Achieving precision and accuracy in ETA calculations
Establishing an ETA (estimated time of arrival) implies speculating on a quantity of variables which is, in essence, quite tricky. A good ETA is dependent on many variables - especially in road transportation - and this introduces a high level of uncertainty. Saying that, there is a number of key elements you will need to achieve precision and accuracy when calculating your ETA:
- First and foremost, a good ETA algorithm is an algorithm that takes as many parameters as possible into account: traffic, order information, the driver's past activity, short breaks and night rests, dwell time on site, etc. The more variables get incorporated into the algorithm, the more accurate its estimations are likely to be.
- Your ETA should be computed using Machine Learning methods. Why? Because Machine Learning leverages patterns gathered from data, collected in past deliveries. An algorithm that predicts ETA based on traditional programming cannot rely on historical data and therefore will not be able to deliver such quality in its predictions.
- Make sure your data is clean. Corrupt data such as inaccurate GPS positioning will mislead the ETA calculation and impact your predictions' accuracy.
- ETA prediction works better with real-time tracking. For example, GPS data flowing in real-time at a high frequency is preferable to data delivered in batches. This allows for more accurate and systematic detection of trucks going in and out of loading and delivery sites, and this is, of course, preferable to manual order status changes.
- Last but not least, cooperative carriers. Carriers must adhere to the solution and keep their information up to date (e.g. order - resource assignment) in order for the system to be able to collect information seamlessly.
Following these guidelines should ensure your ETA is as precise as can be.
by Tarik Agayr, lead product manager at Shippeo.