Abhord Quickstart Guide (2026 Refresh)
What’s new in this edition
- Faster setup: a guided workspace wizard now configures projects, entities, and model connections in one flow.
- Smarter entity detection: auto-suggested synonyms and fuzzy matching reduce missed brand/product mentions.
- Token‑normalized share of voice: SOV now adjusts for model verbosity and response length.
- Model bias diagnostics: per‑LLM skew and variance are reported side‑by‑side with confidence bands.
- Scheduling and alerts: save surveys, run them on a cadence, and push alerts to Slack/email/webhooks.
- Cross‑model consensus: a new score highlights when multiple LLMs agree (or disagree) about your brand.
1) Initial setup and configuration
- Create your workspace
- Go to Admin → Workspaces → New.
- Add teammates with roles: Viewer (read‑only), Analyst (create/edit surveys), Admin (billing, connectors).
- Connect model providers
- Add API keys for the LLMs you plan to survey (you can connect multiple vendors).
- Set a default “LLM bundle” (e.g., 5–8 diverse models) under Admin → Models → Bundles.
- Recommendation: keep at least one smaller, faster model in the bundle for quick smoke checks.
- Pick data sources for enrichment (optional but recommended)
- Connect your docs, GitHub repo, help center, and public pages. This helps Abhord link LLM claims to your canonical content during analysis.
- Define entities and taxonomy
- Add your brand, products, competitors, and key features under Entities.
- Include known aliases/synonyms (e.g., “ACME PM,” “ACME Project Manager,” “APM”).
- New: use Auto‑Suggest to import common variants from recent LLM responses and user queries.
- Privacy and regions
- Choose data region (US/EU) and set retention. Toggle PII scrubbing if you ingest support data.
- Notifications
- In Admin → Alerts, connect Slack/email/webhooks and set thresholds (e.g., “Competitor X share of voice > 25%”).
Pro tip: Start with one product line and 3–5 competitors. You can scale up your taxonomy and bundles once you see stable patterns.
2) Running your first survey across LLMs
- Choose a template
- Go to Surveys → New → Templates and start with:
- “Best tools for [job-to-be-done]”
- “Who are the leaders in [category]?”
- “[Brand] vs [Competitor] comparison”
- Draft the prompt
- Keep it neutral and specific: “List the top project management tools for mid‑market teams (100–1,000 employees) and explain why. Consider pricing transparency, integrations, and security certifications.”
- Add clear constraints: “Answer in up to 120 words. Cite products by their canonical names.”
- Select the LLM bundle and sampling
- Pick your default bundle. Set runs per model (3–7) to estimate variance without over‑spend.
- Turn on fixed seed for reproducibility when comparing runs week‑over‑week.
- Launch and monitor
- Hit Run. You’ll see per‑model progress and estimated completion time.
- Optional: enable link capture to extract any URLs the models reference for later auditing.
Common gotchas
- Over‑broad prompts inflate mentions and muddy