How Does the Estimator Actually Work?
If you’re wondering how the audience estimator calculates numbers—or why you might see different actual results—it’s important to understand that the estimator makes a best-effort prediction based on historical data and statistical modeling.
The estimator works in two stages:
1️⃣ Initial Volume Estimation
The model estimates the total daily volume of people who fit into a given audience segment.
This is based on past behavior and considers:
The number of topics and how they overlap
The time window (how many days are being searched)
The frequency of engagement (how many times an action occurs)
2️⃣ Filtering & Match Rate Calculation
The system runs additional filters on a random sample of people who fit the initial criteria.
From this, a match rate is calculated, which is then applied to the total volume estimate.
The daily estimate is then used to estimate the monthly volume, factoring in how likely people are to move in and out of the segment
How Does Changing Frequency, Topics, and Duration Affect Volume?
Increasing Duration (Time Window) ⬆️ → More Volume
Increasing Number of Topics ⬆️ → More Volume
Increasing Frequency (Number of Events Required) ⬆️ → Lower Volume
Note: These relationships are non-linear—for example, a 14-day duration does not necessarily return 2x the volume of a 7-day window. The estimator counts unique individuals, not total events.
Why Is the Audience Estimate Off? It Predicted X, But I Got Y People.
The audience estimate is based on historical data and typical search volumes for topics and filters. Several factors can cause deviations between estimated and actual results:
✅ Topic Popularity Fluctuates – Some time periods see spikes in engagement (e.g., seasonal trends, industry news).
✅ Behavioral Changes – External events, marketing campaigns, or competitor activity can shift audience behavior.
✅ Sampling Limitations – The estimator uses a subset of data to predict total audience size, and variations occur naturally.
While we strive to provide the most accurate predictions possible, real-world factors always introduce some variability.
🚨 Why is the Audience preview slow to load? Large audiences or segments with many conditions can be slow to load. Don't worry, you can continue editing your segment even while the preview is loading. 😅