Skip to main content
Glarify
Methodology · Glarify Ask

How the Intensity score (1–30) is built

Every suggestion in Glarify Ask carries an Intensity from 1 to 30 — an estimate of how raw the autosuggest demand signal is. Higher means the search engine surfaces the query earlier and under a less-typed seed.

Try Glarify AskAll methodologies
Three tiers

The 1–30 scale is three stacked tiers

Each tier corresponds to a different autosuggest scenario for the same seed. The position inside that scenario maps linearly to the tier's slot.

Tier 1
21 – 30

Raw seed

openclaw
#1
30
#2
29
#3
28
#4
27
#5
26
#6
25
#7
24
#8
23
#9
22
#10
21

Suggestions Google offers the moment you type the bare word — the strongest expression of "this is what people search for after this term."

Tier 2
11 – 20

Seed + trailing space

openclaw␣
#1
20
#2
19
#3
18
#4
17
#5
16
#6
15
#7
14
#8
13
#9
12
#10
11

A trailing space signals "I have committed this word, what comes next." Often returns a completely different list than the bare seed.

Tier 3
1 – 10

Seed + single letter

openclaw a … z
#1
10
#2
9
#3
8
#4
7
#5
6
#6
5
#7
4
#8
3
#9
2
#10
1

Letter-by-letter fan-out captures the long tail. Each letter is a separate autosuggest call, scored within its own slot of 10.

Formula: intensity = tierCeiling − min(rank, 9)
Aggregation

A query can show up in multiple tiers

When the same suggestion surfaces under more than one autosuggest scenario, we keep the maximum intensity and report tierHits so the UI can flag cross-confirmed demand.

Example query: openclaw ai
Tier 1openclawrank #5
26
Tier 2openclaw␣rank #2
19
Tier 3openclaw arank #1
10
Final score: 26
Max of (26, 19, 10) · tierHits = 3 → cross-confirmed across all three scenarios.
Adaptive deepening

We deepen only the letters that earned it

A blind aa … zz fan-out would cost 676 autosuggest calls per seed. Instead, after the first pass we promote only productive letters and modifiers.

Letter promotion

Any tier-3 letter L that returned ≥4 unique suggestions is deepened with {seed} L{a..z}.

  • · Top 5 most productive letters only
  • · Catches second-letter tails like openclaw banned
  • · Worst case: 5 × 26 = 130 extra calls

Modifier promotion

Any aux modifier M (how, why, vs, for, best, …) with ≥3 unique results is deepened with {seed} M {a..z}.

  • · Top 4 modifiers only
  • · Covers comparison and question intents
  • · Worst case: 4 × 26 = 104 extra calls
Cached in Postgres for 7 days

Deepened results land in question_research_cache with a 7-day TTL. The first user pays the ~80 extra calls; everyone after gets it free until the cache expires.

Why tiered, not a single rank

Rank #1 in "openclaw a" is not the same as rank #1 in "openclaw"

RANK-ONLY METER

Punishes long-tail queries that genuinely matter. The top suggestion for openclaw a and the top suggestion for openclaw both look like "#1" — but they represent very different demand.

TIERED INTENSITY

Gives users a comparable yardstick across very different autosuggest states — similar to AnswerThePublic's intensity wheel, but with explicit tier semantics you can debug.

See Intensity live on your own seed

Glarify Ask fans out across all three tiers in seconds, then deepens the productive ones. Every result shows its score and the tiers it surfaced in.

Open Glarify Ask