Abhord Quickstart: A Practical Product Guide

This guide helps new Abhord users go from zero to actionable insights. You’ll set up your workspace, run your first multi-LLM survey, interpret the output (mentions, sentiment, share of voice), monitor competitors, and turn findings into next steps.

Before you begin, gather:

• Your brand and product names (plus common misspellings).
• A short competitor list (3–8 brands).
• 20–50 priority questions/customers’ intents (e.g., “best X for Y,” “pricing of Z,” “alternatives to A”).
• Target markets and languages.

1) Initial setup and configuration

1) Create a workspace and project

• Workspace: your organization.
• Project: one product line or market (e.g., “US SMB, Collaboration Suite”).

2) Define entities

• Brand and products: add official names, acronyms, and known variations.
• Competitors: include product-level entries if relevant.
• Disambiguation: add “anti-keywords” to avoid false matches (e.g., “apple fruit” excluded for Apple Inc. contexts).

3) Build an intent set (your “survey questions”)

• Start with 20–50 queries covering: comparison (“best…”, “vs”), evaluation (“pros/cons”), how-to, pricing, troubleshooting, and alternatives.
• Tag each query with an intent type (awareness, consideration, decision, post-purchase), market, and language.

4) Choose LLM coverage

• Select the models to survey (e.g., leading closed- and open-weight models).
• Regions/languages: pick endpoints aligned to your markets.
• Frequency: snapshot (one-off) or tracking (weekly/biweekly).

5) Configure analysis options

• Matching rules: exact + fuzzy for entity detection; enable canonicalization to group variations.
• Sentiment model: choose neutral/3-class (pos/neu/neg) or 5-class if you prefer finer granularity.
• Weighting: set rank weights (e.g., answers appearing first get higher impact) and model weights (if certain models matter more to your audience).

6) Invite teammates and alerts

• Roles: Viewer (read) vs Editor (manage surveys) vs Admin (billing, model access).
• Alerts: set thresholds (e.g., “Share of Voice drop >5 pts” or “Negative sentiment spike >10%”) with email/Slack/webhook delivery.

2) Run your first survey across LLMs

1) Create a survey run

• Select your intent set.
• Choose models and locales.
• Sampling: set responses per query per model (e.g., 3–5) and a max tokens budget.

2) Prompt strategy

• Use clear, comparable instructions so answers are consistent across models. Example:

“You are an unbiased assistant. For the query: ‘{query}’, provide