A sponsored viewpoint article by Keesup Choe, CEO of PredictXHow do you think the use of data has evolved in travel in the last five years?
Seven years ago,when we first started working with travel data, there remained a strong adherence to the monolithic model of travel data - that is, an architecture driven almost exclusively by TMCs carrying out the majority share of bookings by GDS as well as monthly reporting on those transactions produced by their own post-trip datasets.
This situation has evolved rapidly since, and the current state of play is leaning towards the polar opposite. Firstly, the single monolithic GDS-TMC booking structure is now being increasingly replaced by a fragmented multiple-source environment containing direct bookings, GDS-transacted TMC activity, TMC punch outs, use of the sharing economy, non-TMC aggregators and a host of other sources of booking data. This has dramatically increased the complexity of the analytical environment for travel managers.
This makes it more difficult to calculate total trip costs as well as understand both the true costs of travel to a location which will gain visibility into 100% of T&E spend
Secondly, businesses now understand that infrequent, regular reporting based on post-trip data is not sufficient to manage travel programmes at their optimal level and is particularly ineffective at driving behavioural change. One of the first changes we saw was the understanding that stale data is merely a historical record and only timely data can drive meaningful action. It was now understood that data needs to be in the hands of the manager undertaking that action within hours rather than weeks.
The older "analyse and fix" approach to data analytics has now been moved into a "predict and prevent" model, looking at where issues are likely to occur before they do so and taking action before the event. The move from looking at data seven or eight weeks out of date to looking at what is going to happen several weeks in the future is a significant evolution that we have seen among the leading programmes we work with.
Another area where we have seen significant evolution over the last five years is the understanding that, on a cost basis, booking data only holds around 50-60% of total programme activity. Five years ago we were quite evangelical about this, and it felt a little like we were one of only a few voices with this message. Now I think that understanding has become mainstream.
Moreover, the past five years has seen a proliferation both of areas that are viewed as addressable with data-driven decision making (think ground transportation) and of systems on which activity is being booked. This has been driven at least partly by the development of sharing economy-based platforms, but also by classic transport and accommodation suppliers providing much richer datasets.
Finally, it's not just the demand that has changed, it's also the technology we use to clean, consolidate and analyse the data as well. Artificial intelligence is starting to allow us to automate a lot of the process of data analysis itself, which is especially exciting when you understand this capability can finally allow a machine to tell a human manager "the data says you should do this." I think we're just getting started here - and a lot of the headwind is social acceptance rather than a lack of technical capability. We were the first to apply machine learning to the travel data problem and are confident the results will continue to improve very quickly.
Do you think buyers have a better hold on data now? What's the next stage?
Travel buyers who have embarked on achieving a data-driven travel programme understand that this journey is a long-term one. Integration of travel booking data with card and expense to have a single version of the truth is a straightforward first step, however it cannot transform the travel programme on its own. The integrated data is akin to a weighing scale. The scale by itself does not result in weight loss. But without the scale, it is impossible to monitor whether your weight loss plans are working.
What we are talking about is a data strategy that runs parallel with the travel strategy. More and more organisations are forming their data strategies as their businesses aim to become data-driven for optimal decision making and automation.
Depending on where the organisation is on their data-driven strategy journey, and their alignment with overall corporate strategy, they will have different priorities for their specific programme. A common end-goal for many of our forward thinking customers is, however, the increased adoption of prediction and forward-looking. This moves travel managers away from an "analyse and fix" mentality to a "predict and prevent" mentality, which in turn will allow them to have even greater impact on policy compliance, spend management and duty of care processes.
How are you applying predictive analytics to travel management?
Travel is no different than any other modern business service. It is getting more complex, with ever more diverse content sources and more demanding customers. At the same time companies want greater efficiency and lower cost. It is impossible to succeed on all these fronts without leveraging the latest in technology. If you were the last accountant to use paper spreadsheets and resist moving to Excel, you'd surely be out of business. Today AI is the new spreadsheet. Choosing not to leverage its power is living in the past.
Our predictive analytics module is concerned with not looking at what did happen, but what will happen. This will hopefully drive more travel managers to proactively manage rather than reactively mnanage their programme.
What are the main data challenges that buyers struggle with? How could they overcome them?
Many travel managers believe that the most challenging aspect of data management is to develop a single source of truth. Most buyers operate in global travel programmes that draw from more than one data source. This results in multiple data sources with multiple supplier formats. Added to this, is that many operate in multiple regions with multiple departments - each with their own picture of spend. The result is a lot of fragmentation.
This fragmentation of data is something they all struggle with. Matching these different data sources is time consuming, challenging work. If it is not done, however, we have no accurate picture of spend. Enriching that data by correctly identifying chains, brands, and alliances arms the sourcing manager with accurate volume and value figures to inform more powerful negotiations.
Furthermore, travel data is unique in that the core transaction data is created, stored, and managed by the third party - usually the TMC. In other areas of the business like finance, inventory, or sales, the systems that hold these transactions are usually under the operational control of the business itself.
This lack of control makes it more difficult for travel managers to manage the data as any data gaps or remedial work cannot be performed on the underlying systems. Instead of fixing the system, they must fix the data. Obviously manually cleaning the data can only be performed on a one-off basis. Rather, sophisticated data cleansing automation systems should be deployed to productionalise the process.
Is total cost of trip still important and how are you working with buyers to achieve it?
Total trip cost has always been important as you can't drive a travel programme if almost half of your vision is blocked. Most buyers we spoke to operated from TMC data primarily. This was concerning as TMC spend, while rich in detail, misses corporate credit card spend and expense spend. This missing spend works out to be around 41% of the total value.
