In Alvie, you have two methods for modeling revenue: Daily Average and Separate Models. Each approach has its own strengths, and the best choice depends on how your business operates.
This guide explains the difference between the two and will help you decide which one is right for you.
Method 1: Daily Average (Recommended)
This method first runs a model to predict conversions. It then calculates revenue by multiplying the predicted conversions by your average order value (AOV) over a recent window.
Predicted Revenue = Predicted Conversions × Recent AOV
It's a fast and straightforward approach that is easy to interpret.
This method is a good fit if:
Your average order value (AOV) is generally stable.
You don't run frequent, large-scale promotions that cause AOV to swing dramatically.
You favour speed and simplicity in your modeling.
We generally recommend this option.
Method 2: Separate Models
This method runs two distinct models at the same time: one that predicts conversions and a second, separate model that predicts the revenue associated with those conversions.
This approach is more nuanced because it understands that revenue can fluctuate independently of conversion volume. For example, it can capture how a "50% Off" promotion drives conversions with a lower AOV, while a "Luxury Edit" campaign might drive fewer conversions but at a much higher value.
This method is a good fit if:
Your AOV is volatile or heavily influenced by seasonality, promotions, or tiered pricing.
Different marketing channels or campaigns attract customers with very different spending habits.
You need the highest possible accuracy for revenue optimisation, even if it adds complexity.
Be aware though that this also requires higher data volumes per channel to get results that make sense!
Quick Decision Guide: Key Differences
| Daily Average Model (Recommended) | Separate Models |
How it Works | One model:
| Two models: one for conversions, one for revenue. |
Best For | Stable, predictable AOV. | Volatile, campaign-driven AOV. |
Accuracy | Accurate when AOV is steady. Can be biased if AOV changes. | More accurately captures revenue shifts that diverge from conversion trends. |
Sensitivity | Smoother, less sensitive to daily revenue fluctuations. | More sensitive and responsive to real-time revenue changes. |
Potential Pitfall | Can miscalculate value if a new campaign attracts unusually high or low-spending customers. | Can be noisy or less stable if your revenue data is sparse or inconsistent. |
Practical Recommendation
If you're unsure where to start, begin with the Daily Average method. It's robust and easy to understand.
You should consider using Separate Models if you operate in a promo-heavy environment (like e-commerce with frequent sales) or if your business strategy involves segmenting customers by value.
You can always run both and compare the results to see which provides a more stable and believable profile for your business.