This is a creative fatigue analysis (e.g., Hill function modeling) for user acquisition for finance app in python using advanced machine learning techniques.

Creative fatigue refers to the phenomenon where an ad’s effectiveness (e.g., click-through rate, conversion rate) declines with repeated exposure. It’s important to identify and quantify this to optimize user acquisition campaigns.

 

Modeling Approach

1. Data Requirements

Input:

  • creative_id

  • impressions per time period (daily/hourly)

  • clicks, installs, cost

  • timestamp

  • audience_segment (optional but useful)

  • platform (iOS/Android)

  • campaign_type

2. Step-by-Step Pipeline

Step 1: Preprocessing

Aggregate by creative and day, then calculate:

  • CTR = clicks / impressions

  • CPI = cost / installs

  • Conversion Rate = installs / clicks

Step 2: Hill Function Modeling (Fatigue Curve)

The Hill equation is:

y=Vmax⋅xnKn+xny = \frac{V_{\text{max}} \cdot x^n}{K^n + x^n}y=Kn+xnVmax​⋅xn​

Where:

  • y: performance metric (CTR or conversion rate)

  • x: number of impressions or time (proxy for fatigue)

  • Vmax: maximum performance

  • K: half-max point (fatigue threshold)

  • n: Hill coefficient (steepness of dropoff)

Step 3: ML Layer

  • Use Regression with engineered features (e.g., impressions, past performance, time since launch).

  • To estimate confidence intervals on fatigue thresholds.

Insights:

1. Optimal Impression Threshold for Maximum ROI

The Hill function identifies a clear saturation point (K) where additional impressions yield diminishing returns. For the finance app, this means you should limit ad frequency per user to around this threshold to avoid wasting budget on fatigued creatives that no longer drive conversions efficiently.

2. Creative Refresh Timing to Sustain Engagement

By understanding the steepness parameter (n), you can gauge how quickly users become fatigued. If the curve drops sharply after a small number of impressions, it signals a need to rotate or refresh creatives more frequently to keep user interest and acquisition rates high in a competitive finance market.

3. Maximizing Conversion Potential by Balancing Spend and Fatigue

The maximum performance parameter Vmax⁡V_{\max}Vmax​ represents the peak conversion achievable. Using this, you can forecast the maximum attainable user acquisitions per campaign and adjust spend allocation dynamically—investing more in creatives or channels before fatigue hits to maximize conversions without overspending on tired ads.