datascience

Practical Data Science for the Travel Industry

Are you an Excel jockey? Do several reports sit on your desk every day? It’s time to update your act and adopt the latest technology.

Years ago, I used to think that “Data Science” was a fancy new term for my old, trusted Statistics. Boy, was I wrong! Actually, stats are a component of the complex body called DS, which combines also programming and analysis methods that were not previously available in classic statistics, plus practical expertise in a specific domain. Instead of lengthy -and debatable- definitions, I’ll just affirm that these are two related but distinct fields, and will try to show you what’s the difference between both sciences and what’s each one better for from a practical standpoint… which is why you’re reading this text, I assume.

When I owned a DMC back in the nineties, I noticed that we were inefficiently using our buses. I collected information about the flights arrival/departure times (taking into account delays, queue check-in times, etc), then looked for the average number of passengers per zone we had on each flight, average delays, average customs clearing times, then devised some sort of schedule in order to transport as many passengers per vehicle/flight as possible, without extending waiting times. Hello, parking and fuel savings!

Years later, before 9-11, I was starting an outgoing operator. In that market (rather small), everybody and their grannie were vacationing to the same spot, even to the same 2 or 3 hotels. Every tour operator, every travel agent, were selling exactly the same packages. No kidding! How was I supposed to introduce a differential product and compete against well-established companies? During a couple of Amadeus events I conducted a survey among local retailers; then I went to this oh-so popular destination and surveyed my target holidaymakers. Turned out they loved this place because (in this particular order):

  1.      Hotels were all-inclusive
  2.      It was “only” a 1.5 hours transfer from the airport
  3.      Beach was nice
  4.      Package was fairly cheap

But everyone complained that, since there was just one flight per week, availability was rather limited. So I found a similar destination with much better all-inclusive hotels at same (or lower) rates, even better beaches and -more to the point- only 15-20 minutes’ drive from its airport. Then contracted a local airline with small planes for a charter operation, and ran ads in the national papers with a special package offer for this new, fashionable destination for hype connoisseurs. Result: it took some time, but from the third charter or so, we were fully booked.

Am I super-smart? Alas, I’m just a halfwit drop-out from a computer science degree. The above mentioned are possibly my two only achievements in a decades-span career, and credits go to data. However, I might have what’s called an analytical mind, as I experience a warm feeling towards statistics: I always enjoyed finding patterns, trends, relationships among numbers… and that’s where this branch of mathematics shines. Stats can tell you a story of what was and what’s going on right now, but falls a bit short on predictions. Stats are brilliant if you need to know what a large population did at some point in the past, but would have a hard time telling you what a small group (or even a specific individual!) will do in the future. One of the reasons for that is because Stats do not like much unstructured data, whereas Data Science loves to eat anything from any source and find order, some meaning among a seemingly chaotical poodle of information. That’s why Data Science also excels at forecasting. In short: you either use one or the other depending on what you want to achieve.

Up to this point, you might still ask yourself “what do I need this for?”. Well, you’ve heard or read about all these buzzwords like “big data” (see my article about it >>) and “revenue management”, which are related in more than a way. For instance, in tourism is quite uncommon to process and analyse big data sets without data science; similarly, revenue management techniques are being dramatically improved by using data science methods (even artificial intelligence procedures). Simply put, what you’ll get from data science is competitive advantage… which is exactly what the big players are doing! Check the job offerings from TUI, Expedia, Airbnb, etc. What they are looking for is data wizards!

