How consultative retail conversations actually run — which behaviours occur, how often, and where the widest coachable gaps sit between the top quartile and the median.
Built on a corpus of real, consented, in-person selling conversations. Nothing in it is staged, and the methodology is set out before the findings rather than in an appendix.
Methodology
The corpus consists of real conversations captured at designated counters in participating stores, under visible signage, a QR-accessible plain-language notice, and a verbal disclosure inside the greeting. No conversation was staged or solicited for research purposes.
Behaviours were defined before any scoring took place, derived from top-quartile conversations where the quartile was selected on point-of-sale outcome rather than on managerial reputation. Each definition carries an inclusion and an exclusion boundary, and an eligibility test so that conversations which never called for the behaviour are excluded rather than counted as failures.
Scoring accuracy was validated against human double-scoring and is reported separately by language and accent group. Definitions that produced repeated disagreement were rewritten. No voiceprints or voice embeddings were created at any stage, including model training; attribution was by capturing seat and roster shift.
Key findings
01
The gap between the best and the median is behavioural, not personal
Across every vertical in the corpus, top-quartile associates were separated from the median by how often they performed a small number of specific, nameable moves — not by charisma, product knowledge or tenure. The same six to ten behaviours accounted for most of the observable difference.
02
The largest coachable gap is almost never the greeting
Greeting compliance was high nearly everywhere, because it is the behaviour every existing training and mystery-shopping programme measures. The widest gaps sat later in the conversation: budget framing, the attach conversation, and scheduling a follow-up before the customer left.
03
Follow-up is the most neglected high-value behaviour
Setting a specific next contact before the customer stood up was among the least frequently performed behaviours in the corpus, and among the most strongly associated with top-quartile membership. It costs nothing, requires no discount and is rarely trained.
04
Code-switching is the normal case, not an edge case
A large share of conversations moved between languages inside single sentences. Any measurement system that treats a language switch as a new conversation, or that scores only the dominant language, will systematically misread the floor.
05
Behaviour rates vary more within a store than between stores
The spread between the strongest and weakest associate inside one door was consistently wider than the spread between store averages. This is the empirical reason store-versus-store comparison cannot isolate a coaching effect, and why cohorts are built inside a single store.
Gated, and sent by a person rather than by an automated download.
We use cookies necessary to run this site. Analytics cookies help us see which pages get read, and they load only if you accept. No advertising or cross-site tracking, ever. See our Privacy Policy.