The category, in our own words.

A statutory requirement, in roughly a dozen US states, that every party to a recorded conversation consent to the recording. Several of those states are large retail markets, including California, Florida, Illinois, Pennsylvania, Massachusetts and Washington, so treating them as unavailable is usually an overcorrection. Our design obtains customer consent at the counter through visible signage, a plain-language notice and a scripted verbal disclosure inside the greeting.

The part of a consultative sale in which a complementary item, service or protection is introduced — a chain with a pendant, a strap with a watch, coverage with a frame. It is a conversation rather than a prompt at the till, and it is one of the most reliably coachable behaviours we measure because the top quartile performs it two to three times more often than the median. What "attach" contains is defined per vertical.

A specific, observable move inside a selling conversation that either happened or did not. "Asked what the purchase is for, before presenting a product" is a behaviour; "built rapport" is not, because two careful people can disagree about whether it occurred. Every behaviour we score carries an inclusion boundary and an exclusion boundary so that the near-miss case is decided in advance.

The set of behaviours defined for a vertical, derived from conversations recorded by that vertical’s own top performers rather than from a generic selling framework. A working library holds six to ten behaviours, because beyond that a store manager cannot retain them and a daily prompt has nothing to focus on. Libraries are meant to be edited as assortment, season and customer change.

The structured method by which a behaviour library is produced: select the top quartile by point-of-sale outcome, listen across their conversations for recurring moves, write each move as a scoreable definition, then rank the definitions by the gap between top quartile and store median. The taxonomy captures the move, never the words, because scripts do not transfer across languages or personalities. It is the artifact that makes coaching priority a matter of evidence rather than taste.

The recording of audio at a designated selling position by a fixed device with an always-visible indicator. Capture cannot run with the indicator unlit; that constraint is enforced in signed firmware rather than in configuration. Nothing runs on an associate’s personal phone and there is no camera or image sensor in any Gaincraft device.

The proportion of eligible consultative conversations in a store that were actually captured and scored in a period, published weekly. Coverage is the honest denominator behind every other number: a behaviour rate measured across forty percent of conversations is a different claim to one measured across ninety. We publish coverage per store precisely so that a thin week cannot be quietly excluded from a favourable readout.

A physical area of a store designated for capture during scoping, recorded on a zone map signed by the retailer. Areas excluded by default include fitting-room interiors, optical exam rooms, finance and insurance offices, and all staff areas. Where a retailer’s position is narrower than our proposal, we take the narrower position.

A short excerpt, typically fifteen to forty seconds, of a real conversation in which a behaviour was performed well, used as teaching material. Clips come from the retailer’s own floor, which makes them more credible than produced training video and free to create. Clip use is governed by the same consent and retention rules as any other captured audio.

The per-seat, per-month component of pricing, covering the daily prompt for each associate who receives one. It is funded from Learning or Field Education, which is already budgeted per head. The two-component structure exists because retailers allocate money per store and per person, not in a single line.

The full two-sided design by which consent is obtained and evidenced: four visible layers for the customer, and a separate framework inside the employment relationship for the associate. It is called an architecture rather than a policy because parts of it are physical — signage, an indicator light, signed firmware — and cannot be changed by a configuration flag.

The sixty days of a pilot during which measurement is treated as evidence, as distinct from the thirty-day setup phase that precedes them. Setup is not counted, because scoring a store while hardware placement, taxonomy tuning and floor announcement are still settling produces a baseline nobody should trust. The counted period begins when coverage and cohort assignment are both stable and written down.

The measurement of real, consultative, in-person selling conversations at the point where they happen, rather than of calls, chats or staged visits. It differs from contact-centre conversation intelligence in that the conversation is face to face, unscripted, frequently multilingual, and takes place in a semi-public commercial space, which changes both the consent design and the acoustics. The purpose is coaching an individual, not auditing a store.

The two-minute coaching message an associate receives each morning, containing yesterday’s behaviour result and, in most cases, a short clip of a colleague performing the behaviour well. It is private to the associate and their manager, and it gives before it asks — a prompt that only reports a shortfall is a daily performance warning, and people correctly refuse it.

The behaviour scores, gap analyses and coaching prompts produced from a conversation, as distinct from the conversation itself. Derived insight persists for the contract term; audio is auto-deleted at the end of the retention window the retailer configures. The distinction is what allows a thirty-day, seven-day or twenty-four-hour retention posture without losing the measurement.

