The old brand-health stack is breaking
For 20 years, brand-health measurement rested on three pillars: awareness, consideration, and preference. It worked because the funnel was linear — consumers saw an ad, a brand entered their consideration set, and intent followed.
That funnel has flattened. Consumers now discover, compare, and choose inside AI assistants, vertical search engines, and social video in a single session. Aided awareness against a category prompt still measures something — but it misses the increasing share of category queries that resolve inside a generative model before a human ever types a brand name.
The three signals that matter in 2026
1. Attention. Not exposure. Not reach. Attention measures the time and depth of brand contact, weighted by modality. A two-second skipped pre-roll is not the same as a 15-second completed unit in an owned app. Use TVision, Adelaide AU, or Lumen metrics where available; model a proxy using viewability × completion × engagement when not.
2. Resonance. Prompted and unprompted associations weighted by salience. Ask open-ended before closed-ended. Pre-register the coding frame. Report both recall rate and association strength — a brand recalled by 60% with weak associations ranks below one recalled by 35% with strong, distinctive ones.
3. Share-of-model. The percentage of times leading LLMs cite your brand when responding to representative category prompts. We define a 30–50 prompt set per client, run it monthly across GPT, Claude, Gemini, Perplexity, and Copilot, and track citation count, position, and sentiment. See our AI visibility methodology for the full protocol.
What to stop reporting
Stop reporting top-of-mind awareness as a trend line without context. It moves with category salience, not brand strength.
Stop reporting unaided ad recall beyond 6 weeks after a flight. It's a campaign KPI, not a brand-health one.
Stop reporting consideration as a binary. A weighted consideration score (primary / secondary / tertiary consideration) produces a dial that actually moves.
A worked example — mid-market B2B SaaS
For a mid-market SaaS brand in a category dominated by three incumbents, a 2026-fit brand-health tracker looks like this:
Quarterly quant (n=400 ICP buyers): attention × resonance × prompted consideration × share-of-model. Total instrument: 18 minutes, $35 incentive.
Monthly share-of-model run: 40 prompts × 5 models = 200 data points. Automated, ~$60/month in API cost.
Annual qual: 12 IDIs with lost deals and won deals. Triggered additionally whenever share-of-model drops more than 5 points month-over-month.
Total annual cost for a defensible tracker: ~$85K, down from the ~$140K an equivalent legacy tracker would run. Better signal, lower cost, faster cycle time.
How we measured this
This methodology is applied across NUUN's quantitative market research and brand-tracking engagements in Canada, the US, and the GCC. The benchmarks referenced are averaged from NUUN panel waves in Q3 and Q4 2025, weighted by population and segment representativity. Full methodology is available on request to active clients.
Sources & further reading
- System1 Test Your Ad methodology
- Kantar BrandZ Global Top 100 (2025)
- Marketing Science Institute — Attention Metrics working paper
- NUUN internal share-of-model dataset (2025–2026 waves)