Business wants forecasts, but historical data is unstable. How do you approach this?

< 1 min read

First, I would validate why the historical data is unstable by checking data quality, missing values, and any changes in business processes.
I would discuss the issue with stakeholders to understand if there were events like product launches, policy changes, or seasonal effects causing the instability.
Next, I would clean the dataset by removing obvious anomalies or separating abnormal periods if needed.
If the instability remains, I would consider using a shorter, more recent time window that better reflects current trends.
I would also test multiple simple forecasting approaches, such as moving averages or trend-based projections, instead of relying on complex models.
In one project, our sales data fluctuated due to a promotion campaign, so we built forecasts using post-campaign data only.
I would clearly communicate the limitations and assumptions behind the forecast to stakeholders.
I would present forecasts as ranges or scenarios rather than a single number.
Finally, I would recommend improving data consistency so future forecasting becomes more reliable.

Power BI Essentials