Funding a Gaincraft programme by reallocating budget already committed to floor quality assurance, rather than by creating a new line item. The reference figure is roughly two thousand one hundred and sixty dollars per store per year for two mystery shops a month at about ninety dollars a shop. Where the platform fee does not fit inside a retailer’s existing spend, we say so.

An associate’s formal challenge to a behaviour score, with the underlying audio available to both the associate and the manager and a human review deciding it. An overturned score is corrected in the record rather than footnoted. Dispute rate is a quality signal about the definition, not about the associate: a behaviour that generates repeated disputes is badly written and gets rewritten.

The test applied before a behaviour is scored, establishing whether the behaviour was appropriate in that conversation at all. A follow-up is not expected on a gift purchase collected immediately; a second pair is not expected in a warranty visit. Without eligibility, behaviour scores punish associates for conversations that did not call for the behaviour, and the floor stops trusting the number.

The comparison of a behaviour’s performance rate in the top quartile against the store median, with the outcome difference attached. It converts a list of good practices into a ranked decision about what to coach first, in which store. The ranking rarely matches what management would have guessed.

A measurement design in which coached and capture-only associates are compared inside the same store, so that footfall, catchment, manager, inventory and promotional calendar are shared rather than assumed away. Groups are swapped at the midpoint to test whether the effect follows the coaching or the people. It is the only design we run, because store-versus-store comparison cannot separate a coaching effect from the accident of two different doors.

The guarantee that Gaincraft capture cannot run without a lit indicator, enforced by hardware and cryptographically signed firmware that causes a device to refuse an unsigned image. It is not a policy commitment we could quietly revise in a later release. We decline requests for an evaluation mode with the indicator off, and we lose deals over it.

The posture in most US states, where the consent of a single party to a conversation is sufficient for it to be recorded lawfully. It simplifies the legal review but does not change our design, because the signage, notice and verbal disclosure ship in every state. Deploying a covert-capable programme in a one-party state is legally easier and operationally fatal, since the floor will not cooperate with a system it cannot see.

The per-store, per-year component of pricing, covering capture, hardware, playbooks, scoring and support. It is funded from Store Operations, where it displaces the mystery shopping and retail audit line — existing outsourced vendor spend that measures a sampled, delayed version of the same thing.

The period after which captured audio is automatically deleted — thirty days by default, with seven-day and twenty-four-hour options. Deletion is automatic rather than a manual step someone has to remember. Personal identifiers and payment detail are redacted at ingestion, before any human or model can open a transcript.

The weekly view of behaviour performance for a store, district or region, arranged so that the weakest behaviour per store is legible in a single column. It is deliberately not a live feed: the unit of feedback is yesterday, which removes the possibility of supervision by notification. Whatever a manager can see about an associate, that associate can see about themselves.

Attributing a conversation to an associate by combining the capturing position with the roster assignment for that shift, rather than by recognising the speaker’s voice. It is deliberately less precise than voice matching — a covering colleague or a manager stepping in can be misattributed — and every attribution is disputable, with the record corrected rather than annotated. In exchange, no biometric identifier is created at any point.

The eyewear-specific behaviour of introducing a second pair — sunwear, a reading pair, an occupational lens — inside the same consultation. It is a defined behaviour with an eligibility test, so a conversation where a second pair was never appropriate does not count against the associate. Measuring eligibility rather than raw offer rate is what keeps the score coachable rather than resented.

The moment an associate moves a customer from an initially stated budget to a higher-value option by reframing what the money is for, rather than by pressure. It is scored on whether the reframing happened and whether a budget range was established first, never on whether the customer spent more. Coaching it without the budget-framing prerequisite reliably produces worse conversations.

The practice of finding, inside a retailer’s own conversation data, which behaviours separate the top quartile from the median, and how much each gap is worth. It replaces the assumption that the best sellers have an unteachable gift with a list of specific moves they make more often. The output is both a coaching priority order and a library of clips.

A stored mathematical representation of the physical characteristics of a person’s voice, used to identify that person in future audio. Voiceprints are treated as biometric identifiers under Illinois BIPA and analogous statutes, carrying specific consent requirements and, in Illinois, a private right of action. Gaincraft creates and stores none, for customers or associates, at any stage including model training.