CUSTOMER EXPERIENCE & INSIGHT ARCHITECTURE OPERATING MANUAL Customer insight starts where the score stops A CSAT score tells you what happened; insight explains why, where, who owns it, and what changes next. Practical VoC & CX playbooks for Indian enterprise, retail, and scaling operators
OPERATING DEFINITION Insight is a decision-ready sentence THEME Late pickup PROOF Volume, quotes, or revenue at risk PLACE Store, SKU family, or journey step OWNER Someone who can change the process Research, NPS, and tickets are inputs; insight is the output. 02
Service AI closes contacts; insight AI reduces the reason they exist AI in customer service Bot or copilot that handles the contact. AI customer experience Listening, journey design, personalisation, and the closed loop after contact. A customer experience AI platform may listen through VoC and act through agents, routing, or offers. Core rule: Do not buy a helpdesk and call it insight, or buy a survey suite and call it service. 03
END-TO-END WORKFLOW The insight pipeline joins language to operations 01 Collect surveys, reviews, chats, email, social, and call recordings. 02 Transcribe and translate Hindi, Hinglish, Tamil, and other voice inputs. 03 Theme and score sentiment by aspect, not one emoji for the whole text. 04 Link evidence to order ID, store, SKU, and agent. 05 Rank by volume × severity × money, then alert and assign. 06 Check whether the theme falls next month. 04
Buy the job, not the loudest platform name Listen-first Qualtrics, Medallia, Sprinklr VoC, QuestionPro-class. Analyse-first Chattermill, Thematic, specialist review AI. Service-first Zendesk, Freshworks, Gorgias for Shopify retail. Speech-first Contact-centre analytics across call recordings. Vendor lists vary. The useful frame is gather, analyse, act. 05
India’s platform choice follows operating context OPERATING CONTEXT PLATFORMS FIT Enterprise XM Qualtrics and Medallia Quote-only and longer implementation cycles. Social and digital command Sprinklr Enterprise modules. Contact-centre speech Verint-class and CCaaS analytics Recordings required. Mid-market tickets Freshdesk Freddy, Zoho Desk + Analytics, Zendesk AI INR, GST, and IST support matter. Feedback-only analysis Chattermill or Thematic-class Bring your own export. Reviews and maps Google review tools plus a sheet Until volume explodes. 06
OMNICHANNEL RETAIL Retail insight fails when store and site stay separate Ingest store exit comments, WhatsApp, 1800 lines, marketplace reviews, and return reasons. Tag by banner, city, fulfilment type, and BOPIS versus courier. Split product themes from operations themes: fit and fabric versus a late rack. Use return text as the honest survey you already pay for. Close the loop: a spike in “alteration ready” calls may call for a better SMS, not more agents. Combine telephony, helpdesk, and Sprinklr when the operating model needs the blend: that blend is the market. 07
Calls are insight when they change the operating decision Call transcripts reveal reasons, emotion, and repeat friction, not just handle time. For DTC teams, callers may still reach a person and convert on that call. If “where is my order” dominates the night shift, fix tracking or fulfilment before adding minutes. Voice agents create high-volume signal; they are not automatically a VoC platform. 08
SMBs can start with a monthly theme loop Freshdesk or Zoho Desk reports plus a monthly theme export. Native review reply tools for Google and marketplaces. Looker Studio on a reviews sheet, or Shopify helpdesk AI for D2C. Use a junior human to tag 50 comments a week until volume justifies a suite. Skip the seven-module Enterprise XM contract when one literate founder can own the loop. 09
The weekly VoC stand-up needs one signal and one owner NPS comment: theme + score → late BOPIS → SMS when rack is ready. Ticket: intent + repeat → WISMO three times → tracking page. Call: reason + emotion → order-status rage → contain and fix process. Review: aspect sentiment → fit versus fabric → size chart. Return text: root reason → photo does not match product → improve PDP photo. Social: spike → dye-lot complaint → pause SKU. 10
