Useful signals.
Visible evidence.
Surfacd helps you decide which keywords to investigate. Its metrics are transparent research aids, not predictions of ranking or revenue.
Research workspace: two independent scores
The V1 research workspace uses retained page-one organic search results and the associated app listings. Each distinct organic app counts once. Sponsored placements and editorial content are shown separately and do not enlarge the organic denominator.
Keyword difficulty: lexical coverage
This 0–100 score measures how strongly app copy covers the searched phrase. Exact phrases carry more weight than close formatting variants, reordered words, nearby words or individual components. App names, listing headlines, search subtitles, descriptions, features and contextual taxonomy carry different field weights. A high value means stronger lexical coverage, not a measured probability of ranking or proof that those apps satisfy the merchant's need.
Competition level: estimated intent fulfillment
D1 Ultra assesses each app–keyword pair from the keyword's meaning, search intent, merchant need and retained listing evidence. It distinguishes apps that fully satisfy the need, partially satisfy it, serve a different need, or remain unresolved. The semantic assessment is independent of keyword-match scores.
Full fulfillment contributes one competitor-equivalent, partial fulfillment contributes half, and irrelevance contributes none. An unresolved interpretation contributes a range from zero to one; the displayed estimate uses the midpoint. Competition level expresses that estimate on a 0–100 scale across the distinct organic apps.
Completed estimates can be displayed despite ambiguous judgments. Model decisions and source claims can be wrong. This is an adopted research method, not a claim of universal measured classification accuracy. Lexical difficulty and semantic competition remain separate; they are not blended into a single score.
Availability and interpretation
Verified-empty searches, unavailable source evidence and incomplete assessments remain nonnumeric. Missing evidence is not a zero score or proof of an opportunity. Captures describe their observation time and collection context, not guaranteed current rankings. Synthetic demo scores are labeled illustrative examples, not measured publications.
Legacy explorer methodology
The panel index and difficulty_v4 explanation below describe the historical explorer. They do not define the V1 research workspace's lexical difficulty or semantic competition scores.
Two sources, two different questions
Demand evidence comes from observations in an uploaded panel dataset. Collection period, geography, language and rights/privacy review belong to that dataset. Missing metadata is shown as “Not supplied.” An observation is not necessarily a unique merchant, search, click, install or purchase.
Competition evidence comes from a separate, timestamped search-result observation. Demo evidence is synthetic. Live retrieval requires documented authorization and a verified integration. The browser collector records live organic app titles and subtitles separately from the panel dataset.
Search volume index volume_v1
A relative demand index calculated from observed searches in our panel. It is not an estimate of monthly searches or unique merchants.
Count 1–50: count × 1.75 Count > 50: 87.5 + (count / 20) Clamp to 1–100; display one decimal.
The founder-specified formula intentionally jumps from 87.5 at 50 observations to 90.05 at 51. Counts of 1 and 200 produce 1.75 and 97.5; we do not stretch the endpoints. Sorting uses unrounded values. A phrase absent from the dataset has unknown demand, never zero demand.
Legacy keyword difficulty difficulty_v4
A keyword-competition proxy based on exact-phrase coverage in sampled organic app titles and subtitles. It does not predict your probability of ranking.
N = up to 20 unique organic apps on page one T = apps with a title phrase match S = apps with an app card subtitle phrase match Score = round_half_up(1 + 99 × (2 × T + S) / (3 × N))
The 2:1 title-to-subtitle weighting is a product heuristic, not a measured Shopify ranking weight. Sponsored placements and duplicate app cards are excluded. Each app counts at most once per field. The evidence view retains the original extracted text and occurrence counts.
If Shopify verifies that a search returns no organic apps, difficulty is unavailable because there are no apps in the denominator. The result is labeled “No organic results observed.” This does not mean competition is low. Blocked, incomplete, or uncertain searches also remain unscored.
- 1–33: Lower lexical competition.
- 34–66: Moderate lexical competition.
- 67–100: Higher lexical competition.
Matching is Unicode-aware, case-insensitive and whole-phrase. Whitespace and equivalent hyphen separators are normalized: “Email Pop-Up” matches “email pop up”; “email popup” does not. “review” does not match “reviews”, and “push” does not match “PushOwl”. No stemming, synonyms or fuzzy matching affect scores.
When evidence is missing
Unknown subtitles, unresolved ad classification, blocked pages and incomplete extraction produce no score. Verified empty results remain unscored and are labeled “No organic results observed.” Samples smaller than 10 apps carry a warning. These are evidence limitations, not statistical confidence intervals.
Timing and reproducibility
Snapshots up to 24 hours old are fresh. From 24 hours to seven days, they are stale; older snapshots are historical. A failed refresh preserves prior successful evidence. Shared-query refreshes have a one-hour cooldown. Exports record the dataset ID, snapshot IDs, timestamps, and scoring versions.
Demo and production are separate
The demo uses deterministic synthetic observations and fictional app cards. No demo action retrieves Shopify results. Production does not fall back to demo data when its dataset or provider is missing. Surfacd is independent of and not endorsed by Shopify.
Explore the demo