Why the Parcel Did Not Get In: How an Online Grocery Retailer Used IDIs and Delivery Analytics to Diagnose a Failed Delivery Rate That Was Costing More Than the Delivery Itself

Executive Snapshot

Client

Online Grocery Retailer, Netherlands

Situation/Challenge

The client's last mile delivery operation was recording failed delivery rates in a cluster of dense urban postcodes that were running at nearly twice the network average, driving re-delivery costs and customer compensation that collectively exceeded the margin on those orders. The operations team had attributed the failures to address access problems in older urban housing stock, but customer service records showed a different pattern: many customers in the affected postcodes were complaining that they had been home and received no delivery attempt. The gap between what the operations team believed and what customers were reporting had not been resolved.

Objective

Conduct structured IDIs with customers in the high-failure postcodes who had experienced failed deliveries, paired with delivery attempt analytics examining driver GPS patterns, delivery time window adherence, and doorbell event data across the affected postcode cluster.

Constancy Researchers Solution

Primary Research & VoC through In-Depth Interviews (IDIs) combined with Data Analytics & Business Intelligence, 36 IDIs with customers in high-failure-rate postcodes who had experienced at least one failed delivery, paired with analytics examining driver GPS trace, time window adherence, attempt duration, and doorbell sensor data across all deliveries in the affected cluster.

Impact

IDIs revealed that a significant share of customers reporting failed deliveries had been home and had either not heard the doorbell or had heard it after the driver had already left. Analytics confirmed that average attempt duration in the affected postcodes was markedly shorter than the network average, and that a subset of drivers showed GPS traces inconsistent with a genuine access attempt before logging a failed delivery.

Client Outcome

The client introduced a minimum attempt duration standard with dispatcher monitoring, added a customer SMS alert at the fifteen-minute pre-delivery mark, and implemented driver-level delivery success rate tracking. Failed delivery rates in the affected postcodes halved within three months.

The Situation / Challenge

A failed delivery rate that runs at twice the network average in a specific geography is telling you something precise. The question is what.

The operations team’s working explanation was the first: access difficulty in older Amsterdam housing. Customer service records complicated that story.

The cost was mounting on two fronts: re-delivery operational cost and customer compensation credits, collectively pushing the unit economics of those postcodes into negative margin territory on a proportion of orders large enough to matter.

Key Challenges

  • No direct customer research establishing whether failed deliveries in the affected postcodes reflected genuine access difficulty, time window misses, or inadequate attempt quality.
  • An operations team working explanation of access difficulty that was contradicted by a significant volume of customer service complaints from customers who reported being home.
  • No delivery attempt analytics examining driver GPS traces, attempt durations, and doorbell event data to distinguish genuine from inadequate attempts.
  • Re-delivery and customer compensation costs pushing unit economics negative in the high-failure postcode cluster.
  • A gap between operations and customer service explanations that had not been investigated at the data level.
  • Operations leadership pressure to explain the failure rate differential and implement a lasting operational remedy before the summer peak season.

A failed delivery is a cost event that carries within it a question the failure rate alone cannot answer: did the driver try? The answer matters enormously because access-difficult genuinely failed deliveries and inadequate-attempt failed deliveries look identical in the failure rate metric but require entirely different operational responses to reduce.

Constancy Researchers Solution

Constancy Researchers ran the IDI and analytics workstreams in parallel, using IDI findings to guide which dimensions of the delivery attempt data were most worth interrogating and using analytics patterns to sharpen the questions asked in subsequent IDIs.

In-Depth Interviews (IDIs) with Customers in High-Failure Postcodes
  • Conducted 36 structured IDIs with customers in the high-failure postcode cluster who had experienced at least one failed delivery.
  • Found that a clear majority of IDI participants reported being home and either not hearing the doorbell at all or hearing it and reaching the door within a normal response time only to find no driver present, with several describing watching delivery notifications arrive on their phone with no preceding doorbell event.
Driver GPS Trace & Attempt Duration Analytics
  • Analysed GPS trace data for every delivery attempt in the affected postcode cluster over a six-month period.
  • Found that average attempt duration in the affected postcodes was significantly shorter than both the network average and comparable density postcodes.
Time Window Adherence Analytics
  • Analysed delivery time window adherence for the affected postcode cluster, comparing the proportion of deliveries arriving within the customer’s selected window against the network average.
  • Found that time window adherence in the affected cluster was below the network average during the late afternoon delivery window.
Doorbell Event & Notification Sequencing Analysis
  • Analysed doorbell sensor event data and customer push notification timing for a sample of failed deliveries in the affected postcodes.
  • Found a subset of failed delivery notifications where no doorbell activation was recorded in the sensor data within the expected attempt window.
Operational Remedy Design & Driver Performance Framework
  • Delivered an operational remedy package including a minimum attempt duration standard with real-time dispatcher monitoring.

The engagement replaced two competing internal explanations with a data-grounded diagnosis that distinguished genuine access difficulty from inadequate attempt quality, and delivered a specific operational remedy targeted at the actual cause.

Impact

  • IDIs confirmed a clear majority of home-but-not-answered complaints reflected genuine presence with inadequate driver attempt rather than access difficulty.
  • GPS trace analytics confirmed average attempt duration in the affected postcodes was significantly shorter than the network average.
  • A bimodal attempt duration pattern in a driver subset identified the source of the short-duration failed delivery concentration.
  • Time window adherence below average in late afternoon correlated with the shift handover period and the affected driver subset.
  • Doorbell event analysis confirmed a subset of failed delivery notifications had no preceding doorbell activation.
  • A minimum attempt duration standard with dispatcher monitoring was introduced.
  • A pre-delivery SMS alert was added at fifteen minutes before the window.
  • Failed delivery rates in the affected postcodes halved within three months of the operational changes.

Client Outcome

Failure Rate Reduction

Failed delivery rates in the high-failure postcode cluster halved within three months of implementing the operational remedies.

Root Cause Diagnosed

Inadequate attempt quality in a driver subset was confirmed as the primary cause, replacing the unverified access difficulty explanation.

Re-Delivery Cost Reduced

The failure rate reduction materially reduced the re-delivery and customer compensation costs that had been pushing postcode unit economics negative.

Attempt Standard Introduced

A minimum attempt duration with real-time dispatcher monitoring gave the operations team a specific, enforceable quality standard.

Customer Communication Improved

A pre-delivery SMS alert gave customers fifteen minutes' notice to be at the door, reducing the legitimate missed-attempt share.

Driver Accountability

Delivery success rate tracking at the driver level provided a performance management basis that the prior failed delivery metric alone had not supported.

Internal Alignment

Operations and customer service teams aligned around a shared, data-grounded diagnosis rather than continuing to advocate competing explanations.

Analytics Capability

GPS trace and doorbell event analytics were retained as an ongoing quality monitoring framework for the affected postcode cluster.

Market Positioning

The retailer was repositioned as a last mile operator that diagnoses delivery failure causes at the attempt-level data layer rather than accepting geography as an explanation.

Case Studies

Learn how our success stories power data-driven growth across industries

Speak with an Analyst

    Download TOC