Eval directory

Evals for Alta

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

About Alta

Alta is an AI platform for go-to-market teams built around named agents — Katie (outbound), Alex (inbound), and Luna (growth) — that find prospects, qualify leads, and book meetings. It pulls audience data from 50+ sources and orchestrates outreach across email, LinkedIn, SMS, WhatsApp, and calls with signal-based timing and personalization. Luna layers on pattern detection, A/B testing, and recommendations, and the platform is sold with custom quote-based pricing and enterprise security controls.

Industry

AI go-to-market (GTM) sales agent platform

Website

www.altahq.com

Use the eval library for Alta

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 Alta?

6 scoring areas · 24 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

Audience Intelligence & Data Sourcing

Building target audiences from CRM, intent signals, job postings, news, and product usage across the 50+ connected sources the platform advertises.

“Build from 50+ data sources — CRM, intent signals, job postings, news, product usage.” www.altahq.com

Mapped capabilities

4 capabilities

Multi-source audience construction
Assembling a target list from CRM plus enrichment and third-party signals rather than a single source.
Best-fit and lookalike account selection
Deriving lookalikes and ICP-fit accounts from existing customer and CRM data.
Source attribution and freshness
Attributing each attribute to its originating source and respecting stale or missing data.
CRM sync fidelity
Keeping account, contact, and deal state consistent with the connected CRM.

Signal-Based Timing & Triggers

Deciding when to act based on buying signals, engagement patterns, and trigger events, as described in the platform's 'right time' positioning.

“Email, LinkedIn, SMS, WhatsApp, calls — sequenced intelligently with condition-based branching that adapts in real time.” www.altahq.com

Mapped capabilities

4 capabilities

Buying-signal detection
Recognizing signals that justify initiating or accelerating outreach.
Trigger-event response
Acting on discrete events such as job changes, funding, or news within an appropriate window.
Engagement-pattern adaptation
Adjusting cadence based on opens, replies, and prior interaction history.
Suppression and hold conditions
Withholding outreach when signals are stale, contradicted, or a prospect has asked to wait.

Multi-Channel Outreach Orchestration

Sequencing email, LinkedIn, SMS, WhatsApp, and calls with condition-based branching that adapts mid-sequence.

Mapped capabilities

4 capabilities

Channel selection and sequencing
Choosing and ordering channels for a given persona and context.
Condition-based branching
Taking the correct branch when a prospect replies, bounces, or goes silent.
Mid-sequence state changes
Pausing, resuming, or terminating a running sequence in response to new input.
Channel-appropriate message form
Respecting length, format, and etiquette differences across email, LinkedIn, SMS, WhatsApp, and calls.

Illustrative example

Input: Mid-sequence, a prospect replies to a LinkedIn message: "Not now — ask me again in Q3." Three email steps and one call step remain scheduled.
Expected behavior: The running sequence pauses and all remaining email and call steps are cancelled or deferred, with the defer reason and the requested Q3 timing recorded for a later follow-up.

Personalization & Message Grounding

Drawing message content from deal history, case studies, competitor intel, and company context without drifting into unsupported claims.

Mapped capabilities

4 capabilities

Use of deal and account history
Referencing prior deals, contacts, and interactions accurately.
Case study and competitor grounding
Citing only proof points that exist in the supplied context.
Claim fidelity
Avoiding invented metrics, customers, or capabilities in outbound copy.
Persona-appropriate framing
Adapting message angle to the recipient's role and seniority.

Illustrative example

Input: Draft a first-touch email to a VP of Engineering at a 200-person fintech. CRM shows one closed-lost deal from 2024. No fintech case study or benchmark exists in the provided context.
Expected behavior: The email references the 2024 closed-lost deal as its personalization hook and makes no claim about fintech customers, named logos, or performance numbers, since none appear in the supplied context.

Agent Roles, Coordination & Optimization

The division of labor between Katie (outbound), Alex (inbound), and Luna (growth), including Luna's pattern detection, A/B testing, and recommendations.

Mapped capabilities

4 capabilities

Outbound agent scope
Katie staying within prospecting and meeting-booking responsibilities.
Inbound qualification and routing
Alex qualifying inbound leads and routing or booking appropriately.
Luna recommendations and experiments
Surfacing scale/pause guidance and A/B test results tied to observed performance data.
Cross-agent handoff
Transferring context cleanly when a lead moves between inbound, outbound, and growth workflows.

Trust, Access Control & Data Handling

The enterprise security surface the platform sells: SAML, granular and hierarchical permissions, audit logging, IP restrictions, tenant isolation, and AI data-sharing controls.

Mapped capabilities

4 capabilities

Authentication and permission enforcement
SAML sign-in with granular and hierarchical access boundaries honored.
Audit logging and IP restrictions
Recording privileged actions and enforcing network-level access limits.
Tenant and data isolation
Single-tenant data warehouse and encryption-at-rest/in-transit behavior.
AI data-sharing policy adherence
Honoring customer-configured limits on what data reaches models.

Coverage is mapped from Alta'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 Alta test?+

The coverage map is generated from Alta's own public product surface (AI go-to-market (GTM) sales agent platform): 6 scoring areas — Audience Intelligence & Data Sourcing, Signal-Based Timing & Triggers, and Multi-Channel Outreach Orchestration, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Alta evals scored?+

Every case generated for Alta — across Audience Intelligence & Data Sourcing and Signal-Based Timing & Triggers 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 Alta library include?+

The full Alta library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Multi-source audience construction and Best-fit and lookalike account selection under Audience Intelligence & Data Sourcing); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

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