Businesses often insist on forecasting weekly, based on considerable demand variation from week to week. The belief is that weekly forecasting will be more accurate, will improve their replenishment planning, and will lead to a better managed overall inventory. But will it?
Weekly forecasting is appropriate where you have:
- Frequent short period promotions of less than a week.
- The bulk of items with short lead times and short replenishment cycles.
- Manufacturing is planned weekly with short planning horizons.
- Products have a very short shelf-life.
Now that we understand where weekly forecasting is appropriate, let’s look at reasons why weekly forecasting is not always such a great idea. Issues with weekly forecasting include:
- As seasonality could easily move to a different week (when comparing year on year sales by week), trying to predict the weekly seasonality becomes unreliable and extremely difficult.
- Managing and amending forecasts by week adds unnecessary complexity to your team’s workload.
- Performance measurement becomes much more difficult and time-consuming.
- Converting weekly data into monthly periods for reporting often results in a misalignment between the weeks and months.
- Period to date exception reporting becomes difficult.
Ultimately, creating planning forecasts by week leads to greater instability in the supply chain as we try to respond to weekly variations. It also invariably leads to more considerable strain on the demand planning team.
Forecasts by month are almost always more accurate than by week, as they flatten out the variations and pick up seasonality more reliably. Likewise, monthly replenishment forecasts are more stable and are not subject to over-reactivity.
Let’s look at an example or two:
Example A: if you import a widget on a 60-day lead time, you keep 30 days of safety or buffer stock, and you plan to receipt stock monthly, your total cover forward (horizon) looks like this:
30 days safety stock + 60 days lead time + 30 days cycle = 120 days
When you compute the recommended order quantity required today, you are looking at the forecast for the next four months. In this case, weekly variations will have little or no impact.
Example B: if you purchase washers locally on a seven-day lead time, you keep ten days safety or buffer stock, and you plan to receipt stock fortnightly, your total cover forward (horizon) looks like this:
10 days safety stock + 7 days lead time + 14 days cycle = 31 days
In this example, where an item with a very short lead time is ordered more frequently and you hold less safety stock when you compute order recommendations, you are looking one-month forwards.
Forecasting monthly but tracking daily
Forecasting monthly periods but tracking month to date sales versus forecast on a daily basis as you progress through the month, is a much more accurate way to plan.
Any significant variations between actual sales and forecasts will be highlighted by a structured exception reporting process, enabling your team to review and, if required, amend forecasts. They are able to apply their extensive market knowledge to make considered and beneficial forecast adjustments.
Since order recommendations can be created and viewed by day, the impact of weekly versus monthly forecasts on replenishment is minimal (especially with products that have a planning horizon of one month or more).
In summary
From a replenishment perspective, for the vast majority of products, it is far better to forecast monthly and let your team spend their valuable time (and considerable experience) managing any exceptions that occur via a structured exception reporting process.
Andy Hiscox, CEO, NETSTOCK Australia
