Abhord Quickstart Guide (2026 Refresh)
Last updated: March 5, 2026
What’s new since the previous edition
- Expanded Model Matrix: run surveys across GPT‑4.1/4.1‑mini, Claude 3.7, Gemini 2.0, Llama 3.3, Mistral Large 2, and more.
- Market‑weighted Share of Voice (SoV) now the default, with equal‑weight mode still available.
- Mentions Clustering 2.0 reduces duplicates and unifies variants and misspellings.
- Normalized sentiment scale (−100 to +100) with aspect and certainty scores.
- Scheduling, anomaly alerts, and Slack/Webhook integrations are GA.
- Cost controls: per‑run budget caps and token ceilings; improved audit logs.
- Multi‑locale and language segmentation graduated from beta.
1) Initial setup and configuration
Goal: stand up a clean workspace, connect destinations, and define the entities you’ll measure.
- Create your workspace
- Go to Settings > Organization. Add SSO (Google/Okta) and set roles: Owner (billing + admin), Analyst (create/edit), Viewer (read‑only).
- Configure data region (US/EU) and retention (default 12 months). Enable PII redaction by default.
- Connect destinations and alerts
- Integrations: Slack, Email, Webhooks; optional: BigQuery, S3, and Google Sheets.
- Set default alerts channel and escalation rules (e.g., >10pp SoV swing or sentiment < −30).
- Define brands, products, and competitors
- Navigate to Taxonomy > Entities. Add:
- Primary brand and product lines (include spelling variants, abbreviations, and local names).
- Competitors and their product families.
- Deny‑list generic terms (e.g., “apple” as fruit) and add synonym groups.
- Choose your LLM roster and budgets
- Settings > Models: select the models you’ll survey. Keep at least one frontier model and one fast/low‑cost model for breadth.
- Set per‑run budget caps and per‑model token ceilings to avoid surprises.
- Seed knowledge (optional but recommended)
- Upload core pages (FAQs, product docs, comparison pages). Abhord uses these as grounding hints to improve entity resolution and theme attribution.
Pro tip
Create a “Starter Preset” with your entity list, deny‑list, default models, and sampling size. It saves clicks and ensures methodological consistency across teams.
2) Running your first survey across LLMs
Goal