Worked examples · De'Longhi (US)

Basic usage & advanced analysis

Every chart on this page came out of the Rankscale REST API through Claude Code and the public rankscale-api-skill. Part 1 reproduces the standard dashboard views. Part 2 builds five analyses the web interface has no screen for. The tracked brand is De'Longhi (US), from its first tracked day to yesterday's data.

2026-05-28 → 2026-07-14 29 tracked days, one 20-day gap 7 topics × 9 engines = 63 terms 83 API calls Source: Rankscale REST API

Part 1 of 2

Reproducing the Rankscale web UI

The goal here is trust, not novelty: before building anything custom, confirm the API returns the same numbers the dashboard shows. Each block below is one standard dashboard view, rebuilt from POST /v1/metrics/report (plus /citations and the topics endpoint) with a custom date window - first tracked day to yesterday.

Basic · 1

Topics and prompts under measurement

The setup is deliberately simple: one prompt per topic, run on all nine engines, so every topic × engine cell rests on the same question. That gives 7 × 9 = 63 tracked search terms.

TopicPromptEngines

Basic · 2

Overall Visibility Score across days

Daily Visibility for the full window. The shaded band is the 20-day tracking pause - the line breaks rather than bridging it, because no data exists there.

Avg Visibility, entire period Rankscale's period aggregate (what the dashboard headline shows). It is period-weighted, not the mean of the daily values - the simple mean of the 29 dailies is . Both are correct; they answer different questions.
Visibility climbed from ~44 to ~57 across the window, most of it during the first tracked block. Hover for exact values.

Basic · 3

Standard metric scorecards

Everything /report returns by default for the window, with the trend vs. the previous period of the same length. One reading note: for Avg Position a negative delta means the position number went down, which is an improvement.

Snapshot values are Rankscale's weighted aggregates over the window - they are not sums of the daily series, so don't try to reconcile them by hand.

Basic · 4

Visibility by AI engine

Same metric, cut per engine. [API] = provider API, [GUI] = live browser interface - the same labeling as the consistency study. De'Longhi holds Brand Rank #1 in every engine, but the spread is real: ChatGPT sees the brand at 67.6, Google AI Mode at 48.4.

AI EngineVisibilityDetection %Avg PositionBrand Rank
Solid = [API] engines, dashed = [GUI]. The ordering is fairly stable across the window; ChatGPT and Perplexity stay at the top of the band, Google AI Mode and Google AI Overview at the bottom.

Basic · 5

Visibility over time - top 10 brands

The ten most visible brands of the window, De'Longhi highlighted. One honest limitation: the API's competitor time series only reaches back about two weeks, so competitor lines start on 07-07 even though their window aggregates cover the full period. The De'Longhi series is complete.

De'Longhi sits ~3× above the nearest competitor on every tracked day. The competitor cluster (Breville, Jura, Technivorm) runs between 5 and 21.

Basic · 6

Top 20 competitors

Sorted by window Visibility. Rank when found is the average list position on the runs where the brand actually appeared - a brand can rank well when found but be found rarely (see Fellow Aiden: position 2.6, but in far fewer answers).

#BrandVisibilityDetection %Rank when found

Basic · 7

Top 20 domains by citations

Which websites AI engines cite when answering the 63 tracked prompts, from POST /v1/metrics/citations: . Mentions your brand = at least one cited URL on that domain mentions De'Longhi.

#DomainCitationsShareMentions your brand

Part 2 of 2

Advanced analysis - past the web UI

The dashboard shows one cut at a time. The API lets you re-run the same report with filters and combine the cuts into tables that don't exist as screens. The custom metric used throughout is Brand Rank (by Visibility): the brand's position among all detected brands for a given cut, sorted by Visibility, rank 1 = most visible. It is computed from the competitor list on every call - deliberately not read from any ready-made rank field (see the method notes at the bottom for why).