When it comes to data, if it is not accurate or complete, it is not a measure which can be leveraged to implement change. For example, if we look at TMC data only, air spend may seem like a huge portion of it. If we analyse and examine all data sources we find that this percentage is incorrect. There is, however, no need for action regarding high air spend.
What is also important is that, when it comes to negotiating with suppliers, the supplier should not have more visibility into your own company spend than you do. In spite of this, due to a partial, incomplete picture of spend delivered weeks after a trip has taken place, many travel departments do have less visibility than their suppliers.
Nobody would come to a key negotiation or meeting with only half the information, yet, with the level of data visibility available to most travel managers, often this is the case.
This is why a full, accurate picture of spend can save a travel programme between 5 and 20% if it is leveraged into action. When the numbers we work with are not right, we cannot achieve any of this.
Is 100% programme visibility an achievable goal travel managers should aspire to? How can a total trip cost help with that?
The 100% programme is ambitious and aspirational. This idea incorporates both "total trip" cost and also the reconciliation of this cost to the figures recorded in the general ledger. When 100% of the total trip cost is mirrored in the GL, true 100% visibility is achieved. The importance of this is that any savings realised by the travel manager is "addressable" and reflected in the GL and thus dropping to the actual bottom line - money going out of the bank account..
100% programme visibility is aspirational as the goalposts are constantly changing. There are simply so many data sources popping up all over the place. Meetings data, and sharing economy data are some of the new sources which have recently come into play. Once we get used to working with one data source, another data source is then introduced. We still want to use as much data as possible. We believe the more data we have, the closer we are to achieving 100% visibility of spend so each new data source needs to be integrated in a manner that can deliver value.
A total trip cost measures the spend that takes place across a trip. It is defined as spend going through the TMC, credit card and expense management system. It is important as we can pick up on things that a partial view of spend will not show. Having a total trip cost will not guarantee you 100% visibility against the GL, but it will be a giant first step towards that.
What could travel learn from some of the other industries you work in?
Travel, as a business function, is incredibly unique. The travel department is often quite small yet they handle so much money for so many people. Unlike any other procurement category where there is a list of suppliers with transactions that get approved before they take place, travel is subject to the purchasing decisions of thousands of travellers everyday. Added to this, they often don't have operational control of their own data and have to work with what they get from their supply chain. We think this level of demand is quite unique but is similar to another industry we operate in - the care and health industry. Often they have limited resources with high demand which equals the same complexity and scope that we see in travel. That is why many healthcare authorities are turning to digital solutions and, data, to predict and prevent blockages in the system and direct their resources to the optimal place for maximum growth.
Healthcare is, in fact, increasingly becoming more innovative and digital and, in travel, we can take this spirit and change how they work for the better.
Travel managers constantly feel not only as if they are drowning in data, but also often don't have time for achieving their strategic objectives as they are steeped in day-to-day response to ad hoc requests, lengthy RFPs and many are still using manual data reporting.
Travel managers can use technology to make their lives easier in the same way care and health industries are doing so. Machine learning can, to a large extent, automate data reporting to deliver an accurate picture of spend in a timely manner. With the data covered, travel managers can focus on more strategic planning activity.
How is PredictX applying AI to data analytics for travel?
PredictX uses machine learning which is a type of practice used to achieve artificial intelligence. For example, if AI is teaching a machine to think like a human, machine learning is the method which we use to achieve that. In travel, machine learning algorithms are picking apart the data from multiple sources and grouping transactions based on probabilities to create the total trip cost. Prior to machine learning, this type of data might be aggregated using a static rules set that would miss elements like whether a hotel room transaction was actually part of a meeting expense. Machine learning algorithms can figure out new exceptions and possibilities without manual intervention.
You released a digital assistant to market. How can this type of assistance help travel managers get a better hold on their data?
Using tools like AI and natural language processing (NLP), we are implementing RPA (robotic process automation) and our digital assistant is part of that - almost like a digital agent, or employee tasked to perform specific tasks. Some of the "agents" are an interactive reporting tool. You can ask them questions like which business area is forecasted to exceed budgets and the agent replies to you via chatbot or voice. This way our team or stakeholders can get exactly the answer they want without trawling through reports and dashboards.
The more exciting part, however, is where the agents can be in the background looking for compliance problems, fraud and travel disruptions, etc. The agents will allow us to catch issues when they occur and deal with them straight away. Our digital assistant formed part of our initiative to help travel managers get valuable work done. There is no denying data's ability to facilitate programme growth, but often we found there were constraints preventing managers from putting data into action. Travel managers do not have the time to constantly comb through dashboards to find one piece of relevant data that they need now, whether it be a problematic tax issue arising during a trip or communicating last year's department travel spend to a department head.
What does your partnership with Festive Road entail?
Festive Road and PredictX both share a similar cultural DNA. We want to support travel managers whose ambition is to drive positive change in travel management. We also realise that, while our goals and our culture are similar, our skills are different. We are technologists focused on building better tools that help our clients identify areas for improvement and scenario-modelling tools that help decision making. Festive Road are experts in integrating the data strategy with corporate strategy and can help deliver the change needed.
Through partnering with the consultancy that has been on both sides of the travel industry (buyer and supplier), we believe we can develop a comprehensive programme of solutions for the buyers we operate with and deliver both a data-driven and qualitative-driven solutions.
A new product we are releasing is our Health Check. As the name implies this is a thorough data-driven review of the travel programme to identify the current state of the programme, how it is poised to reflect the strategic future ambitions, and identify quick, medium, and long-term actions necessary to achieve these ambitions. Much like a check-up at your GP, we provide the laboratory results and Festive Road interpret these based on current market as well as practical state-of-the-art.
The health check is just one of our pipeline of solutions to nine key industry problems and are now in the process of developing these solutions through joint workshops.