Manaz Afzal
Logistics & Transport · Pickup ETA messages

A pickup time the customer can plan the day around

An ETA step that runs when the driver accepts the job, tested on 338,524 pickups over 56 later days.

46 minaverage error; window midpoint 77 min
44%within 30 minutes; window midpoint 35%
79%inside the 80% range, median width 120 min

Executive summary

An estimate made when the driver accepts the job is off by 46 minutes on average, against 77 for the middle of the customer’s window, and lands within 30 minutes 44% of the time.

46 minaverage error, against 77 min for the window midpoint
44%of pickups within 30 minutes of the time given, against 35%
220 hof office time saved a month at 5,000 pickups a day

Situation

Customers are given a half-day window and ring the office to ask when the driver will arrive; dispatch answers from memory.

Approach

A gradient-boosted ETA on what is known at acceptance, trained on 755,195 pickups and tested on 338,524 over 56 later days, with a calibrated 80% range for dispatch.

Result

Average error falls from 77 to 46 minutes. The 80% range holds 79% of the time, so dispatch is alerted when a pickup slips outside it.

The problem

A parcel or courier network that gives the customer a pickup window of a whole day gets failed pickups and calls asking where the driver is. Carriers that send a narrower window, updated as the run progresses, report more first-attempt success and fewer of those calls. A window helps only when the driver arrives inside it.

I measured how accurately a pickup time estimated when the driver accepts the job lands against the actual arrival, and how narrow a window can be while still holding most of the time.

Team
Dispatch, customer service, operations manager
Systems
Descartes, Bringg, Onfleet or an in-house dispatch system; Salesforce or Zendesk for customer contacts; SMS through MessageMedia or Twilio
Measure
Pickups within 30 minutes of the time given; where-is-my-driver contacts per hundred jobs

What I built

An ETA step that runs when a driver accepts a pickup. It reads the customer’s requested window, the time of day, the driver’s distance, how many pickups the driver already has open, and how long pickups usually take for that driver and in that area. It returns a single time for the customer message and an 80% range for dispatch.

AcceptedDriver takesthe pickupEstimateTime and80% rangeMessageSMS to the customerwith the timeWatchDispatch alerted if therange is slipping
The estimate replaces the half-day window in the first message. Dispatch keeps the range and is alerted when a pickup runs outside it.

Results

Over 56 test days and 338,524 pickups, the time from the model is off by 46 min on average. The middle of the customer’s requested window is off by 77 min, and each driver’s usual time by 82 min. 44% of pickups happen within 30 minutes of the model’s time, against 35% for the window midpoint.

Courier’s usual time82 minMiddle of the requested window77 minETA model46 min
Courier’s usual time31%Middle of the requested window35%ETA model44%
0%5%10%15%-240-180-120-600+60+120+180+240ETA modelCourier’s usual timeTime given minus actual pickup, in minutes
The model’s errors cluster near zero; the driver’s usual time is often hours too early because it ignores when the customer asked to be collected.

The 80% range contains the actual pickup 78.7% of the time, with a median width of 120 minutes. That is narrow enough for dispatch to spot a pickup going wrong, and wide enough that I would not send it to customers. The customer message carries a single time, updated if the range slips.

What drives the estimate

When the window opens0.264Hour accepted0.040When the window closes0.036Distance to the pickup0.034Typical time in that area0.023Courier’s typical time0.021Pickups already open0.019Region0.007
When the customer’s window opens carries most of the signal. Time of day, distance and the driver’s open load refine it.
Courier’s usual timeETA modelNo other open pickups51 min1 to 244 min3 to 543 min6 or more46 min
The gain over the driver’s usual time is largest when a driver already has six or more pickups open, which is when customers most often chase.
Open pickups at acceptanceTest pickupsETA modelCourier’s usual time
No other open pickups34,96751 min78 min
1 to 250,65744 min68 min
3 to 573,53643 min68 min
6 or more179,36446 min93 min
Courier’s usual timeETA model7:0043 min8:0048 min9:0055 min10:0042 min11:0058 min12:0041 min13:0046 min14:0034 min15:0042 min16:0026 min
Morning acceptances, the busiest period, are where the usual-time rule is weakest.
Under 1 hour21%1 to 2 hours37%2 to 3 hours18%3 to 5 hours13%5 hours and over11%
Pickups usually happen within three hours of acceptance; the median is 107 minutes.

