Convey’s goal is to help our customers improve the delivery experience, including surfacing likely delays as early as possible. Every day, 5-10% of all parcels expected to be delivered will fail to be delivered on time, typically with no warning from the shipper or carrier. This leaves your customers empty handed, wondering where their shipment is, and significantly more likely to call in asking “Where is my order?” (WISMO). To combat this problem, we have created Predicted Delays, an exception type that detects shipments at risk of failing to meet delivery expectations.

Use the information on this page to understand how predictions are made, the kinds of predictions that Convey surfaces and how to make use of this information to get ahead of a bad customer delivery experience. 

What are Predicted Delays?
How It Works
How To Locate Predicted Delays in the Platform
Prediction Accuracy
Timing
Suggested Use
FAQs and Additional Information

Have questions that are not answered here? Let us know!

What are Predicted Delays?

Predicted Delays is an exception type that detects shipments at risk of failing to meet delivery expectations.There are two types of Predicted Delays powered by Convey, based off two different customer expectation models:

• Predicted to Miss EDD (PMEDD): EDDs are dynamic delivery estimates provided by the carrier. These EDDs are typically what is communicated to end customers on their tracking pages, setting the customer expectation. PMEDD detects shipments that are likely to miss this EDD, but have not received an updated EDD from the carrier.
  
  
• Predicted to Miss Promise Date (PMPD): Promise Dates are static delivery estimates provided to the consumer by the shipper, often at the time of order. These Promise Dates are not considered by the carrier, so when a shipment is expected to miss the Promise Date, the end customer is not alerted to the delay. PMPD detects shipments that are likely to miss their Promise Date before that Promise Date passes.

Predicted Delays can help you mitigate these experience failures by notifying your customers of the issue before they are left empty handed. Every predicted delay represents a potential customer churn and WISMO call.

How It Works

Convey continually monitors in-flight shipments and  identifies several data points about them, including the time passed since the last event, status of the last event, distance to destination, carrier, service level, time of day, time remaining until the delivery expectation, package weight and more. 

Convey then compares these data points to patterns we expect from on time deliveries through machine learning models. When a shipment is expected to miss its delivery date with a high level of certainty, Convey tags it as a predicted delay and assigns a percentage-based risk level. These percentages translate to the probability that a shipment will end up missing its EDD or Promise Date.

How To Locate Predicted Delays in the Platform

Where and how you access shipments that have a Predicted Delay will depend on the Convey product(s) that you use.

Discover

A dashboard depicting a birds-eye view of in flight Predicted to Miss EDD exceptions is available in Exception Insights:

This dashboard summarizes the number of currently impacted shipments, a timeline of impacted shipments over the past two weeks, and breakdowns of flagged shipments by carrier, risk score, and location.  These breakdowns can be used to triage and filter down to a subset of shipments that you can then act on. Click on the chart elements to see the underlying shipments.  

Discover also offers in depth reports on Predicted Delays upon request.  The predicted delay dashboards in Advanced Reporting provide a detailed view of prediction accuracy and outcomes and a lane-by-lane breakdown of the operational impact of these delays:

Model Results

Operational Impact

Extend

If you are utilizing the Retrieve Shipment Events API, predicted delays will appear as event objects with ‘prediction’ as the event_type. These objects will contain the score, prediction type, and whether the predicted delay is being created, updated, or resolved. Here is an example:

“event_details": {
          "score": "96.1",
          "status": "Updated",
          "shipment_insight_type": "predicted_miss_edd", }

Recover

Predicted Delays are available as a filter on the ‘Shipments’ page in Convey under the Exception Types filter. Users may choose which delivery expectation the predictions were scored against, and the risk score to display.

Predicted Delays also appear as a badge on the Shipments table results, along with the risk score. To enable the Prediction Column on the shipments table, enable the column in Admin Settings.

Finally, the predicted delay also appears on the shipment detail page timeline as a Convey-identified exception event: 

Prediction Accuracy

Both PMEDD and PMPD are reported with a risk score - higher risk scores are more likely to result in delays. Convey reports on the accuracy of predicted delays using precision, or how often predicted delays resulted in real delays. 

