Skill: HG Modeled vs. Observed Data
Numbers framed honestly — modeled estimates labeled, observed facts stated, no false precision.
Overview
Frame every quantitative claim with the right level of confidence. Claude labels modeled spend as 'modeled', observed installs as 'observed', and derived signals like FAI percentages as the statistics they are — so a CRO challenging a number gets a clear source frame instead of a guess.
Use cases
Briefs that survive a 'how do you know that' challenge
Every figure in the output carries the right qualifier on first mention: 'HG models spend at $50M' (modeled), 'HG observes Salesforce running at 12 locations' (observed), 'FAI: Sales 58%' (derived). The CRO's challenge gets a one-sentence answer.
Sales prose that doesn't read as overconfident
Modeled deal values become 'an estimated $1.2M annual run-rate'; high-confidence observed installs become 'they run X across 12 locations'. The reader knows which is which without reading the source notes.
View full skill
HG Modeled vs. Observed Data
When to use
- A workflow output makes a quantitative claim and needs a hedge.
- A prompt is producing prose for a CRO / VP audience who will challenge the source.
- An author is reviewing a deliverable for fabrication risk.
Per-tool classification
| Tool | Modeled / Observed | Caveat language |
|---|---|---|
company_firmographic | Mostly observed (filings, scraped sites) | None for high-confidence; "approximately" for low. |
company_technographic | Observed-and-modeled (HG verifies installs, models intensity) | "HG observes…" for the install; "intensity" is a modeled signal. |
company_spend | Modeled | "HG modeled spend of $X" |
company_contracts | Modeled deal value, observed expiry/source | "HG-modeled contract value of $X, expires [date]" |
company_fai | Derived (modeled from observed headcount + product signals) | "FAI: Sales 58% / Engineering 8%" — no extra hedge needed; the percentages are a derived statistic |
company_intent | Observed activity, modeled score | "Intent score 87" is fine; "active research" implies the score is real |
| SEC filings | Observed | "Per the FY2025 10-K" |
| Web search | Observed (but the source may be modeled) | "Per [domain.com]" |
Decision rule
Use the strongest qualifier the source supports. When in doubt, say "modeled". The reader will trust the brief more if the qualifier is upfront than if a CRO catches a precise-looking number that turns out to be an estimate.
When to drop the qualifier
The first mention of each source carries the qualifier; subsequent mentions in the same section can drop it. "HG models the company's IT spend at $50M; spending in security accounts for ~6% of that" — second sentence drops "models" because the source frame is established.
In a markdown table sourced entirely from company_spend, one introductory sentence ("HG-modeled spend, May 2026") covers every cell; do not repeat per row.
Common pitfalls
- Saying "they spend $X" without qualifier. Reads like an invoice extract; it's a model.
- Saying "modeled estimate of $X" twice in a row. Once is enough.
- Calling intent "active research" when it's stale. Pair with
hg-recent-signalsto keep the qualifier honest.
Reference
hg-insights-api.md#data-sources— per-endpoint data-source noteshg-citation-discipline— citation density + source priority