Demand dashboard: orders, trends and forecast gaps

A demand dashboard compares actual customer orders with the plan for the same product, location and period. Show the underlying rows, the prior-period trend and both signed and absolute gaps so offsetting errors cannot make an inaccurate plan look close.

Define the demand event before choosing measures

For an order-based review, specify whether actual demand means orders booked during the month or quantities requested for delivery during it. Those are different period assignments. Choose one for the plan and actuals, and show the date basis beside the results.

This example uses net booked orders for September 2026 at a fictional US distributor, Midwest Service Parts. Orders are assigned to Dallas or Columbus according to the receiving location requested by the customer. Product quantities are expressed in saleable units, with returns handled separately from new demand and cancellations removed from the booked-order total.

The Census Bureau's manufacturing definitions distinguish new orders from shipments and describe new orders net of cancellations. This example adopts that net-order principle for the distributor's internal review; it is not a reproduction of the manufacturing survey.

Booked orders still do not reveal all potential demand. A customer who never places an order because stock is unavailable leaves no booked quantity. Call the measure net booked orders, and investigate known stockouts separately before assuming the order history captures every customer need.

Prepare the actual, plan and comparison files

Bring the order-line export with order ID, line ID, product code, customer location, booking date, quantity, cancellation quantity and unit of measure. Bring a plan file with product, location, planning month, forecast quantity and the date the plan was approved. Keep a price reference if the review needs a dollar comparison.

Use the plan that was available before the reviewed month began. The forecast snapshot definition explains why that dated version matters. Replacing it with a later revision changes the question from how the original plan performed to how a newer view compares with known events.

Standardize product codes and location names before totaling. If one export uses cartons and another uses individual units, apply the documented conversion before computing gaps. Unmatched combinations belong on an exception list; dropping them would remove genuine misses from the result.

A worked product-and-location review

The following table contains the complete inputs for this example. August orders provide the previous monthly comparison. The reference price is held constant within each product across both locations and both September views, separating quantity differences from price changes.

Actual gap = Actual orders - Planned orders

Absolute error = Absolute value of Planned orders - Actual orders

Order growth = (Current orders - Prior orders) / Prior orders

September actual orders exceed the plan by 1,440 - 1,400 = 40 units. The month-over-month change is also 40 units because the August total equals the September plan. Growth is 40 / 1,400 = 2.86%, rounded to two decimal places.

The row results reveal a different level of variation:

For Dallas filter cartridges, growth is (600 - 550) / 550 = 9.09%. Planned value is 500 × 20 = $10,000, while actual value is 600 × 20 = $12,000. Apply the same formulas down the table; calculate company growth from the company quantities rather than averaging the row percentages.

The value gap is 46,800 - 46,000 = $800. It is a quantity comparison at reference prices, not invoiced revenue or a profit estimate. This distinction helps a commercial manager assess exposure without mixing price changes into the order-volume result.

Separate error size from its direction

The absolute errors sum to 100 + 60 + 40 + 40 = 240 units. Define weighted absolute percentage error, or WAPE, for this review as total absolute unit error divided by total actual units:

WAPE = Sum of absolute row errors / Sum of actual orders

Here, 240 / 1,440 = 16.67%. The modest net gap of 40 units therefore coexists with much larger misses beneath it. WAPE weights the saleable units in this example equally; it is not a dollar-weighted error score.

For directional bias, use plan minus actual, divided by actual. That sign convention gives (1,400 - 1,440) / 1,440 = -2.78%, indicating underplanning overall. NOAA's forecast-verification guidance uses forecast minus observed for error and explains that compensating errors can conceal misses in mean error. The same arithmetic distinction is useful here, even though NOAA's application is weather and water forecasting.

Use the existing demand forecast bias by product and site guide for a deeper directional review. Keep the dashboard focused on finding the combinations that need a conversation with the demand planner and commercial owner.

Change one planning input

Suppose the original Dallas filter-cartridge plan had been 600 units instead of 500. All actual orders, other plan rows, reference prices and August quantities remain unchanged. This is a comparison of alternative original plans, not a reason to rewrite the historical snapshot after results arrive.

The alternative reduces absolute error but changes the direction of overall bias to overplanning. It also makes the signed company unit gap larger in absolute size: the plan is 60 units above actual instead of 40 below. A single headline gap would miss that the row-level plan became closer overall.

Reconcile the inputs before reading the gaps

Actual quantities sum to 600 + 240 + 360 + 240 = 1,440 units and must match the net booked-order export under the stated September filters. Planned quantities sum to 500 + 300 + 400 + 200 = 1,400 units and must match the frozen plan. Retain all four product-location combinations when summing the 240 units of absolute error; exclude none simply because another row offsets its signed gap.

Where it goes wrong

• Using shipments as a substitute for orders. Fulfillment constraints can reduce shipments without reducing customer requests. Preserve the chosen order event throughout the review.
• Comparing different forecast versions. A revised plan can contain information unavailable when purchasing or staffing decisions were made. Keep the original snapshot and date visible.
• Letting positive and negative errors cancel. Inspect absolute errors alongside directional bias, using the same product-location rows.
• Ignoring zero actuals or missing combinations. A planned item with no orders still contributes absolute error. An absent file row needs investigation before it becomes a zero.
• Treating a forecast gap as a purchase instruction. Inventory, lead times and existing commitments belong in a separate supply decision. The demand review identifies a question; it does not settle every replenishment constraint.

The demand planner owns the plan version and definitions. Sales explains customer changes, while operations checks whether availability affected observed orders. The forecast-accuracy glossary provides the supporting distinction between a score and the business explanation behind it.

Compare your order and plan files in Covirage

Bring a sample of your own order export and dated demand-plan file to Covirage. Covirage maps the columns, and its tools compute and check the figures; the external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring the product, location and period rows behind your demand review.

See Covirage for forecast analysis for the supporting context.