Predicted to Miss EDD

Shipments are flagged as PMEDD with 96% precision--this means that 96% of all shipments flagged as PMEDD end up missing their estimated delivery date.

We report PMEDD by a probabilistic risk level--at all risk levels shipments are more than 94% likely to miss their EDD, but higher risk levels are even more precise. When tagged with the highest risk level, shipments miss their EDDs 98% of the time.

Predicted to Miss Promise Date

Shipments are flagged as PMPD with 91% precision--this precision is slightly lower than for PMEDD because it is provided further in advance.

At all risk levels shipments are more than 86% likely to miss their Promise Date, but higher risk levels are even more precise. When tagged with the highest risk level, shipments miss their Promise Dates 97% of the time.

Further results, impact, and accuracy analysis is available through detailed reports in Convey’s Discover product. 

Timing

PMEDD exceptions are only reported on the day of the EDD. The predicted delays occur throughout the day, but most occur around midday. On average, PMEDDs are detected 26 hours before the carrier revises the EDD to a later date and 49 hours before delivery. 

PMPD exceptions can be reported as soon as the shipment is trackable, not just on the day of the promise date. PMPDs are detected on average 36 hours before the end of the Promise Date, and 65 hours before delivery. 

Suggested Use Cases

Proactive Alerting

Convey suggests using predicted delays to alert customers with flagged PMEDD/PMPD shipments that a potential delay has been detected in their delivery. Convey research shows that 87.5% of customers expect retailers to push alerts to them by email or text if a delivery will miss its EDD. 

Convey suggests alerting all customers with predicted delays above a desired threshold for risk scores. A good place to start is by alerting all customers with the highest risk scores (95%+) at first, and then expand into lower risk scores as desired.  Alerting based on lower thresholds will allow you to capture more potential delays and to capture them earlier, so this should be weighed against the slight increase in false positive risk.  

To alert customers, use the exception filter to find all affected shipments above the appropriate risk score and apply any other desired filters to the search results (like customer segment, origin, destination etc.) From there, use Convey Bulk Alerts to communicate to all affected customers. 

Network Problem Detection 

Another common use case for Predicted Delays is to detect and diagnose larger network problems in real time. In the Exception Insights page (part of Convey’s Discover product), we report predicted totals by current location, carrier/service level, and risk level. Often delays are concentrated in a few locations, so these Exception Insights provide a leading indicator of a delayed carrier facility, lost truck, weather event or other exogenous event causing delays:

This high concentration in Glenn’s Ferry, ID showed a large weather related issue before any weather alerts, exceptions or delays were reported--and before missing any customer expectations.

In this instance, the shipper could alert all 200 of the affected customers to let them know their shipments were being impacted by the winter storm and to keep an eye on the tracking page for an updated EDD.  The shipper could also buffer any customer promise dates to the impacted areas until operations return to normal, or contact the carrier to request a trace for these shipments.

FAQs and Additional Information

Which modes and carriers are supported for this feature?
Predicted Delays are only available for the following parcel carriers at this time: UPS, USPS, FedEx, Canada Post, DHL, DHL eCommerce, and Purolator International.

Will other modes and carriers be supported in the future?
Convey plans to expand this to other carriers and delivery modes as soon as possible, and to the extent that the available data supports. 

Why don’t I see Predicted to Miss Promise Date in the app? 
Predicted to Miss Promise date requires a Promise Date to be sent to Convey via an order feed or Order API integration.

When does the PMEDD filter apply?  Does the filter capture previous predicted delays?
PMEDD only applies to shipments that are expected to be delivered today, meaning shipments that have already missed their EDD and shipments with a future EDD are not eligible to be flagged as active PMEDD.

What about my shipments that have already received revised EDDs, or exceptions?
PMEDD only applies to shipments that don’t already have exceptions, meaning PMEDD helps find distressed shipments that otherwise appear to be on the happy path.

Will my customer see this prediction on their shipment tracking page?
No, predictions are not currently exposed to your customers. They are only visible in the operator application.

Can I automate an alert to my customer, based on a prediction? 
At this time, manual (bulk or one-off) alerts are the only ways to alert customers. We will offer the option to automate these alerts in the future.