Chris Conway
Chief Architect, Quantiv
I often travel by train as part of my work. But rather than having a daily short commute, my journeys tend to be less frequent and longer.
Ahead of each trip, I receive a reminder email with ticket confirmation, as well as advice about timings, luggage, connections, and Covid-19 precautions.
Since the pandemic began, we’ve all come to accept plans can often change. I now regularly find I’ve bought tickets for one date, only to have to change them later for travel on another day. But although the ticket changes work correctly, I still receive reminder emails for the original dates.
Receiving the right information… at the right time
Leaving aside that an email entitled “you’re travelling with us tomorrow” should probably arrive the day before the journey, it should also only be sent to those who are actually travelling on that date.
I presume this problem is just the result of an omission of a connection between systems. And it’s perhaps a forgivable mistake, because changing a ticket isn’t seen as being as important as buying a new one.
However, the fact the discrepancy can exist at all reminds me data is often not shared in the way it should be. Many IT solutions now rely on messages being sent between applications. But if a message isn’t sent (or received), an inconsistency arises.
Rather than duplicate data in this way, a better solution is to share information.
Sending a message requires the sender to know about all possible recipients, and it's all too easy for one to be left out. But sharing only needs the sender to remember to publish the information, and that can be done independently of how it will be used. Consumers can then use the information in any way they see fit.
Big data and AI
In an age of big data and artificial intelligence (AI), it would be easy to assume this process could happen almost automatically. But rather than these technologies being a solution to sharing information, they can introduce more problems.
As their names suggest, big data and AI often relate to large data sets used to provide intelligence about unusual patterns within an organisation’s operation. In effect, they help identify ‘unknown unknowns’ within the data, i.e. future outcomes that haven't been predicted.
Identifying these patterns doesn’t depend on data being entirely complete, up to date, or even fully consistent. But it does require data to be comprehensive enough to allow patterns and dependencies to be seen.
In short, publication in this context can be quite a loose and varied process.
Operational management and data
For other uses, those characteristics are the opposite of what’s required. Operational management relates to well-understood metrics - the ‘known unknowns’ - so it’s important the information used is based on data that is complete, up to date and consistent. Conversely, the information only needs to include specific details, not all possible data.
Here, publication is a concise and unambiguous process.
When simply sharing isn’t enough
Using one form of data interchange where another is more appropriate leads to processing becoming inconsistent and unreliable. For example, receiving a reminder email for a train journey on which you’re no longer travelling.
Information sharing can offset many of the problems caused by data duplication or omission.
But on its own, sharing isn’t enough. Information must also be shared in the right way, using a coherent information format and service. Doing this will prevent problems that can happen when incomplete, delayed or inconsistent data is shared.
Analytical data and operational information both have their places, but they shouldn’t be confused.
An issue with a travel email might just lead to some anxious moments. But if the same problems were to apply to accounting for revenue or reporting on performance (especially against compliance targets), the consequences could be much more serious.