Doing a sales or project monthly forecasting is a good option to keep track of the latest evolution of the activity. But why doing a monthly forecast and not a quarterly or yearly? What does a monthly forecast require? Before jumping in and going through this post, you might want to read everything about sales forecasting.
But first, what is a 'monthly forecast' ?
You should make the distinction between frequency and garnularity and level of aggregation when it comes to name a forecast type. A "monthly forecast" could be different things:
- ... a forecast made with daily data and then aggregated at a monthly level?
- ... a forecast made with monthly level data for the next 12 months?
- ... a forecast made every day for the next 30 days (rolling forecast)?
This post is defining a 'monthly forecasting' as a forecast made for the next 30 days or next calendar month using daily level aggregated data.
Even if you need to report to your management only a single aggregated number (E.g. the aggregated forecast for March will be $100M), you will need to have analyzed the various possible effects at a daily level, if you want to be accurate.
The forecasting process should always imply working with data at a daily level (if possible) and projecting at a daily level. Then, you will be able to consolidate the daily forecast obtained into weeks, months, quarters and years. Daily aggregate data makes it possible to choose the level of aggregation of your forecast suiting your needs. Trying to break a monthly aggregated forecast into a weekly one is almost impossible and not a common practice. Same if you have a forecast already aggregated at a weekly level and you try to transform it into a monthly forecast.
So why doing a monthly forecast (and not a quarterly or yearly one)?
You have a fast-growing project
Sometimes, the project you are working on might be in a launch phase: you just realeased a new product or service or opened a new country. One single marketing campaign can boost the overall performance and be a game changer and you might need to align ressources such as call center or support. As such, you might want to focus only on the next 30 days to forecast the performance.
You aim at precision and accuracy
Forecasting over a time frame of 30 days only is a good option to get an accurate forecast, especially if you work with time series. The further in time a forecast will be, the more likely your forecast will be wrong or not 100% accurate and the level of incertainty increases.
On the contrary, the closer to the present, the better accuracy will be: long term trends and latest evolutions will be correctly picked up by the time-serie algorithm forecast used, provided you implement all the correct "special events" (such as any bank holiday or a special trade fair happening every year) for the next 30 days.
You need to compare the latest trend with your central yearly forecast
Monthly forecast helps driving operational and tactical actions
Given the latest trends observed in the data, do you think that the next 30 days will be inline with your overall yearly plan? Yes - you might want to double down on effort to beat forecast or change nothing. No - short term actions will need to be made to meet your targets.
Doing a monthly forecast will help you to steer your business and implement short-term tactical actions in case you anticipate any unexpected variation.
You need to give an outlook to your management
Finance or the management team might need your outlook on your project results for the next calendar month. Just to get a feel at where you're heading. Doing a monthly forecasting then makes sense - but be quite daunting.
How to make a monthly forecast?
You need data at a daily level
Don't go more granular unless necessary.
Data overflow is never a good thing and including hourly variations might not be relevant. Daily level data when doing a monthly forecast helps capturing weekdays versus weekend patterns.
Monthly forecasting building will logically imply to get a good and robust business intelligence system with a good data management platform. The data captured need to be properly stored, easy to get access to and in the right format required.
Choose between calendar month or 30 days rolling period
You have two options.
First, you could run a monthly forecast every month at more or less the same date and focus on the next calendar month (Jan, Feb,...). This might be great to liaise internally with other departments, such as finance, analytics, logistics or the board of directors.
Second option, you might want to run every day or every week a forecast for the next 30 days. A rolling monthly forecast. This is typically something you would need to do for instance to forecast demand and price correctly your product. If you rent a car, demand for the next 30 days will be forecasted for each pick up location (or store). This will ensure that the pricing and car fleet availability will be ok to meet demand. You will also get the benefit of doing a rolling monthly forecast by comparing forecast D-7 and D by understanding what are the latest trends.
You need to automate the monthly forecasting process
If you plan to run a monthly forecast, you might want to work with time series and have a look at our basic forecasting solution. You need some tool or routine (Excel template for monthly forecasting, or an algorithm being executed on schedule) and a standard data visualization to make it work.
You should work on making your forecast accurate and not on updating it.