Using AI as a Time Machine for Brand Value
How much do people really value a brand in their everyday lives, not on a survey scale, but in terms of what they’d actually give up?
How much do people really value a brand in their everyday lives, not on a survey scale, but in terms of what they’d actually give up? That’s the question Felix Eggers (Copenhagen Business School) and collaborators have been working on, and it leads to a very different way of thinking about brand tracking: one where AI can reconstruct brand value over time, simulate alternative futures, and turn tracking from static snapshots into a dynamic “time machine.”
From “do you like this brand?” to “would you give it up?”
Traditional tracking asks people how much they like a brand or how likely they are to recommend it. Useful, but indirect. This research asks something much more concrete:
“Would you give up access to [Brand X] for one month if we paid you $Y?”
By varying the amount, we see what people actually need to be paid to temporarily give up a brand. That give-up price is a direct, behavior-anchored signal of the brand’s real value in their lives.
Training AI to behave like thousands of consumers
In work presented at the MSI conference, Felix showed how to go a step further:
- Collect incentive-compatible survey data. Real consumers make real trade-offs.
- Train a large language model on these choices. The model learns to behave like realistic consumers, segment by segment and year by year.
- Use the calibrated model as an AI consumer panel. Once calibrated, it can simulate thousands of realistic choices across conditions and time periods.
The key result: the AI’s simulated choices align closely with real survey data, which means it can credibly fill in missing years, explore what-if scenarios, and track brand value through time.
Why this matters for brand tracking
A deeper brand value metric. “What would you need to be paid to give this up for a month?” captures how embedded the brand is in people’s lives: a richer core KPI than stated liking.
Faster, more affordable tracking. Use real survey waves as anchor points and let the calibrated model generate high-quality simulated waves in between, grounded in real respondents, without exploding the budget.
Trends, not snapshots. Build a brand-value time series: how value moved around launches, campaigns, or crises; whether a competitor’s entry eroded your value or just your share of voice.
Scenario planning. Simulate how usage shifts, bundle or positioning changes, regulation, or new competitors might affect brand value, stress-testing strategy before the market does it for you.
From research to product
This “AI time machine” approach is built into BrandSight™: track not just what people say but what they’d need to be paid to give you up, combine survey rigor with calibrated AI panels, and explore historical trends and future scenarios in one platform.
Want to explore it for your category? Book a demo to discuss a pilot or an integration with your existing tracking.