Data science is a team sport

Does the above last paragraph mean that you -a mid/small-sized operator or independent hotel, maybe- cannot afford to apply data science to your particular environment? Absolutely not! Although it might sound as an IT-specific field, that’s not the case. Thing is, so many skills are required to perform data science projects that a team has to be assembled. Nobody (except a few geniuses, perhaps) can do this alone. The simplest, most basic data science team should include:

  • 1 developer
  • 1 statistician or mathematician (possibly versed on data mining, wrangling, etc)
  • 1 analyst
  • 1 expert in the domain (that would be you)

You will be the captain of this team, because:

  1.      You know exactly which business questions is the project designed for
  2.      You’ll know how to interpret the project’s results
  3.      You’ll know how to implement the solutions proposed based on results
  4.      You pay for the toy

The question or questions are relevant to the goal you intend to achieve with the project, because at the end of the day what you need is to take action based on whatever results the project brings. That’s why the second reason is paramount: no machine can (yet) tell you exactly what’s the course of action, so you’ll need to understand how/if/when/ the goal will be actionable, given the project results. I suppose the third reason is pretty obvious: nobody better than you will know how to apply whatever the results are, in the current context. So, delving a bit into the technical side, you’ll be providing the nerds in the team a framework to build their “model”.

Data collection

Even before planning to convert data into profits, you’ll probably need a data warehouse. You surely heard that these days, data is gold. Experts would be right to kill me for this blasphemy, but I’ll extremely simplify one of data science’s basic rules (especially if your goal is to get predictions): the more data you can process, the better (and the more accurate your predictions). Hence, in order to profit from your data, you’d better start collecting it properly, on a single depot preferably. I guess you have an ERP or accounting software, a booking system, a PMS, a fleet management system, a CRM, Google analytics from your e-commerce site and so on, each generating large amounts of information from all types of transactions. Now, while all these “data marts” manage really well their own figures, they usually can’t exchange information to each other. The idea then is to extract data from every source and convey it to a single “data warehouse”. In this central repository, your tech guys will “Extract, transform, load” your data, so it will be ready to be processed by Business Intelligence tools or to conduct data science experiments. This used to be expensive: not anymore. With a modest investment, you have now a Ferrari that runs on data fuel… leave those spreadsheets to old-school managers who prefer to drive horse-pulling carts.

Examples

With a data warehouse in place, you’ll be better positioned to ask those business questions that will convert information into profits. A basic example: you might find no correlation between your social media followers and your website users. If you look into your Google Analytics insights, your Facebook page and Instagram stats, and try to run some sort of comparison, you’ll probably end up understanding even less about your audience. Conversely, having all your audience data in a central location and a proper analysis tool, will help you not only to establish the right demographics for each channel, but to find out their preferences as well. Hence, your product management and marketing campaigns will be much more efficient.

Now, think about your own business: can you imagine further useful applications? I’m sure you can! In fact, there are dozens of areas in which data science can provide fantastic insights and help you take better (data-driven) decisions, aside from marketing optimization. I’ll just list a few for you to consider, should you run a tour operator or DMC:

  • – Fleet efficiency: how about saving fuel and other costs by improving planning, for starters?
  • – Dynamic pricing: are you still giving the same price to everybody, the whole season?
  • – Revenue Management: when was the last time you flew on a half-empty plane?
  • – Predictive analytics: would you like to know where will your customers go next time?
  • – Demand creation: would you like to decide where will your customers go next time?
  • – Sales forecasts: would you like to know EXACTLY the allotments you need to buy next year?

Conclusion

The two modest examples of statistics usage I mentioned early on in this article took a significant amount of time to deliver usable results, perhaps in the order of days. And if I needed to replicate the experiment back then, the second time would be quicker, of course, but still quite long. With applied data science, the same exercises would take just a few hours to obtain results, but replication will be immediate and indefinite (as long as the data sources don’t change). In addition, I’ll be able to get truthful forecasts for the same price!

Technology might be giving an edge to the big players; nevertheless, is becoming more and more affordable to every kind of business. Data science levels the playing field, so small operators can go to war with the same secret weapons the monopolists are brandishing. Does all this sound too difficult to you? Cheaper and easier to use automation tools are entering the market (just like REVVA); and data-obsessed nerds like me can provide free advice to open-minded brains like yours.

Go on, drop me a line >> or comment below and let’s talk data science for the travel industry!

Marcello Bresin


Comments

One response to “Practical Data Science for the Travel Industry”

  1. GuQin Avatar
    GuQin

    Hi! Do you use Twitter? I’d like to follow you if that would be ok. I’m absolutely enjoying your blog and look forward to new updates.