10 questions to ask when selecting a provider for ocean predictive ETAs

By Alex Charvalias

March 24, 2021

In my previous article, I explained the top five things to consider when selecting a vendor for ocean predictive Estimated Time of Arrivals (ETAs). Hopefully, the pre-work is done and now you have your business-specific requirements at hand. Now it is time to reach out to vendors. Most of them might not be that transparent when it comes to claims on their ETA forecasting accuracy. If they are not, you should look elsewhere. If they actually give you some detailed ETA accuracy information, question the following:

1. Is accuracy calculated compared to the Actual Time of Arrival (ATA)?
   
   

   First and foremost, you need to understand what they compare against to calculate ETA accuracy. 

   1. 1. In some cases the comparison is against the Carriers’ ETA, which traditionally is not that great, and also makes comparison between vendors impossible as the Carrier also changes their ETA while InTransit, and you won’t ever know which one they are comparing against. And obviously, if both the Carrier and vendors ETAs are wrong, that doesn’t make the prediction right!

      2. In some cases, it is against the ETA reported by the crew through AIS. Similarly, this ETA also changes while InTransit and is not exactly comparable. And obviously, if both the crew reported ETA and vendors ETAs are wrong, that doesn’t make the prediction right!

      3. The best option is to compare to the Actual Time of Arrival (ATA). In that case, just verify the ATA definition used by vendors is comparable. i.e. entrance at outer port limits or docking, and whether this ATA is calculated by the vendor through geospatial mappings or taken from Carrier EDI messages. Traditionally, the geospatial mappings are more reliable.

   
   
   
2. How frequently is the ETA recalculated?
   
   

   Is it refreshed after each milestone such as a vessel’s port arrival/departure or is it more of a real-time re-calculation?

3. How frequently are the vessel's positions updated?
   
   The more frequent the updates the better, as ETA recalculations can quickly identify a potential delay due to events such as an engine failure or an accident.

   
   
4. Which vessel types and sizes are included in the calculations?
   
   

   In general, the smaller the vessels, the less predictable. If your cargo is primarily in feeder vessels for example, it is possible that these vessel types are not included in the provided accuracy information, or that they represent such a small percentage of the total number of vessels included in the accuracy information, that accuracy might be skewed positively simply because larger vessels are more easy to predict in general.

5. What is the timeframe of voyages included in the calculations?
   
   Is the timeframe sufficient to have statistically valid results? Is the timeframe recent? What if the voyage timeframe is the whole of 2019? Historical patterns of previous years might not be sufficient to predict COVID-19 era voyages, since many patterns were disrupted.

6. Is the port congestion considered?
   
   Is the ETA forecasted to the outer port limits (e.g. anchorage area) or to berth considering when the berth will become available?

   Related: Port congestion explained

   
   
   
7. Is the weather forecast considered in the ETA forecast?
   

   
   If so, is it just extreme weather events such as tropical storms, or more? For example, headwinds of 6Bf versus 8Bf means a potential speed drop of 2 knots, which translates to 3 hours delay per day for a typical container vessel sailing through bad weather at 15 knots. Would this delay be factored in by the forecast model?

   
   
8. Does the provider cover all relevant types of cargo you are interested in?
   
   

   If not, the benefits should be extremely significant to make the business case for having more than one vendor.

   
   
9. Are canal delays considered in the ETA forecast?
   
   

   When vessels pass through a canal such as the Suez or Panama Canals, they have to wait in line, and also traverse in the canal slowly. The wait time especially is dependent on the canal traffic which is affected by seasonalities and other factors.

10. How is the platform uptime promise calculated?
    
    

    You won’t get any ETA if the system is regularly down. Let’s compare two similar promises of 99% uptime, one vendor calculating uptime per month and the other per year.

    • 99% uptime per year = 3.65 days total downtime per year: Is it acceptable not to get updated ETAs for almost four consecutive days? I believe not.

    • 99% uptime per month = 7.2 hours total downtime per month: Is it acceptable not to get updated ETAs for almost seven consecutive hours, every month? I believe yes.

Alex Charvalias