Quick Summary: While most forecasts function automatically, situations like promotions, lifecycle changes, or large variances require planner judgment. Recognize when system forecasts diverge from reality and utilize specific tools to decide when intervention is necessary.
Why Forecasts Sometimes Need Intervention
Netstock’s statistical models correctly forecast the majority of stocked items.
However, no system can anticipate everything that affects demand. Business events, unusual trading conditions, and data irregularities can all create gaps between algorithmic forecasts and the real world.
Planner intervention adds value because:
The system cannot know about business decisions such as upcoming promotions or lost customers.
Some products lack sufficient history for reliable statistical modeling.
Certain items go through lifecycle or mix changes that algorithms can’t predict.
Human review ensures forecasts remain realistic, actionable, and aligned with current business insight.
Common Situations Where Forecasts Need Help
End-of-Lifecycle Items
When an item is being phased out, the forecast must be reduced to match the exit plan. If you leave the system forecast unchanged, the app will continue generating replenishment recommendations, creating unnecessary stock and write-offs.
What to do: Lower the forecast gradually until the final month of sale. Confirm that safety stock also winds down in line with the depletion plan.
Promotions
Promotions produce short-term sales spikes that the forecast engine cannot anticipate from historical data. If you are planning a discount, marketing campaign, or bundle deal, the forecast should be increased ahead of time so stock is available when the event begins.
Best practice: Make these adjustments at least one full lead time before the promotion starts so purchase orders can arrive in time. After the event, return the forecast to normal to prevent inflated replenishment.
New or Lost Customers
Customer changes are one of the most common reasons for forecast error.
A new customer increases demand immediately. Without manual correction, the system will under-forecast and risk stockouts.
A lost customer reduces demand sharply. If you don’t lower the forecast, Netstock will continue ordering as if the customer still exists.
Without manual adjustments, replenishment will continue based on outdated demand patterns.
New Items
Brand-new items have little or no sales history, so the app cannot generate a meaningful forecast. Leaving them unadjusted means the Recommended Order Quantity (ROQ) will remain zero.
What to do:
Start with a reasoned estimate based on market knowledge or similar products.
Use the Supersessions tool (Refer to article: Supersessions Explained) to transfer history from a discontinued item to its replacement, creating a realistic starting point.
Review forecasts for new items frequently and adjust as sales data accumulates.
Large Forecast Variance
Large variances occur when sales and forecasts differ significantly. Whether this is a problem depends on context.
Seasonal items, such as umbrellas or sunscreen, often show large forecast variances when comparing short timeframes, but the forecast is still accurate when viewed across the full cycle. In this case, the variance is expected and not a concern.
Consistent items, such as soap or a generic pharmaceutical, should track closely to the forecast. A large variance here is a red flag that the forecast is misaligned.
💡 The Sales exceeds forecast and Forecast exceeds sales variance panels highlight where action may be required.
The Variance Panels
Variance panels help you quickly identify forecasts that may need attention:
Sales Exceeds Forecast: Items where sales are running higher than forecast (risk of stockouts).
Forecast Exceeds Sales: Items where sales are lagging behind forecast (risk of excess stock).
These panels allow you to manage by exception, instead of reviewing every SKU.
How To: Interpret The Variance Panels & Reports
⚠️ Watchouts
Seasonal Variances: Not every variance needs to be corrected. Seasonal products may look off month to month but be accurate over a longer horizon.
Ignored Business Events: Ignoring events such as promotions or lost customers can cause the system to over-order or under-order significantly.
Documentation: Adjustments should be documented so other users understand why changes were made.
💡 Tips
Prioritize Exceptions: Focus on items flagged by variance panels or exceptions first, rather than manually checking every item.
Act Quickly: Use the month-to-date forecast exception reports to identify and act on problems quickly.
Collaborate with Teams: Collaborate with Sales and Marketing to align forecasts with planned business events.
New Items: For new items, use Supersessions if appropriate to avoid starting from zero.
Forget about these 👇 😞 😐 😃 Have your say here!