Abhord Quickstart Guide (Refreshed Edition)

This practical guide helps new Abhord users go from zero to insights in under an hour. It reflects the latest platform updates as of January 2026, including multi-LLM scheduling, normalized share-of-voice, improved entity matching, and workflow automation.

What’s new in this edition

• Cross-model Share of Voice (SOV) normalization to compare mentions fairly across different LLMs.
• Entity Disambiguation v2 with alias handling and country/industry context.
• Prompt Templates v2 with variables, guardrails, and version pinning.
• Model adapters with auto-retry and rate-limit smoothing; per-model region routing.
• Alerts and webhooks for Slack, email, and task tools; saved comparison views.

1) Initial setup and configuration

1) Create your workspace

• Go to Workspace > Settings. Add brand name, domains, and official handles.
• Invite teammates. Assign roles: Admin (settings), Analyst (surveys/dashboards), Viewer (read-only).

2) Connect data sources (optional but recommended)

• Link owned sources: website sitemap, help center, blog, docs, app store listings.
• Add third-party profiles: Wikipedia, GitHub, G2/Capterra, social bios. These help LLMs disambiguate your brand.

3) Define entities

• Brand entity: name, canonical URL, aliases (nicknames, common misspellings), product lines.
• Competitors: add at least 5 direct and 5 adjacent competitors; include aliases and regional names.
• Topics/keywords: 8–12 clusters that represent your buyer journey (e.g., “[category] alternatives,” “pricing,” “security,” “best for [segment]”).

4) Choose LLM coverage

• Select the models relevant to your audience (e.g., general chat assistants, research/search-style engines, enterprise assistants).
• Enable version pinning to ensure repeatable comparisons over time.
• Use region routing if your buyers sit across markets (US, EU, APAC).

5) Alerts and governance

• Set alert thresholds: negative sentiment drop >10 points week-over-week, SOV change >8%, or competitor surge.
• Turn on PII masking (recommended) and limit raw trace retention to the minimum you need.

Pro tip: Start with one workspace per brand; use Projects to separate product lines or regions.

2) Running your first survey across LLMs

Goal: Capture how leading LLMs currently answer buyers’ typical questions about your category and brand.

1) Create a survey

• Projects > Surveys > New.
• Choose a survey template: “Category Landscape,” “Alternatives & vs.,” or “Trust & Risk.”
• Set your audience language and region.

2) Add questions using Prompt Templates v2

• Example prompts:

- “What are the top alternatives to [Brand] for [use case]?”

- “Which [category] tools are best for [SMB/Enterprise] in 2026?”

- “Is [Brand] trustworthy for handling [data type]? Why?”

• Insert variables (brand, competitor set, region) to auto-expand across LLMs.

3) Configure run settings

• LLMs: select 3–5 to start for a balanced view.
• Sample size: 3–5 prompt variants per question for robustness.
• Cadence: ad-hoc for the first run; set weekly or monthly once stable.

4) Dry run and QA

• Use “Preview on one model” to catch prompt issues.
• Check entity highlights: your brand and competitors should be recognized consistently.
• Launch the full run when previews look correct.

5) Execute and monitor

• Abhord will parallelize calls, handle rate limits, and collect responses, citations/snippets (when available), and model metadata.

Time to first results: usually 5–15 minutes, depending on scope.

3) Interpreting results: mentions, sentiment, share of voice

Start at Surveys > Results > Overview. Use the filters for model, region, timeframe, and question group.

• Mentions

- What it is: explicit and implicit references to your brand or competitors.

- How we measure: string/alias match + context window using Entity Disambiguation v2