Advanced · 1

Topics: Brand Rank and Visibility

One filtered /report per topic. This is the table that breaks the "rank #1" headline apart - the four automatic-machine topics are won outright, manual espresso and drip are not close.

TopicBrand Rankof brandsVisibility

Advanced · 2

AI engines: Brand Rank and Visibility

Same computation, cut per engine instead of per topic. Rank #1 across the board - averaged over all topics, every engine puts De'Longhi first. Keep A1 in mind while reading this: the engine averages are carried by the automatic-coffee topics.

AI EngineBrand Rankof brandsVisibility

Advanced · 3

Pivot: Brand Rank per topic × engine

The full 63-cell map - one filtered /report per cell. Darker = better rank. A hyphen means De'Longhi was never detected in that cell during the window. This is the single most useful table on the page: it shows which engine loses you which topic, not just averages.

Read row by row: bean-to-cup is rank #1 on eight engines and simply absent on Claude 4.5 Haiku (0 mentions in 28 runs). Drip coffee makers barely register anywhere - two engines rank De'Longhi at #20 and #22, the other seven never name it.

Advanced · 4

Best-ranked brands per GUI engine

For each browser-based engine, the brands sorted by Visibility - the answer to "who is winning in each engine users actually see". De'Longhi highlighted; the interesting part is places #2-#4, where Jura and Breville swap depending on the engine.

Advanced · 5

Citations per day - top 20 URLs that mention De'Longhi

From the URL-level citation data: the 20 most-cited individual pages that mention De'Longhi, and how often engines cited each one per day. 510 of the 3,724 cited URLs mention the brand; these 20 carry a large share of that volume.

Citations/day1 → 15  ·  blank = not cited that day
The heavy hitters (homecoffeeexpert.com, coffeeness.de, techradar.com) are cited every single tracked day - these pages are structural to how AI engines answer coffee-machine questions, not one-off sources. Hover any cell for the exact count.

Method & caveats

How these charts were made

83 API calls in total: 1 topics list, 1 search-terms report (for the prompts), 1 overall daily report, 1 citations pull, 7 per-topic reports, 9 per-engine reports, and 63 per-cell reports for the pivot. Every response was saved to a local JSON file first and all tables were computed from the files - the page never re-queries the API. The whole pull runs in about a minute, well inside the 200 requests/minute limit.

The only request shape that matters, with filters swapped per cut:

POST /v1/metrics/report
{
  "brandId": "<id>",
  "isoStartDate": "2026-05-28",
  "isoEndDate": "2026-07-14",
  "aggregation": "daily",
  "includeNotFoundExecutions": true,
  "selectedTopic": "<topic id>",     // per-topic and pivot cuts
  "selectedEngine": "<engine id>"    // per-engine and pivot cuts
}
Read with these in mind
  • The end date is exclusive. Asking the API for a window ending 07-14 returns data through 07-15. We keep the requested end date and slice the extra day off client-side rather than compensating in the request.
  • Aggregate ≠ mean of dailies. The window Visibility of 56.7 is Rankscale's period-weighted figure; the simple mean of the 29 daily values is 50.0. Headlines use the aggregate, trend reading uses the dailies.
  • Brand Rank is computed, not read. Rank = 1 + (brands with strictly higher Visibility in the cut). The API's ready-made rank fields don't reproduce the dashboard reliably (the MCP's own_brand_rank is always 1), so we don't use them.
  • Competitor daily series are short. The per-competitor time series only reaches back ~2 weeks regardless of the requested window - verified by re-querying the first block alone. Window aggregates are complete; only the competitor lines in chart B5 are truncated.
  • The 20-day gap stays a gap. No interpolation anywhere on this page.
  • One field oddity: the daily brandNotFound flag is true on all 29 days even though detection rate is 62% - it doesn't match the observable data and we ignored it.

Data: De'Longhi (US), first tracked day 2026-05-28 through 2026-07-14, pulled 2026-07-15 via the rankscale-api-skill. Raw responses cached as local JSON; reproducible against the same workspace.