Delivery window uncertainty and satisfaction statistics

Updated 6 source citationsSources checked 13 November 2026

On this page
  1. Quick answer
  2. Key Takeaways
  3. What changes when the expected window becomes wider?
  4. Did wider uncertainty always reduce satisfaction?
  5. How can a merchant test a revised delivery message?
  6. Why can an on-time arrival still disappoint?
  7. Methodology and Sources
Download study text

Quick answer

Delivery window statistics show why widening an estimate can change expectations even when the midpoint stays the same. In a refrigerator delivery scenario reported in 2022, the median arrival time that felt neither satisfactory nor unsatisfactory was eight days after a six to eight day estimate, compared with seven days after a two to twelve day estimate. The randomized experiment involved 384 US adults, not real deliveries (Back and Spann, Oct 2022).

For a merchant answering an uncertain arrival question, the width of the window carries information of its own. A broader range can soften reactions to a poor outcome while also making a good outcome less rewarding. That tension matters when a support reply replaces a specific date with a broad range. Honest uncertainty and reassuring language need to work together, because the revised message changes what the shopper expects.

Key Takeaways

  • The two estimates had the same seven day midpoint. Their widths were two and ten days, an original calculation from the experimental conditions (Back and Spann, Oct 2022).
  • The wider window moved the neutral median one day earlier. The reported medians were seven days for the wide range and eight for the narrow range (Back and Spann, Oct 2022).
  • There were 384 people and 2,688 ratings. Each participant evaluated seven arrival scenarios, so ratings were not independent customers (Back and Spann, Oct 2022).

What changes when the expected window becomes wider?

The range expands in both directions. Its midpoint alone cannot describe the expectation a shopper receives, because a possible early arrival and a possible long wait can influence the same judgment.

Delivery window statistics from the refrigerator scenario reported in 2022: narrow estimate six to eight days with neutral median eight, wide estimate two to twelve days with neutral median seven. Both estimate midpoints are seven days.
Figure 1: The window became five times as wide while its midpoint stayed at seven days. The reported neutral median moved earlier. Source: Back and Spann, The Impact of Uncertainty on Customer Satisfaction, Oct 2022. Original analysis of the cited public source. Source periods and limits accompany the finding.

Why is the shared midpoint insufficient?

A merchant might treat both messages as equivalent because each centres on day seven. The first message concentrates the promised experience around that date. The second introduces an attractive early possibility and a much longer potential wait. Those are different propositions to someone who needs a replacement refrigerator.

The original comparison therefore does not support choosing a delivery window by averaging its endpoints. A useful message should let the shopper understand the range itself, the operational basis for it and whether anything has changed since checkout. If an estimate becomes wider because a carrier has lost certainty, simply retaining the old midpoint can conceal the information the shopper needs.

Does the neutral median mean a delivery target?

No. It describes the time respondents said would feel neutral in the stated scenario. It is neither a measured service level nor a threshold that guarantees satisfaction. It also does not mean that eight days is acceptable for every product.

That distinction matters for a consumable required tomorrow, an occasional accessory and a replacement appliance. The same elapsed time can have different consequences. The merchant's own delivery distribution and the shopper's deadline remain necessary inputs when applying the finding.

Did wider uncertainty always reduce satisfaction?

The result depended on the arrival outcome. The paper reports an interaction between uncertainty and departures from each person's neutral point, rather than a universal penalty for a wide window.

What did the regression compare?

In Table 1, the loss coefficient was negative 1.204 and its interaction with the wider range was positive 0.403. Adding them gives a wide range loss slope of negative 0.801. The gain coefficient was 0.517 and the interaction was negative 0.104, giving 0.413. These are original sums of published coefficients on the study's satisfaction scale, not percentage changes in customer satisfaction (Back and Spann, Oct 2022).

The practical tension is that cushioning disappointment can also dampen the benefit of an early result. A merchant should therefore evaluate the entire experience after an estimate change. Measuring only angry replies after late arrivals could miss a less enthusiastic response among shoppers who receive an order quickly.

What would contradict a simple window rule?

A wider range could be the truthful representation of a genuinely less predictable route. A narrow range could be misleading if fulfilment cannot support it. The experiment does not justify making promises more precise than the underlying evidence allows.

The alternative to unjustified precision is useful explanation. Say which milestone is known, what remains uncertain and when the estimate will next be updated. Whether that explanation improves satisfaction needs a separate test. It was not an arm of this experiment.

How can a merchant test a revised delivery message?

Keep the service offered constant while testing how its uncertainty is communicated. Otherwise a faster route, a different price or a changed customer mix could explain the result.

