Sendy Logistics Challenge
$7,000 USD
Predict the estimated time of arrival (ETA) for motorbike deliveries in Nairobi
1555 data scientists enrolled, 431 on the leaderboard
LogisticsPredictionStructuredLocation
Kenya
23 August 2019—26 November 2019

The dataset provided by Sendy includes order details and rider metrics based on orders made on the Sendy platform. The challenge is to predict the estimated time of arrival for orders- from pick-up to drop-off.

Sendy provides an API as well as a web and mobile application platform to link customers who have delivery needs with vetted transporters. The customers select their vehicle of choice, get their price quote upfront and pay using various payment options. The system optimises the route, looks for the closest available riders and dispatches the orders in the most efficient way.

The training dataset provided here is a subset of over 20,000 orders and only includes direct orders (i.e. Sendy “express” orders) with bikes in Nairobi. All data in this subset have been fully anonymized while preserving the distribution. Furthermore, we have added data about the weather, that corresponds to the time of the order.

You are also able to use any data from the ArcGIS Africa GeoPortal (details below).

You may use only the data provided here in your solution. However, if there are other publicly available datasets that you believe can help improve predictions, please notify us at zindi@zindi.africa. We will assess whether the data should be allowed in this competition. We will post links to these new datasets below in this section and also announce them on the discussion forum. Until we respond, please assume that you are not allowed to use the dataset in your solution.

The files for download are:

  • Train.csv - is the dataset that you will use to train your model.
  • Test.csv - is the dataset on which you will apply your model to.
  • Riders.csv - contains unique rider Ids, number of orders, age, rating and number of ratings
  • VariableDefinitions.csv - Definitions of variables in the Train, Test and Riders files

Additional datasets:

Variables

Order details

  • Order No – Unique number identifying the order
  • User Id – Unique number identifying the customer on a platform
  • Vehicle Type – For this competition limited to bikes, however in practice, Sendy service extends to trucks and vans
  • Platform Type – Platform used to place the order, there are 4 types
  • Personal or Business – Customer type

Placement times

  • Placement - Day of Month i.e 1-31
  • Placement - Weekday (Monday = 1)
  • Placement - Time - Time of day the order was placed

Confirmation times

  • Confirmation - Day of Month i.e 1-31
  • Confirmation - Weekday (Monday = 1)
  • Confirmation - Time - time of day the order was confirmed by a rider

Arrival at Pickup times

  • Arrival at Pickup - Day of Month i.e 1-31
  • Arrival at Pickup - Weekday (Monday = 1)
  • Arrival at Pickup - Time - Time of day the rider arrived at the location to pick up the order - as marked by the rider through the Sendy application

Pickup times

  • Pickup - Day of Month i.e 1-31
  • Pickup - Weekday (Monday = 1)
  • Pickup - Time - Time of day the rider picked up the order - as marked by the rider through the Sendy application

Arrival at Destination times (column missing in Test set)

  • Arrival at Delivery - Day of Month i.e 1-31
  • Arrival at Delivery - Weekday (Monday = 1)
  • Arrival at Delivery - Time - Time of day the rider arrived at the destination to deliver the order - as marked by the rider through the Sendy application

  • Distance covered (KM) - The distance from Pickup to Destination
  • Temperature -Temperature at the time of order placement in Degrees Celsius (measured every three hours)
  • Precipitation in Millimeters - Precipitation at the time of order placement (measured every three hours)
  • Pickup Latitude and Longitude - Latitude and longitude of pick up location
  • Destination Latitude and Longitude - Latitude and longitude of delivery location
  • Rider ID – ID of the Rider who accepted the order
  • Time from Pickup to Arrival - Time in seconds between ‘Pickup’ and ‘Arrival at Destination’ - calculated from the columns for the purpose of facilitating the task

Rider metrics

  • Rider ID – Unique number identifying the rider (same as in order details)
  • No of Orders – Number of Orders the rider has delivered
  • Age – Number of days since the rider delivered the first order
  • Average Rating – Average rating of the rider
  • No of Ratings - Number of ratings the rider has received. Rating an order is optional for the customer.

Supplementary data: ArcGIS & Africa GeoPortal

Delivery logistics is geographic by nature, so location data and spatial analytics should play a key part in your response to this challenge. Esri and Zindi are working together to give you access to world leading spatial analytics software, ArcGIS and additional contextual datasets such as base-mapping, satellite imagery, traffic data, population data and more. Esri’s ArcGIS technology can be accessed via desktop trials, documented APIs, notebooks and online tools, the best starting point is a free account on the Africa GeoPortal (www.africageoportal.com).

To get an understanding of the sample data, you may wish to plot the pick up points and Sendy distribution centers. Also what about roads, speed limits, traffic congestion? All these should play a factor in your models. There are a number of ArcGIS tools such as route optimisation, nearest point analysis that might help you.

Here is some useful information to get you started on this challenge:

  • To access the ArcGIS Technology, free accounts can be made here www.africageoportal.com. Further instructions can be found here.
  • CSVs with location tags in them such as addresses or coordinates can be quickly displayed via drag & drop – how to is here.
  • To check out one key data set, traffic data, see more info here, be sure to create an Africa GeoPortal account first, then you can access this data.
  • Other useful datasets such as imagery, population data covering Tanzania can also be accessed for free via Africa GeoPortal – see here.
  • For data scientists, integration between ArcGIS and Jupyter Notebooks may be of great value. More info here, tools here and set up here.

If users want ArcGIS desktop products, you can download a trial here. Trials are 21days but you can request an extension if needed. Or for full unlimited access to Esri’s technology for use from home, check out this package. Any questions please send an email to zindi@zindi.africa or africageoportal@esri.com.