Respondents quality

How do we ensure the quality of responses ?

We take the quality of our data seriously, and several safeguards are in place to ensure trustworthy and meaningful responses:

• Respondents who complete the survey too quickly are automatically excluded.

• No respondent can answer the same questionnaire twice.

• Each respondent may only be surveyed once every three months.

• We conduct regular quality checks on our respondent base to maintain data integrity.

Targeting logic

The survey is not broadcasted randomly. Happydemics sources its respondents through country-specific advertising inventories, ensuring that surveys reach users within the same national digital ecosystem as the campaign.

From there, broadcasting is continuously adjusted and optimized based on real-time performance signals. As Ad Recall responses are collected, the system identifies which inventory environments generate the highest-quality responses and concentrates delivery accordingly. Sources that consistently produce low-quality responses are progressively excluded as part of our ongoing optimization process.

Threshold consistency

The 100,000 unique impressions is a recommended minimum, not a strict requirement. Its primary purpose is to ensure a sufficiently fluid collection pace so results can be delivered within expected timelines.

In practice, we remain flexible: campaigns that fall slightly below this threshold can still be activated.

It is also important to note that 100K uniques is already a relatively low baseline, consistently achievable across all markets where Happydemics operates. For example, markets such as Poland regularly reach this level, confirming that this threshold is realistic and operationally reliable.

Probability calculations for CTV

For Connected TV (CTV), Happydemics collects IP addresses rather than device IDs. This is a deliberate choice adapted to the CTV environment: since a CTV device is linked to a household rather than an individual, the IP address is the most reliable way to connect exposure to survey responses across devices within that household.

This allows us to bridge the CTV exposure environment with the mobile/desktop survey environment at a household level, without requiring a retargeting pixel.

That said, without a retargeting pixel, we do not calculate the exact probability that a given respondent was exposed to the campaign. Instead, our methodology relies on a different principle: we measure whether respondents remember the ad.

This recall signal — combined with the household-level connection — is the core measurement unit of our methodology.

Respondent identification & speed filters

Unique respondents are identified using a combination of:

• Hashed IP address

• Declared age

• Declared gender

If multiple respondents share the same hashed IP, age, and gender, they are treated as a single respondent to prevent duplication.

Survey completion speed is tracked via server-side timestamps. Any respondent completing the questionnaire below the minimum time threshold is automatically excluded from the dataset.

Age & gender in a panel-free environment

Age and gender are collected as self-declared data directly within the survey flow.

This approach does not rely on third-party cookies, device fingerprinting, or probabilistic modelling, making it both privacy-compliant and robust in the context of evolving data regulations.

Respondent motivation & self-selection

Happydemics does not rely on incentivized panels. Respondents are not paid to participate.

Instead, surveys are designed as short, low-friction interactions integrated within the advertising experience — making participation feel natural and lightweight rather than intrusive.

The consistent ability to collect respondents at scale across all studies demonstrates the effectiveness of this model.

Regarding self-selection bias: because surveys are distributed within live advertising inventories (rather than panel databases), the respondent pool reflects real, active users across diverse digital environments, naturally limiting overrepresentation of specific profiles.

Ongoing quality audits

Happydemics operates exclusively within brand-safe environments and maintains data quality through several automated controls:

• Anti-bot algorithms that detect and blacklist fraudulent inventory sources

• A frequency cap of one impression per user per campaign, ensuring fresh and non-fatigued responses

• Internal quality filters excluding inconsistent or abnormal responses

• Automatic exclusion of any unverifiable data (e.g. consent issues, hashing failures)

These mechanisms are continuously reinforced and updated to remain compliant with evolving regulations (such as GDPR updates).

👋 If you don't find your answer, feel free to reach out to [email protected] for further consideration.