# Exponential Moving Average

**Acronym:** EMA  
**Category:** metrics  
**Short Description:** Weighted moving average giving more importance to recent data.  
**Last Updated:** 2026-07-07T00:00:00Z

## Definition

An exponential moving average is a type of moving average that places greater weight on more recent data points, making it more responsive to recent changes while still smoothing out noise. Each period's EMA blends the newest value with the previous EMA, so older data fades exponentially rather than dropping out abruptly. This is particularly useful for metrics that require faster reaction to changes, such as fatigue inflections or post-launch drops.

## Formula

**Formula:** `EMAₜ = α × Valueₜ + (1 − α) × EMAₜ₋₁`

A weighted running average where the smoothing factor α sets how fast old data fades — higher α reacts faster, lower α smooths harder. The result stays in the same units as the metric being smoothed.

## Calculation

**Formula:** `EMA = (Current Value × Smoothing Factor) + (Previous EMA × (1 - Smoothing Factor))`

**Explanation:** Combines the current value and the previous EMA, with the smoothing factor α determining the weight of each. α is commonly set as 2 / (n + 1) to approximate an n-period window — e.g., α = 0.25 behaves like a 7-day average. The result stays in the same units as the metric being smoothed.

### Components

- **Current Value**: The metric value for the current period
- **Previous EMA**: The exponential moving average value from the previous period
- **Smoothing Factor**: The weight given to the current value compared to the previous EMA

## Examples

- With α = 0.25, yesterday's EMA of CPA at $50 and today's CPA of $60: EMA = 0.25 × 60 + 0.75 × 50 = $52.50 — the spike registers immediately but is damped
- EMA of conversion rate to quickly detect performance drops
- Bid adjustments based on EMA of CPC trends

## How AdSights Helps

**Tracking Exponential Moving Average:** Fatigue and performance breaks show up in the newest data first, which a simple average dilutes with weeks of history. AdSights leans on recency-weighted trends when reading each variant's trajectory, so a hook that stopped working this week is flagged this week — while the damping still filters out one-off bad days.

## FAQs

### How do you choose the smoothing factor for an EMA?

Use α = 2 / (n + 1) to make the EMA comparable to an n-period simple average: α = 0.25 approximates a 7-day window, α ≈ 0.067 approximates 28 days. Raise α when reaction speed matters (launch monitoring, fatigue detection), lower it for stable baselines. If the EMA whipsaws on every noisy day, α is too high; if it confirms declines weeks after your team spotted them, it's too low.

### When should I use an EMA instead of a simple moving average?

When catching turning points quickly is worth some extra jitter: creative fatigue onset, conversion drops after a landing-page change, CPC shifts after a bid change. A simple moving average of window n lags a trend change by roughly (n − 1) / 2 periods — about 3 days for a 7-day window — while an equivalent EMA responds sooner because the newest days carry the most weight — with α = 0.25, the last three days carry over half the signal. For steady reporting baselines, the simple average wins.

### How is an EMA used to detect creative fatigue?

A pattern borrowed from technical analysis: track a fast EMA (α ≈ 0.4, roughly 4-day) and a slow one (α ≈ 0.067, roughly 28-day) of a creative's CTR or thumbstop rate. While performance is stable the two lines travel together; when fatigue sets in, the fast line crosses and stays below the slow line — a persistent-decline signal that filters out single bad days and pairs well with a frequency check.

## Related Terms

### Similar Terms

- **[Moving Average](/resources/glossary/metrics/moving-average)**: Standard moving average provides stable baseline while EMA adapts faster to changes
- **[Anomaly Detection](/resources/glossary/metrics/anomaly-detection)**: EMA baselines adapt to drifting metrics, keeping anomaly thresholds current

### Component Terms

- **[Statistical Significance](/resources/glossary/metrics/statistical-significance)**: Validates whether EMA changes represent genuine performance shifts versus random variation
- **[Standard Deviation](/resources/glossary/metrics/standard-deviation)**: Establishes adaptive thresholds for EMA-based monitoring

## Related Resources

- [Moving Average Interactive Example in AdSights Creative Fatigue Article](/blog/topics/creative-strategy/creative-fatigue-in-meta-ads-detection-and-management-strategies) - Demonstrates how moving averages help identify creative fatigue patterns in advertising performance