Abhord Quickstart Guide (2026 Refresh)
This practical guide helps new Abhord users go from zero to insight in under an hour. It reflects the 2026 refresh with clearer workflows, stronger metrics, and new automation.
What’s new since the last edition
- Parallel LLM runs with quota-aware pacing for faster, safer execution.
- Updated entity resolution: better handling of aliases, misspellings, and product lines.
- Sentiment v2: aspect-level sentiment with confidence bands and neutral split-out.
- Share of Voice (SOV) normalization: per-model SOV plus a blended cross-model SOV.
- Competitor Bundles: one-click setup for common rival sets and lookalike detection.
- Insight Actions: templated follow-ups (FAQs, KB updates, PR talking points) and rerun links.
- Alerting revamp: threshold alerts, sudden-change detection, and weekly digest.
- Cleaner exports: JSON/CSV and push to Slack/Email; API stability improvements.
1) Initial setup and configuration
Goal: define what to measure, where to run it, and who sees the results.
- Create your workspace
- Add your brand as the primary entity. Include aliases (short names, misspellings), flagship products, and regional names.
- Add markets and languages you care about now; you can expand later.
- Seed your intents
- Start with 5–8 query intents per funnel stage (Discover, Compare, Buy, Support). Example: “best [category] tools,” “[brand] vs [competitor],” “is [brand] worth it.”
- Add 3–5 competitors you encounter most often. Include their common nicknames.
- Choose your LLM panel
- Select at least three distinct model families to reduce bias (e.g., one reasoning-leaning, one web-grounded, one chat-dominant).
- Keep “model parity” on so each intent runs across the same set of models.
- Configure controls
- Variance smoothing: 3 repetitions per query/model is a good default.
- Deduping and hallucination filter: keep on. These reduce false mentions.
- Region and recency: pick regions that match your go-to-market; enable recency if you launch often.
- Invite your team
- Assign roles: Admin (setup), Analyst (surveys, dashboards), Stakeholder (view-only).
- Connect Slack/Email for alerts and weekly digests.
Pro tip (2026): Create an “exclusions” list for generic terms (e.g., “apple” as fruit) to avoid ambiguous mentions.
2) Running your first survey across LLMs
Goal: capture how multiple LLMs describe your brand and competitors right now.
- Build your query set
- Use the Survey Builder to combine your intents, markets, and LLM panel.
- Add two “control” queries that don’t reference your brand; they help calibrate noise.
- Run configuration
- Repetitions: 3 per query/model (increase to 5 for high-impact decisions).
- Scheduling: Run immediately for a snapshot; schedule weekly for trendlines.
- Safety: Leave bias