and confirms fixes; one-off is cheaper but blind to decay.

When it helps, and when it misleads¶

Its strength is finding gaps from the user's own words, including ones no inspector would predict, and it comes with impact built in — volume quantifies which gaps matter, and the post-fix decline confirms the fix worked rather than merely shipping. It is the natural complement to a forward audit, and often the more persuasive of the two because the evidence is real behavior.

Its central blind spot is survivorship: it can only see users who tried and were logged. The person who never suspected a capability existed leaves no failed search and files no ticket, so a fully-hidden affordance can be invisible to this method precisely because it is invisible to the user — the deepest gaps generate the least evidence. Noisy clustering can also conflate distinct intents, and the classic misuse is to read every failed search as a missing feature (build it) when it is a surfacing gap (label or route it), or the reverse. The discipline is a careful exists-vs-missing match, and pairing the method with a forward inspection audit to catch the silent gaps behavior can't reveal.

How it implements the components¶

Failed-Search and Helpdesk Query Analysis realizes the behavioral-feedback side of the archetype — reading logged intent as evidence of the gap:

desire_path_feedback_capture — its raw material: failed searches and support contacts are the desire paths users wore into the system while reaching for something the surface didn't offer.
discovery_measurement_loop — it quantifies each gap by volume and tracks the post-fix decline, turning surfacing into a measured, closeable loop.
effective_option_set_gap — matching intents against real capabilities yields the gap from the user's side: capabilities that exist but went unfound.

The forward inventory and sampling map, the cue design, and eligibility — latent_capability_inventory, user_sampling_path_map, affordance_signifier_design, intended_user_and_eligibility_model — belong to Affordance Visibility Audit, Contextual Hint or Tooltip, and Eligibility or Rights Notice; this diagnostic reads gaps from behavior, it neither inventories the system nor authors the fix.

Instantiates: Latent Affordance Surfacing — it surfaces gaps from the behavioral evidence users leave when a surface fails them.
Consumes: Affordance Visibility Audit — the audit's capability inventory is what lets this analysis tell exists-but-unfound from genuinely missing.
Sibling mechanisms: Affordance Visibility Audit · Search and Autocomplete Suggestion · First-Attempt Discovery Test · Call-to-Action Placement Test

Notes¶

This diagnostic and the Affordance Visibility Audit are complements, not substitutes: failed-search analysis finds the loud gaps, where users tried and were logged, while the audit finds the silent ones, where users never knew to try. Running only the behavioral method leaves the most-hidden affordances — the ones that generate no evidence — permanently in the dark, which is why the two belong on the same worklist.

References¶

[1] A desire path (or desire line) is the worn trail that appears where people actually walk, revealing a route the built environment failed to provide. Failed searches and "how do I…" tickets are the digital equivalent: a recorded trace of where users tried to go, and therefore of the affordance the surface should have made findable. ↩