I was recently boarding a flight on an airline I had zero mileage status with. In addition, my ticket was non-refundable coach class. Of the six or seven different boarding groups, I was second to last. And I had my suitcase with me. Are you beginning to see the problem I was preparing myself for?
Sure enough, it was announced this was a full flight, and they would be happy to check my bag to its final destination. No sooner had this announcement been made than the mood in the long boarding line further declined.
Before I continue, allow me to share I haven’t suffered from lost luggage very often–perhaps three times in more than 25 years of airline travel. In years past, though, most airlines have made checking a bag an added cost. I know that not all airlines engage in this practice, but this particular one I was traveling with does.
Returning back to the line, tense conversations had now developed both at the gate as passengers argued about releasing their bags, as well as at the gate desk. The boarding pace had slowed, as many passengers attempted to make a case about how important it was for them to take their luggage with them and the added time necessary to print and affix baggage tags. In the end, the flight departed late and it’s possible customers missed their connections (while I’m sure their luggage did not).
This story is unfortunately typical of modern air travel. I had plenty of time in line and on the plane to consider how this could be avoided. It got me wondering why airlines hadn’t sought to better address this frequent occurrence. But airlines aren’t alone in this regard; other companies don’t use the information they have readily available to try to solve recurring issues. Using the carry-on luggage situation as an example, let’s take a look at how reviewing available data to build a predictive model can help to avoid frustrating customers and avoid service issues.
Start By Admitting The Problem
While I have no idea how often this occurs, there’s no doubt it occurs often, and it creates frustration for customers and airline staff alike. So why not attempt to mitigate the problem earlier and minimize the customer frustration?
Though we’re using an airline example here, it seems like most companies just don’t take the time to recognize situations that periodically and repeatedly develop that frustrate customers. Like any self-help program, companies must begin with admitting they have a problem. From there, the indicators that signal the potential for a customer problem can be identified.
Examine Individual Parts
The accuracy of predictions will benefit from building a predictive model from the smallest possible measurement point. For our carry-on luggage situation, this is the traveler–or really each traveler, their habits and actions, and some inferences. For this, we can consider the ticket the traveler has purchased and what they do at check-in.
If it’s a roundtrip ticket with the return a day or more in the future, they will probably be taking luggage on their trip. When they checked in for the flight, did they indicate they would be checking one or more bags? If not, there’s a high likelihood they intend to bring their luggage on the plane. There are some assumptions here, but those assumptions will get better over time as results can be compared (more on that later).
Aggregate And Constrain The Data
With an idea of each individual traveler’s luggage situation from the analysis of data at the individual level, it’s now time to get the big picture.
Synthesize how it all fits together–in this case, the amount of carry-on luggage likely to be brought by each traveler from the prior step. Constraining this is the specific aircraft. It has a finite amount of space, so limitations must now also be taken into consideration: how many overhead bins are available for this type of aircraft?
Add Situational Considerations
We have considered each individual traveler’s likelihood of bringing luggage onto the plane, combined that data together, and compared that to our limited available overhead bin space. But that’s not enough.
There will be some additional outside influences to consider, where other random influences are taken into consideration to further improve accuracy. Are all overhead bins available and not in use by aircraft personnel or out-of-service? From past experience, what percentage of customers indicate they will be checking bags during flight check-in but later carry them onboard? Is there any seasonality–day of week, a holiday, etc.–that might also influence the model?
Refine And Improve The Model
The data feeding ongoing predictive exercises change all the time. For continuous improvement, the results of each completed flight must be reviewed–evaluating everything from the individual, aggregate, and circumstantial levels–to verify the accuracy of the prediction as well as to further improve the model and build in adjustments to the calculations.
Proactive Service Replaces Negative Announcements
I realize this was a very simple approach, and even then it sounds like a big data analytics exercise for each and every plane departure. You might argue it’s not feasible for airlines (or your own business), but consider how much time, loyalty, and money is lost in occurrences that might be unavoidable but can at least be predicted and customers can be notified? If I had been warned during check-in, due to the full flight and my late boarding order, that I was likely to not get overhead space and I was offered free luggage handling or another incentive, the airline has done its part to inform me and give me options to avoid frustration.
The point is many businesses experience recurring customer service issues that might be avoided. But if the data is available, what are you doing to identify those specific situations, analyze the indicators and reasons, and working proactively to notify customers of a potential problem to set expectations and put mitigations in place ahead of time? It’s not enough to simply offer more pretzels.
