The Travel and Expense Data mess: S = k log Ω (E)

Keep the above in mind. We’ll come back to it.

I was showing my passport in Charles de Gaulle Airport recently when I was struck by a thought: you know what? Waiting in lines is boring. We could improve efficiency and the comfort of travelers by letting them wander in airports – in the shops and bars - and making security agents go to the traveler and proactively request them to show their passport. If I were to suggest this in the real world, it’s pretty likely that you would stare at me as if I were a complete bozo. After all, this is going to require 10 times more people and we are sure to miss some of the passengers who might (for example) hide in toilets. There is no way for this method to be exhaustive. Nobody with common sense would propose such a process.

Dumb statistics

One of the most frequent sources of bad decisions comes from using wrong data, or interpreting correct information with bias, like making false logical connections: 100% of theists believe in God, 100% of theists are human, therefore 100% of humans believe in God! Statistics are a science, but of course we don’t live in a perfect world and seldom do we have complete information. When the information is too fragmented or when what is built upon that information is not meaningful, we can come to absurd conclusions.

On the 10 top business trends of enterprise architecture as analyzed by Forrester, 5 are directly related to data, business intelligence and so forth. With Web 2.0, user generated data, social networks and computing power available thanks to Moore’s law, the amount of available information to us is growing exponentially. In my last blog post, “don’t throw the baby out with the bathwater” I had the opportunity to explain that when it comes to travel mobile computing, a lot of it is about the back-end where all data reside. 

The expectation of quality data, however, is not new and often falls short of satisfaction. One thing which needs to be asked is what are the right performance indicators? It’s a fundamental question, but not the one I will address today. Before defining what these right performance indicators are, you need to have a good and complete set of data.

Hands Up: I want it all!

In a remarkable buyers roundtable event organised recently by The Beat, a full third of the discussion centered around dealing with “getting good data”.

The Travel and Expense universe is challenging, with the first hurdle we need to jump being that data often comes from numerous sources. Yet without the completeness requirement being fulfilled, we only get meaningless results. And for a lot of players, that’s all they offer. Examples?

We can start with full transparency by examining what KDS was like once upon a time, and considering reports which typically come form an Online Booking Tool: “here are your 10 top city pairs, customer. But wait: what I mean is your top ‘city pairs’ for which travelers booked online; unfortunately I know nothing about the rest, i.e. offline bookings. I’ve to assume the patterns are identical.”  This inference does not pass the cut, however: offline is mostly used for complex bookings, so the ratio will be different. In the end, what business meaning can you give to reports based only on the subset of your trips which were booked online?

Next, let’s look at Travel Management Companies. “Here is your spend, customer. But wait. Low Cost Carriers flights won’t be included. Your employees are not calling the Travel Agency for those bookings. You know, many of them are not using our services to reserve hotel rooms either. Total cost of trip? Sure, that’s what you want in an ideal world but where the hell can I know how much you spent on meals or taxis?”

And what about Expense Management Systems? “Of course we have a more holistic view of employee paid expenditures. But most of the time we know little about what is centrally paid. Consequently we are missing flights, 50% of your costs. If we do, it’s with little detail. Don’t expect meaningful travel policy statistics.”

Of course, CFOs, Travel managers, Procurement Managers and so on could not care less if a trip has been booked online or offline, directly on the website, through a direct link or through a GDS. It does not make much of a difference to them either if a hotel room sometimes go through a Travel Agency and sometimes not. And the fact that buying lunch or taking a domestic train happens during trip time whilst booking a flight is performed beforehand does not lead these expenditures to fall in a different category.
 

Filling a barrel with a spoon

Sisyphus rolling a rock up the mountain, then letting it roll down … for ever

The usual remedy is to create a central database, built upon a unified data model, fed with as many sources as necessary. Whilst this is a classical approach for implementing business intelligence or performance management, it does not work that well when applied to the travel dimension.

There are two reasons for this:

• When dealing with more global business information systems, a well defined and limited number of production systems provide the data. If a CEO wants to have indicators on sales and revenue performance, the sources are likely to be the Financial, Supply Chain and Sales Management software, sometimes even gathered into a single ERP. Whatever the precise list is, a small number of systems host 100% of the required information. There is a path for Management Information Systems to be accurate and exhaustive. As described above, this is hardly the case when it comes to travel.

Travel and Expense, while undoubtedly very important, is not the core business of enterprises. Millions are invested in general purpose MIS. The same level of investment for Travel and Expense only is harder to justify even though the issue is not necessarily a lot simpler. As a result, the budget you need to get a fully working data management environment for T&E seldom exists.

…And this is where I link in the thesis story which I used to start this blog post. Executives or Travel Managers are not willing to become “University experts” in the art of gathering travel data. They do not have the manpower to spend months on searching where to take the information from, then building or asking their partners to supply dozens of interfaces. To push the comparison further, they are closer to teachers, who have a number of texts to refer to in order to give courses, which for any given level are supposed to contain the same elements all over a country. More often than not, our industry acts as if we are asking our clients to become Doctors in Travel and Expense Data.

This is all my - maybe convoluted - way of preaching for Industrialization.
 

Columbus Egg

It’s shocking that we continue to ignore a simple principle: the further you go from the origin of the information, the more the process is prone to errors. This is an equivalent to the second law of thermodynamics as applied to information systems: disorder/entropy only grows with time. 

If I extract received orders from the sales management systems, I’m sure to have all of them, and the data is 100% accurate. That’s why - and forgive me for stating the obvious - there are passport lines in airports. You have to go through the line, control can’t miss you. 10 meters down from the checking point, entropy grows, errors would start to flourish. Reliability would drop sharply.

In plain English: if you want to have a reliable Management Information System, data must be taken right from the Production System(s). Not that you don’t create a distinct database, of course; in fact in most cases you do: you want to avoid slowing down production with MIS requests which are one or several orders of magnitude heavier than production ones. But gathering data by hand from here and there and hoping to reconstitute a complete set is a losing proposition. It’s never going to be reliable.

The Alma Mater is the production system(s). But can we identify it or them in this field? Historically, the answer has been No. Today, however, we can say a resounding “Yes!” I’m talking, of course, about Travel and Expense Management Solutions. If well implemented they are the control gate - and by “well implemented,” I mean they should manage:

• Travel bookings going through them
• Expenses paid by employees
• Travel items paid centrally through BTA (Company Credit cards)

There is no escape and no possible hole: either you have booked your trip using the system, or you had to get reimbursed for your expenses, or you have called the Travel agency and it has been paid by the Company directly. There is no way none of these three actions happened and there is no way it did not create corresponding data inside the T&E package ie the production system.

Can we succeed without having a real production system, i.e. an end-to-end Travel and Expense Management Solution? As Michel Audiard, a famous French movie dialogist, once said: “flying fishes exist but they don’t constitute the majority of the specie.”

Of course you still have to extract the right indicators and it’s nothing but easy. But once in place, once you get rid of the “shit in shit out” problem, at least one issue is gone: where to get the data from.