Which outcomes belong together?

Record the estimate shown at checkout, every revision, the actual arrival and the shopper's response. Distinguish a change in the midpoint from a change in width. Preserve the promised window in the record even when the tracking page later overwrites it.

Judge messages against arrival outcomes and repeat contact, as well as postdelivery satisfaction. Do not average away late failures with many early successes. An early parcel does not refund the disruption of a different shopper's missed deadline.

What should a delivery conversation explain?

Use the current shipment evidence to answer the question the customer actually has. A shopper asking whether an order can arrive before departure needs a bounded answer about that deadline, not reassurance based on an average. When the deadline cannot be supported, explain the available choices accurately.

For related operational context, see the historical analysis of delivery promises at carrier handoff and the distinction between dispatch and transit time.

Why can an on-time arrival still disappoint?

The two communicated ranges share a midpoint of seven days, but the reported neutral satisfaction benchmark did not occupy the same position within them. For the six-to-eight-day range it was day eight, the upper endpoint. For the two-to-twelve-day range it was day seven, the midpoint. Re-expressing those positions puts the neutral benchmark at 100% of the narrow range and 50% of the wide range. This is a description of the scenario results, not a probability model for real parcels. (Back and Spann, Oct 2022)

A broad deadline and the customer evaluative reference are different; being inside a window need not feel satisfactory.
Figure 2: A broad deadline and the customer evaluative reference are different; being inside a window need not feel satisfactory. Source: Back and Spann, Oct 2022. Original analysis of the cited public source. Source periods and limits accompany the finding.

The distinction helps explain why a deadline-only service report can miss the customer's experience. An arrival on day nine would be beyond the narrow range but still inside the wide range. That classification alone does not say it feels satisfactory. In the wide-range scenario, the median reported neutral point was already day seven. Being inside a broad promise and meeting the customer's evaluative benchmark are different conditions.

Nothing in that example requires assuming deliveries are uniformly distributed across the quoted range. A two-to-twelve-day estimate does not establish that every date is equally likely, nor that day seven is a measured arrival median. The source manipulated the communicated range in a hypothetical refrigerator task. It did not observe a delivery distribution from which to infer those probabilities.

For a merchant, three fields deserve separate treatment: the promise shown before purchase, the most recent order-specific estimate and the actual arrival. Preserve revisions and the time they were communicated. A wider later estimate may improve the chance that the record meets its newest deadline while leaving the original expectation unresolved.

This also changes the question a conversational answer should resolve. If the shopper needs an item for a particular event, quoting a very broad valid range may be insufficient. Establish the required date and explain only the delivery option the actual order and service support. If the range has changed, describe the change clearly instead of presenting it as though it was the original commitment.

The scenario evidence does not justify narrowing every estimate or promising a date the operation cannot support. It shows that range width, midpoint and the customer's satisfaction reference are distinct. A current test should evaluate truthful expectation setting and the subsequent experience together, rather than treating a higher on-time percentage under a wider window as proof of improved service.

Methodology and Sources

We transcribed the experimental conditions, medians and Table 1 coefficients from Back and Spann's October 2022 working paper. We calculated window widths, their ratio, the shared midpoint and sums of main and interaction coefficients. No participant data were reanalysed. The 2,688 observations are repeated ratings by 384 respondents; the source clusters regression standard errors by person. The scenario concerned a refrigerator that had already been purchased. It does not measure actual checkout conversion, carrier performance or repeat purchasing. The paper's publication date identifies this source version; the fieldwork calendar is not reported in the inspected experimental section.

Glossary

  • Delivery window: The earliest and latest arrival dates communicated together.
  • Midpoint: The average of the two endpoints.
  • Window width: The distance between those endpoints, not the number of inclusive calendar labels.
  • Neutral point: A reported outcome that feels neither satisfactory nor unsatisfactory.
  • Loss: An arrival later than an individual's neutral point in this model.
  • Gain: An arrival earlier than that point.
  • Interaction: A model term indicating how one relationship changes across conditions.
  • Repeated rating: Another judgment by the same person, not another independent person.

Sources

Cite this page

Use the citation below for this analysis, preserving the original source and period.

Convo Rescue. "Delivery window uncertainty and satisfaction statistics." Convo Rescue, November 2026. https://convorescue.com/guides/delivery-window-uncertainty-and-satisfaction-statistics/
<a href="https://convorescue.com/guides/delivery-window-uncertainty-and-satisfaction-statistics/">Delivery window uncertainty and satisfaction statistics</a> (Convo Rescue, November 2026)