Top Mention Rate (TMR)
TMR answers one question: When your brand is mentioned, are you mentioned first?
It represents Preference — winning the race. AI users disproportionately act on the first brand listed, so being named first is materially different from being named third.
Formula — graded credit
TMR uses a graded credit model, not a binary "first or nothing". Credit is awarded based on the brand's first_position in the answer (1 = first brand named, 2 = second, etc.) and your site's TMR Tolerance Level (1–5).
TMR = average graded credit across all runs where the brand appears
The tolerance level controls how quickly credit drops off as the brand moves down the list:
| Tolerance | Behaviour |
|---|---|
| 1 (strict) | Only position 1 gets full credit. Positions 2+ get near zero. |
| 3 (default) | Position 1 = full, position 2 ≈ 60%, position 3 ≈ 30%, positions 4+ ≈ 0. |
| 5 (lenient) | Credit decays slowly — top 5 positions all score meaningfully. |
Pick strict tolerance if you're competing in mature categories where users only consider the top brand. Pick lenient if your category routinely lists 4–6 alternatives.
How to read it
- TMR is only computed over runs where the brand appears (so it's not penalised by VS gaps). A brand that appears 3 times out of 100 runs but always at position 1 will show TMR = 100%.
- Combine TMR with VS to get the real story:
- High VS + High TMR = dominant brand.
- High VS + Low TMR = "always-listed but never first" — usually a positioning/content authority issue.
- Low VS + High TMR = strong in a narrow niche.
Why first_position matters
LLMs are trained to surface the strongest signal first. Being mentioned first correlates with:
- More citation clicks in Gemini and the Google AI surfaces.
- Higher likelihood of being the brand the user remembers from the answer.
- Better downstream conversion when the user follows up with "tell me more about X".
Pairs with
- VS — defines the eligible runs.
- SOV — measures volume; TMR measures rank within the answer.