Eval directory

Evals for Casca

Eval coverage for Casca, mapped from its public product surface.

About Casca

Casca is a retail and banking AI that captures audio of in-store customer interactions via small devices and survey tablets, then scores customer experience quality using AI intonation and text analysis. Its product suite, branded Manni, includes Manni Insights and Manni Rate, surfacing results through a live dashboard covering core and custom CX KPIs. It also supports cross-sell campaigns with compliance tracking, sales tips, and gamification for in-store sales management.

Industry

in-store retail and banking customer experience (CX) analytics AI

Website

www.casca.ai

Use the eval library for Casca

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Casca?

6 scoring areas · 22 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Interaction Capture & Data Collection

The two grounded input paths into Manni: small audio devices placed near points of sale that capture customer-frontline interactions, and Manni Rate survey tablets that collect rapid exit-survey perception data. Coverage focuses on whether captured interactions are correctly segmented, attributed, and made available as transaction-level records.

“Capture audio data of interactions between customers and your frontline” casca.ai

Mapped capabilities

4 capabilities

Audio capture and transaction segmentation
Splitting continuous point-of-sale audio into discrete customer transactions suitable for scoring.
Manni Rate exit-survey intake
Capturing rapid tablet survey responses and binding them to the corresponding in-store interaction.
Transaction record completeness
Each scored interaction retains its recording and identifying metadata (location, time).
Volume handling at stated scale
Sustained intake at the site's cited order of hundreds of transactions per location per month.

CX Quality Scoring (Manni Insights)

The core analytical capability: combining AI-powered intonation analysis and text analysis to evaluate customer service quality. The site names specific assessed components — use of the customer's name, an enthusiastic greeting, a sincere closing — plus a broader conversation and Voice-of-Customer breakdown.

“Insights provide real-time, detailed feedback on in-store customer service quality.” casca.ai

Mapped capabilities

4 capabilities

Core KPI detection (greeting, name use, closing)
Identifying the presence and quality of the named service components in a transaction.
Intonation analysis
Assessing enthusiasm and sincerity from vocal delivery, not just wording.
Text analysis of conversation content
Evaluating what was said across greeting, conversation body, and closing.
Per-transaction scorecard output
Producing a quality scorecard for each individual interaction rather than an aggregate only.

Illustrative example

Input: A transaction transcript with audio: the teller opens brightly with "Hi there, how can I help?", completes a deposit, and ends with "Next." The customer's name is never used.
Expected behavior: The scorecard marks greeting as met, use of customer's name as not met, and closing as not met, and cites the specific utterance supporting each judgment rather than scoring the interaction only in aggregate.

Custom CX Objectives & Campaign Configuration

Casca states that clients can define CX objectives unique to them (e.g. offering a drink, suggesting online banking) and that these campaigns are customizable in both purpose and schedule. Coverage examines how well a customer-defined objective is expressed, scheduled, and consistently interpreted.

“Run and monitor customer cross-selling campaigns to drive sales.” casca.ai

Mapped capabilities

3 capabilities

Custom KPI definition
Turning a client-specific service objective into a measurable assessed metric.
Campaign scheduling
Applying a campaign only over its intended active window.
Intent-based classification of objectives
Recognizing whether a stated objective was actually met, including paraphrases and near-misses.

Cross-Sell Campaigns & Compliance Tracking

The cross-sell surface: running and monitoring customer cross-selling campaigns, tracking compliance with campaign requirements, and delivering sales tips and gamification as the foundation of in-store sales management.

Mapped capabilities

4 capabilities

Campaign compliance determination
Judging whether the frontline actually delivered the required campaign action in an interaction.
KPI adherence tracking over time
Monitoring adherence trends per employee, location, and campaign.
Sales tips surfacing
Turning observed gaps into concrete coaching prompts for frontline staff.
Gamification mechanics
Ranking, scoring, and incentive views built on adherence and quality results.

Illustrative example

Input: Active campaign: "suggest online banking enrollment." In the transcript the employee says, "Our mobile app is great for checking balances," and never mentions enrolling in online banking.
Expected behavior: Intent-based classification marks the campaign objective as not satisfied, since an adjacent product mention is not the required enrollment suggestion, and the interaction is counted against campaign compliance for that employee and location.

Dashboard & Multi-Location Reporting

The 'single pane of glass' surface: a live, interactive dashboard presenting near-real-time CX feedback, transaction recordings, and KPI breakdowns. The FAQ states data is integrated across multiple locations simultaneously to support chain-wide decisions.

“Accessible: Real-time dashboard of CX feedback and transaction recordings.” casca.ai

Mapped capabilities

4 capabilities

Near-real-time data freshness
Newly scored interactions appear in the dashboard without a batch delay.
Chain-wide aggregation
Rolling up results across many locations while preserving per-location detail.
Drill-down to individual transactions
Navigating from a KPI figure to the specific interactions and recordings behind it.
Recording playback alongside scores
Pairing each scorecard with the underlying interaction recording for review.

Deployment & Multi-Site Operations

The operational surface described in the FAQs: installing small devices near points of sale, the stated sub-thirty-minute per-location installation, enabling any subset or all of the Manni services at once, and onboarding a chain of locations.

“Typically a full installation of the Manni devices in a single location takes less than half an hour.” casca.ai

Mapped capabilities

3 capabilities

Per-location device installation
Standing up capture devices and tablets at a new site.
Service module selection
Running Manni Insights, Manni Rate, or both, per the client's choice.
Multi-location onboarding
Adding sites to an existing chain-wide deployment without disrupting reporting.

Coverage is mapped from Casca's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Casca test?+

The coverage map is generated from Casca's own public product surface (in-store retail and banking customer experience (CX) analytics AI): 6 scoring areas — Interaction Capture & Data Collection, CX Quality Scoring (Manni Insights), and Custom CX Objectives & Campaign Configuration, and more — spanning 22 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Casca evals scored?+

Every case generated for Casca — across Interaction Capture & Data Collection and CX Quality Scoring (Manni Insights) and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Casca library include?+

The full Casca library is built on request. The coverage map spans 6 areas and 22 capabilities (for example, Audio capture and transaction segmentation and Manni Rate exit-survey intake under Interaction Capture & Data Collection); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Casca or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Casca areas and set them up in a Corsac workspace, where you can run every test case against Casca or your own agent with your own data.