ML Trends (forecasts)

ML Trends uses machine learning to predict future data by using historical data about your business.

We use this to forecast certain key business metrics such as:

We plan to support more forecast types over time, but are currently prioritising forecast accuracy before expanding to other statistics

Where can I see these forecasts?

These forecasts are available on the Mega Report if you have ML forecasting available on your plan. If you can't see them then it's possible that they are hidden in your current view, you don't have permission to see those rows, or your account has not been set up correctly. 

All ML Forecast rows can be identified with the "AI Generated Forecast" icon and typically start with 'ML Forecast'. The only available dropdown on these rows is the Location dropdown.

How often are forecasts generated?

Forecasts are automatically generated for the next 12 months at the start of every month. Additionally, a special control panel to generate updated forecasts is available to users on some plans. 

Generating forecasts more frequently than once a month is not as insightful as it may appear because even a "record setting" great week or "worst we have ever had bad week" is only one week of data that contributes to a forecast that is potentially looking at multiple years of historical business performance. 

Forecast Tagging ('quarterly' forecasts)

By default each forecast row shows the latest data we have forecast for that metric. For example, as you get closer to December the ML forecast will be updating monthly (and when you trigger it) to ensure that you have the most accurate forecast based on the most up-to-date data. 

Some of our clients also like to "lock in" a forecast, typically for a quarter or year so they can then measure actual business performance against the original forecast. For this we have Forecast Tags. This allows you to tag a specific forecast run with a label. In the screenshot below we have generated a forecast on October 1st, 2023 and then created the Quarterly tag. This allows us to add a second set of ML forecast rows to the Mega Report (as seen in the above screenshot):

• ML Forecast Revenue: our latest (most accurate) forecast for this period
• ML Forecast Revenue Quarterly: the most recent forecast containing the 'Quarterly' tag (aka. the 'locked in October 1st) forecast. 

This tag can be added by any manager accounts by visiting Company Settings -> ML Forecasts then creating an appropriate tag.

Can all data be forecast?

The short answer is "No". Our forecasting looks for trends across multiple seasonalities (such as day-of-week, month-of-year, etc) plus an underlying growth rate. This is not useful when forecasting erratic values or ratios such as Hours Per Client, which has been demonstrated to vary month-to-month in most of our clients without a clear trend.

How is the trend calculated?

The BOS forecasts time series data based on an additive model where non-linear trends are fit with yearly, weekly, and daily seasonality, plus holiday effects. It works best with time series that have strong seasonal effects and several seasons of historical data. We only use 'completed' data currently (IE: billed hours) and don't currently look at future data (IE: booked hours for future periods).

In simple terms: it looks at patterns in weekly, monthly, yearly business data to estimate future performance. Generally speaking it is more 'long term' focused so if your business has been growing steadily month-on-month for years and then you have one bad month then the trend won't 'nose dive' until it has seen that downward trend over a longer period of time.

Questions and Feedback

This is new technology that is constantly evolving. If you have any questions, feedback or suggestions please Contact Us and let us know.