Which signals are evaluated
These signals are evaluated as aggregated usage patterns – without individual tracking. The result: continuous feedback on where the catalogue and advisory logic can be optimised.
Identify product catalogue gaps
Analytics signals deliver concrete business insights: what customers really search for, where the assortment doesn't fit, and which needs the catalogue doesn't yet cover.
BUSINESS INSIGHTS
- Understand the intent behind search queries
- Reveal catalog gaps
- Identify customer needs & use cases
From these signals, three strategic insights emerge:
AGINITY AI condenses these signals into concrete recommendations: which categories should be expanded, which customer needs the assortment does not yet cover and where purchase barriers block conversion.
Use-case-based optimisation
Three concrete patterns from practice:
Users search for the term waterproof but find no IP-certified devices
High abandonment volume after this query
Kundensprache 'wasserdicht' wird mit IP-zertifizierten Produkten verknüpft
Zero results for this phrasing are eliminated
Users ask about compatibility with model X but receive no model-specific answer
Follow-up loop until abandonment
Range is extended with compatibility information for model X
Agent can deliver direct compatibility check
Agent responds too formally for consumer goods, too informally for B2B requests
Frequent follow-up questions about style, low close rate
Answer style rule configured by request context
Bounce Rate After Recommendation Decreases
Comparison: Without vs. With analytics signals
| Aspect | Without analytics signals | With analytics signals |
|---|---|---|
| Request Understanding | Static rule knowledge, no feedback from usage patterns | Query intent patterns continuously flow into the inference logic |
| Hit quality | Assortment gaps and purchase barriers remain invisible | Zero-result patterns and click dropoffs make catalogue weaknesses visible |
| Explainability | Answer texts based on initial configuration | Texts are adapted to frequent follow-up questions and misunderstandings |
| Consistency | Answer style depends on initial prompting rules | Formatting rules and tone specifications are adjusted with versioning |
| Handling unclear queries | Follow-up strategy statically predefined | Follow-up patterns optimised from abandonment analyses |
Expert knowledge trainable to customer requirements
The agent's consulting style is configurable – not as a one-time setup, but as continuously adjustable rule knowledge:
This rule knowledge is structured, versioned and protected by approval workflows.
Governance and Quality Assurance
Every change to rule knowledge goes through a controlled process:
This prevents uncontrolled quality drift and makes changes traceable.
FAQ
What data does AGINITY AI specifically analyse?
AGINITY AI captures search queries, drop-off points, zero-result patterns, click paths and recurring phrasing – evaluated as aggregated usage signals, not as personal tracking data.
How are catalogue gaps identified?
Suchanfragen ohne Ergebnis (Zero Results) und häufige Abbrüche machen sichtbar, welche Produkte Kunden suchen aber nicht finden – ein direktes Signal dafür, wo das Sortiment gezielt ausgebaut werden sollte.
What does 'trainable expert knowledge' mean?
The agent's response and advisory style can be adjusted through configurable rules: which criteria come first, which phrasing is preferred, which reasoning chains the agent should use. This rule knowledge is versioned and can only be changed by authorised persons.
How does a retailer control the agent's response style?
Through configuration rules: tone specifications (informal/formal, brief/advisory/technical), order of recommendation criteria, do/don't lists for phrasing and product-specific reasoning logic. Changes go through an approval workflow.
How is quality assured when rule knowledge is changed?
Every change to rule knowledge is versioned. An approval workflow ensures that only authorised roles can activate rule changes. An audit log documents who changed which rule and when.
Can the agent also support the checkout step?
Yes. The sales agent can actively guide through to the checkout step at the end of the advisory journey – through hints about availability, delivery time, compatibility or configurator options. This is a configurable step in the advisory flow.
