Pickup vs Delivery Metrics: Check the Fulfillment Mix

By Voosh

When a restaurant’s overall cancellation rate or average order value changes, separate customer pickup from delivery before changing operations. First compare each fulfillment type with its own earlier result. Then calculate what the overall result would look like if the fulfillment mix had stayed the same.

This check helps answer a specific question: did performance change within pickup or delivery, or did the restaurant simply receive a different combination of orders?

Fulfillment type is a separate question from ordering platform

An order’s platform tells you where it originated. Its fulfillment type tells you how it reaches the customer. A platform total can contain more than one fulfillment type.

DoorDash’s merchant pickup guidance describes pickup as an order the customer collects and explains that pickup appears alongside delivery in reporting. Its Marketplace order FAQ documents an integration field distinguishing Dasher delivery, merchant delivery, and customer pickup.

That documentation does not prove the same field is available in every restaurant export, analytics connection, or account. Inspect the data you actually receive. If a report says only “online orders,” do not infer fulfillment type from the restaurant, basket size, or customer name.

For the analysis below, customer pickup means customer collection. Delivery means the delivery categories included in the restaurant’s chosen scope. Separate merchant delivery from platform delivery when their processes differ enough to affect the decision and the data supports it.

Write down the numerator, denominator, and population

Before comparing two periods, define one metric and preserve the same definition in both.

For this article’s cancellation example:

Cancellation rate = canceled accepted orders ÷ accepted orders × 100.

Both counts belong to the same accepted-order cohort. Use the acceptance date to assign the period and allow enough time for those orders’ outcomes to settle. Do not combine cancellations recorded this week with a different set of orders accepted this week.

Keep the following consistent:

• Stores and ordering platforms included.

• Fulfillment-type mapping.

• Start and end dates, time zone, and daypart scope.

• Treatment of test orders, duplicates, and orders without a final outcome.

• Definition of cancellation and accepted order.

• Treatment of unknown or missing fulfillment types.

Preserve an unknown category rather than quietly assigning missing records to delivery. Show its count and share. If missingness could change the conclusion, resolve the classification before acting on the comparison.

A hypothetical example: the total worsens while both types improve

The following numbers are hypothetical and designed to explain the arithmetic. They are not restaurant benchmarks or customer results.

Period A

• Pickup: 800 accepted orders, 16 cancellations. Rate = 16 ÷ 800 × 100 = 2%.

• Delivery: 200 accepted orders, 12 cancellations. Rate = 12 ÷ 200 × 100 = 6%.

• Combined: 1,000 accepted orders, 28 cancellations. Rate = 28 ÷ 1,000 × 100 = 2.8%.

• Order mix: 80% pickup and 20% delivery.

Period B

• Pickup: 200 accepted orders, 2 cancellations. Rate = 2 ÷ 200 × 100 = 1%.

• Delivery: 800 accepted orders, 40 cancellations. Rate = 40 ÷ 800 × 100 = 5%.

• Combined: 1,000 accepted orders, 42 cancellations. Rate = 42 ÷ 1,000 × 100 = 4.2%.

• Order mix: 20% pickup and 80% delivery.

The combined cancellation rate rose from 2.8% to 4.2%, an increase of 1.4 percentage points. Yet pickup improved from 2% to 1%, and delivery improved from 6% to 5%.

More orders moved into the type with the higher cancellation rate. That change in composition outweighed the improvement within both types.

A manager looking only at the total could conclude that both workflows deteriorated. The type-level results support a more precise question: why did the delivery share grow, and what explains the remaining delivery cancellations?

Hold the mix fixed to isolate the comparison

Use Period A’s mix as a reference and apply Period B’s type-level rates to it:

Fixed-mix Period B rate = (80% × 1%) + (20% × 5%) = 1.8%.

Compare that with Period A’s actual rate of 2.8%. At the earlier mix, the new type-level rates would produce a rate 1.0 percentage point lower.

Next compare Period B’s actual rate, 4.2%, with its fixed-mix rate, 1.8%. The difference is 2.4 percentage points associated with the change in mix under this calculation.

The two pieces reconcile:

Within-type component: −1.0 percentage point.

Mix component: +2.4 percentage points.

Total change: −1.0 + 2.4 = +1.4 percentage points.

This is a descriptive decomposition, not proof that fulfillment type caused the cancellations. Menu changes, promotions, staffing, order size, and other factors may have changed too. The chosen reference mix also affects how the total change is divided between components. State the reference period and keep it consistent.

Use a small worksheet before a big operating response

For each fulfillment type, record:

1. Period A accepted orders and cancellations.

2. Period B accepted orders and cancellations.

3. Each period’s type-level cancellation rate.

4. Each period’s share of accepted orders.

5. Period A share multiplied by Period B rate.

Sum the fifth value across types to get the fixed-mix rate. Reconcile the sum of the type counts to the overall count before interpreting anything.

For a metric such as average order value, use the appropriate order population and consistent sales definition instead. If average order value is sales divided by completed orders, weight type-level averages by completed-order shares. Do not reuse accepted-order shares from a cancellation worksheet.

Do not take a simple average of two rates unless the relevant order counts are equal. And do not calculate percentage change when the baseline is zero; report the counts and percentage-point difference instead.

Match the next action to the pattern

The total worsened, but each type improved. Investigate why the mix changed and where the remaining issues occur. Preserve the within-type improvements while reviewing the higher-issue workflow.

The total improved, but one type worsened. Do not let the combined result hide a pickup or delivery problem. Review the affected type’s cancellation reasons, counts, and operating context.

Both types worsened. Look for a shared change, such as a menu rollout or staffing disruption, while still checking whether the reasons differ by type.

Counts are small or classification is incomplete. Show the limitation. A descriptive calculation on a small sample is still arithmetic, but it is not dependable evidence of a lasting operating change. Use a longer comparable window where appropriate and inspect individual cases.

None of these patterns, by itself, justifies steering customers away from one fulfillment type. The commercial decision also needs demand, customer experience, cost, and operating context.

Put the result beside the ordinary scorecard

A short review note should contain the actual combined rate, the pickup and delivery rates with counts, the fixed-mix result, and one question to investigate. Keep the actual rate visible because it describes the restaurant’s real order population. The fixed-mix view explains the comparison; it does not replace the actual outcome.

For broader location comparison, see Restaurant Benchmarking: Compare Delivery Locations Fairly. Use that framework after checking whether a fulfillment-mix change is distorting the headline trend.

Voosh’s VooshGPT page describes sales and operating analysis across connected store, channel, and daypart data. Confirm that your account provides the fulfillment classification and counts this worksheet needs. This article does not claim a native Voosh fixed-mix calculator or universal pickup-versus-delivery reporting.

Book a Voosh demo to discuss the operating questions you want your restaurant data to answer.