Data and methodology
Where every number comes from.
Competitor data is only useful if you know how it was gathered and where it stops being reliable. This page lists every source Figo uses, how often it refreshes, and what it cannot tell you.
Every source, in one table
| What you see in Figo | Where it comes from | How fresh | Kind |
|---|---|---|---|
| Keyword positions and estimated monthly traffic | Google search index data, queried for the country set on your workspace | Weekly | Estimated |
| New and changed pages, blog posts | The competitor's own sitemap and pages, read by Figo's crawler | Weekly | Measured |
| Blog publish dates | The page's own markup: article dates, structured data, time tags, then the byline | Weekly | Measured |
| Google Ads | Google Ads Transparency Center | Weekly | Measured |
| Meta ads (Facebook and Instagram) | Meta Ad Library | Weekly | Measured |
| TikTok ads | TikTok Ad Library | Weekly | Measured |
| Instagram and Facebook posts | Public profile posts | Weekly | Measured |
| Reddit mentions | Reddit post and comment search, then a language model decides which threads are really about the business | Weekly | Model-judged |
| Google rating, review count and star spread | Google Business Profile listing data, matched on the competitor's website domain | Weekly, from an index that can lag the live profile by a few weeks | Measured |
| Trustpilot rating and review count | The public Trustpilot profile, where one exists | Weekly | Measured |
| AI visibility | ChatGPT, Claude, Gemini and Perplexity, asked your buyer questions with web search on | Every Monday | Measured |
| The weekly briefing text | Claude, writing only from the data above | Every Monday | Model-judged |
Measured means Figo read it from a public source and stored what it found. Estimated means it is modelled from search data and should be read as a size and a direction, not a precise figure. Model-judged means a language model made a decision, either about relevance or about how to summarise stored facts.
The rules Figo follows
A change needs two observations. The first check of a competitor is a baseline. Until a second weekly check exists, at least five days later, the report is labelled a first look and never claims something stayed the same or moved.
Dates are what the source says, not when we looked. A blog post carries its own publish date where the page states one. When it does not, Figo shows when it first saw the page and labels it that way rather than guessing.
Named and cited are different. In AI visibility, the headline percentage counts answers where an assistant named the business. Being read as a source is tracked separately, because a citation is not a recommendation.
Quiet weeks stay quiet. When nothing meaningful changed, the briefing says so and gives no action. Manufacturing a task every week is how a report stops being read.
Every number links to its evidence. The ad in the ad library, the page on the site, the keyword and its position, the full AI answer with its citations. Nothing asks you to trust a summary.
What Figo does not do
- Figo cannot see spend, revenue or results. Ad libraries publish creatives, not budgets, so a long-running ad is a hint and never proof that it works.
- Only public marketing is visible. Email, direct mail, partnerships, events and referrals do not appear.
- Google review data comes from an index that can lag a live profile by a few weeks, so a brand-new review may not show up in this week's check.
- AI answers vary between runs and between countries. One run is a sample, not a ranking, which is why Figo asks the same questions every week and reports the share across all of them.
- Ad libraries have gaps by country and by platform. A business can be absent from one library and still be advertising.
- Traffic estimates get less reliable for very small sites, where a handful of keywords swings the number.
Common questions
Are the traffic numbers real visits?
No, and no tool outside a business can give you that. They are estimates built from the keywords a site ranks for and the search volume behind them. They are dependable for comparing businesses and spotting a trend, and they should never be quoted as exact visits.
Why did my first report say 'first look' instead of showing changes?
Because a change needs two observations. Until Figo has checked a competitor twice, at least five days apart, it describes what they do now rather than inventing a comparison.
What is the difference between named and cited in AI visibility?
Named means the assistant recommended the business by name in its answer. Cited means the assistant read a page on that business's website while answering, without necessarily recommending it. Figo counts them separately, because only one of them is a recommendation.
How do you keep Reddit mentions relevant?
A keyword match is not enough when a brand name is also an ordinary word. Every candidate thread that does not mention the domain is read by a language model with the business name and website, and only threads genuinely about that company are kept. Most weeks, most candidates are discarded.
Does Figo see private data?
No. Everything comes from public sources: public ad libraries, public websites, public profiles, public reviews and public AI answers. Nothing behind a login, and nothing personal about a competitor's customers.
How often does Figo check?
A full check the day you add a competitor, then one automatic check a week. The AI assistants are asked every Monday at 5am Sydney time and the briefing is written at 7am.
See the rules applied to a real report
The sample briefing shows the evidence, the labels and the limits in context.
Keep reading
- Weekly competitor report example the output these rules produce.
- For marketing teams how the week gets prioritised.
- Privacy what Figo stores about you and about competitors.