Data Coverage
Everything the forecasts learn from, and how much of it exists so far. Each signal is defined up front and matures only as real observations accumulate — every count below is computed live from imported data, never estimated.
Observations
481,393
across all signals
Signals tracked
22
defined in the registry
Mature
0 / 22
signals at their full observation target
Data confidence
9%
mean signal maturity
Journey dataLearning
Position movement across trailing 28-day windows — the strongest in-journey signal of trajectory.
3,607 / 100,000journeys with computed 28-day velocity
Ranking outcomeCalibrating
Journeys that broke into the top 20 — evidence of what this site can rank for.
7,397 / 25,000journeys reaching top 20
Ranking outcomeLearning
Journeys that held a stable top-20/top-10 ranking — the target event the forecast predicts.
2,123 / 25,000journeys with a stable milestone
Journey dataLearning
Journeys slipping meaningfully below their best-ever position — decay pressure on rankings.
2,127 / 25,000journeys >5 positions below their best
Data qualityLearning
Page–query pairs with too few impression-days for GSC position to be trustworthy.
3,926 / 50,000journeys with <10 impression-days
Data qualityCalibrating
Multiple pages on one site competing for the same query — splits ranking signals.
8,585 / 25,000journeys sharing a query with another page
Data qualityLearning
Journeys already ranking when the import window opened — excluded from time-to-rank training because their onset is unobserved.
1,614 / 50,000left-truncated journeys
Data qualityLearning
Brand-navigational queries — excluded from training because rankings are not earned by SEO work.
0 / 25,000journeys flagged as brand queries
Ranking outcomeLearning
Journeys observed without reaching the target — kept as censored data, never coded as failures.
6,666 / 50,000censored journeys
Journey dataLearning
Total page–query ranking journeys available as comparables for forecasting.
11,199 / 100,000all extracted journeys
Journey dataCalibrating
Daily page × query × position rows imported from Search Console — the raw substrate.
425,998 / 1,000,000daily GSC rows
Journey dataLearning
Queries with a classified search intent (informational, commercial, transactional, navigational, local) — used to match comparable journeys across the network.
3,175 / 50,000queries with a non-null intent
Data qualityLearning
Ranking milestones reached within ±14 days of a confirmed Google update — kept in training but flagged, because regime shifts can distort time-to-rank.
4,782 / 25,000milestones inside update windows
SERP dataLearning
SERPs observed where one domain holds 2+ of the top 10. Google's site diversity system generally caps listings per site — a structural constraint on additional pages targeting the same query.
1 / 10,000snapshots with a domain holding ≥2 top-10 spots
SERP dataLearning
Ground-truth SERP checks — corrects GSC's impression-weighted position blending.
3 / 50,000stored SERP snapshots
2 / 25,000snapshots with our position found
SERP dataLearning
Competing results observed per tracked SERP — who occupies the positions being targeted.
147 / 500,000competitor result rows
SERP dataLearning
External-link measurements (referring domains, backlinks) for pages currently ranking on tracked SERPs — real index counts, not authority scores.
21 / 25,000profile rows with backlink-index counts
SERP dataLearning
On-page structure of pages currently ranking on tracked SERPs: internal/external links on the page and word count, fetched from the live pages.
20 / 25,000profile rows with fetched page data
SEO actionsLearning
Logged content rewrites, expansions, and refreshes — future time-varying covariates.
0 / 10,000content-category work log entries
0 / 10,000link-category work log entries
SEO actionsLearning
Logged technical fixes, redirects, and structured-data changes.
0 / 10,000technical-category work log entries
Confidence grows as more websites and history are imported. Signals below their maturity target still inform forecasts, but with wider intervals and a lower confidence tier — the product says “insufficient data” rather than inventing a number. Each signal has its own observation target (10,000 up to 1,000,000, scaled to how fast that kind of observation accrues); these are provisional planning bars — the real thresholds should be set empirically from backtest calibration as the network grows. Every signal is defined in plain English in the glossary.