How Keyword Difficulty Is Calculated# Link to this article
Rankato Keyword Difficulty (RKD) is a deterministic 0-100 score computed with no LLM and no paid data, broken down per factor with a confidence band and a model version.
Rankato Keyword Difficulty (RKD) answers "how hard would it be to rank for this?" — and it is deliberately the opposite of a black box. The score is computed by a deterministic service with no LLM and no paid data. An assistant may later explain an RKD; it never computes one.
What the score is made of# Link to this section
RKD blends six factors. The weights are fixed for the model version — changing one is a model-version bump, not a silent tweak:
- Link strength of ranking pages — 30%
- Authority of ranking domains — 22%
- Domain saturation (brand dominance) — 14%
- SERP feature crowding — 12%
- Commercial intent — 12%
- Exact-term title competition — 10%
Missing inputs are dropped, never guessed# Link to this section
When an input is unavailable, its factor is marked unavailable, removed from the blend, and its weight is redistributed across the factors we do have. If there is nothing to score at all — no organic results and no SERP features — RKD returns no score rather than a misleading 0. A 0 would read as "easy"; "insufficient data" is the truthful answer.
Why v1 is labelled Experimental# Link to this section
The strongest difficulty signal is the link authority of the pages that already rank, and that backlink graph is not yet ingested. So in rkd-v1 the link factor is unavailable and its weight is redistributed across the SERP-structure factors. Because of that, confidence is capped at medium — high confidence structurally requires the link factor, so v1 can never reach it.
Three factors are structural proxies pending better data and are labelled Experimental even within v1: authority of ranking domains, domain saturation, and exact-term title competition.
Reading a score honestly# Link to this section
Every RKD carries its per-factor breakdown, its confidence band, and its model version. Compare RKD to RKD within the same model version; do not compare it to another tool's KD, which is computed from different inputs with different weights.
Written by Rankato