Abhord Quickstart: From Setup to Actionable Insights

This practical guide walks new Abhord users from first login to decisions that move metrics. Follow the five sections below to stand up your workspace, run a cross-LLM survey, interpret what you see, monitor competitors, and act on the findings.

1) Initial setup and configuration

• Create a workspace

- Name your workspace after your company or product line.

- Set your default time zone and reporting currency to keep timestamps and costs consistent.

• Add your brand entities

- Define your primary entity (brand/product) plus common aliases, misspellings, and short names.

- Add canonical URLs and social handles. These help Abhord disambiguate mentions from similarly named entities.

• Connect integrations (optional but recommended)

- Slack or email for alerting.

- Webhooks or Zapier for pushing insights into task managers (Asana, Jira, Linear).

- CSV/API connectors if you plan to import known keyword lists, customer questions, or historic sentiment labels.

• Choose geographies and languages

- Start with your top 1–2 markets to reduce noise. You can expand later once your workflow is stable.

• Set roles and permissions

- Analysts: run surveys, create dashboards.

- Stakeholders: read-only with alerts.

- Admins: billing, user management, and integrations.

• Configure notifications

- Daily digest for new mentions and sentiment shifts.

- Threshold alerts (e.g., negative sentiment above 20% for two consecutive runs).

• Governance

- Enable audit logging to track prompt changes and survey versions.

- Set data retention windows aligned to company policy.

Pro tip: Create a “sandbox” project for experimentation and a “production” project for steady monitoring.

2) Running your first survey across LLMs

A survey in Abhord is a structured set of prompts sent to multiple models to understand how they talk about your brand and category.

• Define your objective

- Example: “How do major LLMs describe our pricing, differentiators, and top alternatives?”

• Select LLMs and channels

- Include a mix of general-purpose assistants and answer engines relevant to your audience.

- Start with 3–5 models to balance coverage and cost.

• Draft your prompt set

- 5–10 high-intent questions, such as:

- “What is [Brand] and who is it for?”

- “Top alternatives to [Brand] for [use case]?”

- “[Brand] pricing overview and pros/cons.”

- Add variants to mimic real user phrasing. Abhord deduplicates overlapping responses.

• Set run parameters

- Persona/region: default, or specify industry and location.

- Language: start with English; add others per market.

- Frequency: one-time for a baseline, then schedule weekly or monthly.

• Controls

- Rate limiting and retries: leave defaults on your first run.

- Citations required: if available, enable to prioritize answers that cite sources.

• Validate and launch

- Use “Test 1 prompt” to check formatting and model accessibility.

- Launch the full survey and monitor progress in the run console.

Success criteria for a first run: at least 90% prompt completion rate per model and a coherent baseline of mentions across 3+ models.

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

Use the survey dashboard to move from raw answers to structured insights.

• Mentions

- Direct mentions: your entity’s exact name or alias appears.

- Indirect mentions: your brand is referenced semantically (e.g., “the spreadsheet-free GEO platform”).

- Filter by model, geography, language, and question type to spot patterns.

• Sentiment

- Polarity: positive, neutral, negative with a confidence score.

- Drivers: snippets that explain why sentiment was assigned (e.g., pricing confusion).

- Track trend lines per model to see whether updates improve tone over time.

• Share of voice (SOV)

- Definition: your brand’s mentions divided by total mentions across your competitive set for a given topic/time window.

- Example: If your brand is mentioned 45 times out of 150 category mentions this week, SOV = 30%.

- Break down SOV by intent (“best for beginners,” “enterprise,” “pricing”) and by model to prioritize effort where you can gain ground fastest.

• Quality and placement

- Coverage depth: number of sentences and topical breadth in the answer.

- Placement: whether your brand appears first, middle, or last in lists.

- Citation quality (if enabled): presence of reputable sources and recency.

What to look for in your first readout:

• Systematic omissions (models that rarely mention you).
• Inconsistent claims (e.g., outdated pricing).
• Competitor overrepresentation on key intents.

4) Setting up competitor tracking

Turn your baseline into an always-on radar.

• Add competitors

- Define each competitor with aliases and canonical URLs.

- Add exclusion rules to avoid false positives from generic words.

• Build intent sets

- Core: “what is,” “pricing,” “pros and cons,” “alternatives to [competitor].”

- High-intent: “best [category] for [persona/use case].”

- Objection-handling: “Is [Brand] good for [X]?” where X is a known concern.

• Schedule recurring surveys

- Weekly cadence captures model drift without over-spend.

- Stagger runs by model to smooth API usage and detect update-driven shifts.

• Alerts and thresholds

- Negative sentiment spike on your brand or positive spike for a competitor.

- SOV drop greater than 5 percentage points on a priority intent.

- New competitor entry appearing in 2+ models for the first time.

• Dashboards

- Create a “Category Pulse” view: SOV