Retailers collect extensive customer information but often receive the analysis after the commercial decision has already been made. First Insight wants to replace that delay with dialogue. Its Ellis assistant lets merchandising, pricing, and planning teams question predictive consumer data in natural language.
After a three-month beta, the system is available to brands and retailers using First Insight’s platform. A user can ask whether six or nine products will perform better in a market, how removing a material affects appeal, or which price is most likely to maximise full-price sales.
The model is not valuable because it speaks. Its promise is to put established analysis inside the moment when a buyer, planner, or executive must choose.
Dialogue Reduces the Distance Between Insight and Action
First Insight has worked with retailers including Boden, Family Dollar, and Under Armour to analyse survey feedback, demand, price sensitivity, and product performance. Traditionally, those findings appear in dashboards or reports that specialist analysts interpret.
Ellis places a conversational layer over the predictive models. Teams can test scenarios during line review or early concept development instead of submitting a request and waiting. Faster access may prevent insight from losing relevance as demand, costs, or competitor activity change.
Natural language also broadens the user base. Senior executives and commercial teams can explore data without learning a reporting interface. That accessibility should not obscure methodology; users still need to know which consumers were surveyed, how uncertainty is expressed, and what information the answer excludes.
Conversation changes the analyst’s role rather than eliminating it. Specialists can spend less time retrieving standard cuts of data and more time designing studies, testing surprising results, and helping teams frame questions that distinguish genuine demand from a noisy sample.
Assortment and Pricing Decisions Offer Measurable Value
Under Armour has described using consumer data to refine assortments and pricing, reducing markdown exposure and supporting full-price sales. Boden uses customer insight to balance trend-led products with core items. These are high-value decisions because mistakes create inventory that must later be discounted.
Ellis is designed to answer questions about optimal price, likely sales rate, assortment size, segment preferences, and competitive position. The outcomes can be measured through forecast accuracy, sell-through, margin, and unsold stock rather than conversational engagement.
The system should complement, not flatten, merchandising judgement. Historical consumer responses may favour familiar products and understate an original concept. Commercial teams need room to pursue a strategic idea while understanding what the model predicts.
Teams should record when they override a recommendation and why. Comparing those decisions with later sales can reveal where human judgment adds value, where confidence was misplaced, and which product categories require a different model or research design.
Democratised Analytics Requires Shared Interpretation Rules
Gartner research cited in the source links wider analytics access with adoption and return, but warns that output must be grounded in robust data and interpreted correctly. A confident answer can travel quickly through an organisation even when the underlying sample is narrow.
Retailers should show source periods, market scope, confidence, and relevant comparisons alongside each response. Questions and answers need an audit trail so teams can revisit what informed a decision and learn when reality differs.
Access controls also matter. Pricing strategy and unreleased product concepts are commercially sensitive, while different regions should not automatically see every analysis. Conversational convenience must inherit the platform’s permissions.
First Insight competes with EDITED, DynamicAction, RetailNext, and other retail analytics providers. Conversational interfaces are becoming common; differentiation will come from data quality and whether recommendations improve decisions.
Ellis makes analytics easier to reach, not inherently correct. Its strongest role is as a fast route into evidence that teams can question, contextualise, and act upon while the decision is still open.
