Seeing a low confidence score (e.g. 34%) in Databox forecasts even after collecting over two years of data can be confusing, especially when you're relying on forecasts for budgeting or performance planning.
Why this happens
Databox forecasting uses Facebook’s Prophet algorithm. Your score is based not just on how much data you’ve collected, but also on how stable and predictable that data is. The biggest factors are:
Historical data volume (max 24 months is used)
Volatility in your data (e.g. spikes in ad spend)
Seasonality
Forecast range and granularity
For example, errratic spending patterns in Meta and Google Ads can make it hard for the model to detect consistent trends.
What to try if your forecast score is low
If you’re seeing low forecast scores, try these adjustments:
Shorten the forecast range. Choose 1–2 months instead of 3+.
Switch granularity to “Monthly”. Avoid “Yearly” if the data varies a lot.
Clean up anomalies. Remove extreme spikes or gaps that aren’t representative.
Stabilize your inputs. More consistent patterns in ad platforms improve model accuracy.
Even with complete data, unpredictable behavior can lower your score. These tips help the model generate a more confident projection.
For more on how forecasts work, visit: How to set up forecasts in Databox