Data and method

Data
LaDe last-mile dataset (Cainiao; Wu et al., KDD 2023; Apache-2.0): 1,146,781 pickup records from Yantai, China, with acceptance time, the customer’s requested window, courier and area identifiers and the actual pickup time. 1,093,719 kept after removing records with missing times or pickups more than 12 hours after acceptance. No customer details are used
Split
By day. The first 70% of 184 days train (755,195 pickups), the last 56 days test (338,524)
Model
Gradient-boosted trees (scikit-learn HistGradientBoosting) on log minutes to pickup, fitted on 400,000 training pickups
Range
Split-conformal 80% range from the last 15% of training days (106,814 pickups), set by predicted-time band
Baselines
Middle of the requested window; each courier’s median time in training
Inputs
Only what is known at acceptance: window, hour, distance, open pickups, courier and area medians from training, area type and region

Limitations

How I would take it live

Financial impact

About 220 hours of where-is-my-driver contacts avoided a month

The saving is office time spent answering customers who chase. The share of contacts avoided by a sharper time is an assumption until measured.

InputValueSource
Pickups a month110,000 about 5,000 a day across a national network
Share of pickups where the customer chases10%
Chasing contacts avoided with a sharper time30%
Minutes per contact4 call or message, look-up and reply
Average error, window midpoint77 minMeasured
Average error, ETA model46 minMeasured
Loaded cost per hourA$45 customer service or dispatch

Calculation

  1. Chasing contacts a month today11,000
  2. Contacts avoided3,300
  3. Office time saved a month220 hours
  4. Annual valueA$118,800

Annual value of office time savedA$118,800

Fewer missed pickups and second attempts are a further gain and are not counted here.

Implementation and rollout

Six weeks from job export to ETA messages

  1. Week 1

    Export and baseline

    Three months of jobs with acceptance, window and completion times from the dispatch system; count today’s chasing contacts.

    Operations manager, Manaz

  2. Weeks 2 to 3

    Train and set the range

    Train on the business’s drivers and suburbs, test on the latest weeks, agree the message wording.

    Manaz, customer service lead

  3. Weeks 4 to 5

    Silent run

    Estimates logged against actual pickups without being sent; dispatch reviews the misses.

    Dispatch

  4. Week 6 onward

    Live messages

    The time goes in the customer SMS; dispatch alerted when a pickup leaves its range. Monthly retraining.

    Operations manager, Manaz on support

Tools

Track-POD, Onfleet, Circuit or Shippit as the dispatch record; ServiceM8 for field teams, Python (scikit-learn) ETA service behind a small API, MessageMedia or Twilio for the customer SMS, Make or n8n to trigger the estimate on acceptance

Team

  • Operations manager owns the message wording and signs off go-live
  • Dispatch acts on range alerts and flags bad estimates
  • Manaz data work, model, integration and monitoring

The timeline is a plan for a typical engagement; it depends on data access and how many systems are involved.

Risks and mitigations

What could go wrong in production, and the control for each

RiskMitigation
A time is given and missedThe message carries one time and is updated once if the range slips; dispatch is alerted first.
New drivers with no historyArea and fleet medians are used until the driver has two weeks of pickups.
Customer phone numbersMessages go through the business’s existing SMS provider; no customer details leave it.
Seasonal peaksRetraining before peak periods and a weekly accuracy check.

Key takeaways

Key takeaways

  • An ETA at acceptance cuts the average error from 77 to 46 minutes.
  • When the customer’s window opens carries most of the signal; driver load and distance refine it.
  • The 80% range is for dispatch alerts; customers get one time.
  • The model is retrained on the business’s own drivers and suburbs before any message is sent.