Revenue management? Nothing new to learn about that - it has been around for so long, and its logic is so widely understood, that it is no longer a competitive issue and all its relevant problems have been solved. Right?
Wrong. To be sure, since American Airlines introduced revenue management to the airline business, carriers and suppliers of information technology have developed countless refinements and advances in revenue management systems. These will continue. But, even as information technology becomes more powerful and the steering algorithms more sophisticated, the basic logic of revenue management remains unchanged. So do the weaknesses that afflict it.
There will always be three basic ways to squeeze more revenue out of that highly perishable product, the airline seat. First, value the revenue contributions of various passengers more precisely. Second, match the available prices more closely to customers' willingness to pay. Third, forecast passenger demand more accurately.
Despite the progress of the last 15 years, there is still considerable room for improvement in all three areas. A look around the industry shows that forecast errors still routinely exceed 30 per cent, rising to as much as 40 per cent, on individual flights. Pricing-to-market continues to be more an art than a science, even in highly competitive markets such as the US. In more regulated markets, most carriers are still getting used to the idea that pricing is a competitive tool and not an administrative task. Most perversely, even the best revenue management systems suffer from weaknesses that cause them, in certain situations, to displace higher-value passengers in favour of lower-value passengers.
These common faults of revenue management can be traced directly to a small number of unresolved problems that will be the focus of revenue management departments in the next few years. Although technologists and software engineers have yet to come up with specific tools to deal with them, the outlines of a solution can be discerned.
The problem of valuing the revenue contributions of passengers more precisely has long been a thorn in the side of revenue managers at network carriers, since the true value of a passenger on a connecting route is not always apparent to a system that just looks at individual legs. The last real quantum leap in revenue management technology - the introduction of true origin and destination (O&D) seat purchasing control - was a major advance in this area.
O&D yield management systems are able to value passengers according to the total network contribution of the travel itinerary. Whereas a conventional yield management system always prefers a customer in a higher booking class to one in a lower booking class, an O&D-based system is able to say, for example: "Wait a moment - the business class passenger on the shorthaul domestic flight is going to displace the economy passenger who would connect on to one of our longhaul flights, thus bringing more total revenue."
For a carrier with high load factors and extensive networks, the multi-million dollar cost of a new generation system can easily be outweighed by an increased revenue potential of as much as two per cent. Quite a few airlines, including Lufthansa, SAS, Cathay Pacific and the US majors, have been willing to lay out the required sums.
But there is a problem here. O&D-based yield management systems do not function properly on codeshare flights, which is to say on an ever-growing proportion of flights. Although codesharing is relatively new to Europe, it already accounts for almost 13 per cent of all scheduled flight numbers in the 1996 European winter flight schedules, up from less than 10 per cent a year earlier. The reasons why O&D yield management and codesharing do not yet work well together are partly technical and partly regulatory, but the results are the same - codeshare partners fail to get the maximum passenger revenue for their product.
For example, this midyear, a passenger could have booked a flight between the European and US hubs of a certain transatlantic alliance. The lowest available fare from the operating carrier was US$1,012. The same flight could be had from the codesharing partner for US$662.
Does this mean that the yield management departments made a mistake? Not necessarily - both airlines may well have optimised their seat contingents perfectly. Yet the fact that the European partner was presumably rejecting passengers who would have been willing to pay US$1,000, who might then have flown on the same flight with the US partner for US$662, indicates that the two partners were not reaching a mutual optimum.
The significance of this problem should not be underestimated. By and large, codeshare flights are filled closer to capacity than single-carrier flights. This means that the expensive new yield management systems are most likely to fail on precisely the flights where they are needed most. As codesharing grows to cover more flights, the sums of money left on the table are bound to grow ever larger. Airlines must now solve this problem without revamping or scrapping expensive investments in revenue management technology.
Easy answer
There is one easy answer, but like most easy answers, there is a catch. By sharing inventory and using a dated revenue pooling arrangement, codesharing partners could unify their yield management and let one partner take over the booking control. The catch is that antitrust authorities frown on revenue pooling.
A second solution would be to exchange and harmonise forecast data, so that both airlines have the same revenue expectations for the marginal seat. This could help eliminate the most crass examples of contradictory inventory control, but would be at best a half-solution. When it comes to practical implementation, any solution that required revenue management departments to rely on a partner airline's data would run up against psychological obstacles. At most airlines, it is difficult enough to get one department to accept data from another department in the same company. Usually both say the other's data is garbage. Often, they're both right. Imagine the same position between different companies on different continents with different languages, and the practical difficulties are clear.
A final possible solution to the codeshare problem might be found in what could be called "virtual contingents", in which codesharing partners would essentially bid for seats in their partner's contingents. This solution would require substantial technical changes, but the principle is quite simple. O&D yield management systems judge reservation inquiries individually, a capability known as "seamless availability". With a virtual contingent, the yield management system would not reject a reservation inquiry without first inquiring if the same booking could be made on the partner's contingent at a lower fare. If it could, the first airline would automatically "purchase" the seat from the partner's contingent and sell it at the higher fare. The partner airlines would then split the difference - the money that would otherwise have been left on the table - according to a pre-agreed formula.
Fully integrated
This solution comes close to giving the advantages of a fully integrated yield management system without requiring airlines to give up the strategically critical revenue management function. It would have the further advantage of taking into account the different network contributions of passenger itineraries in two different networks. For example, Austrian might have a higher yield expectation for a marginal seat than Delta because Austrian expects transfer passengers on East European connections that are not even offered by Delta. Virtual contingents would allow the airline with the higher yield expectation to "bid up" the price of the partner's seats.
The second way to improve the effectiveness of yield management is pricing-to-market, or matching the available prices more closely to customers' willingness to pay. Everyone's first lesson in economics is that different customers value the same product differently, and that a producer can maximise his total take by selling to each customer at or near the highest price he would pay. This idea is at the root of airline revenue management.
The problem with putting this into practice is that no one really knows what the demand curve looks like. When airlines match prices and restrictions to availability, they are largely working in the dark, and anyone who is waiting for a technological magic bullet to solve this problem is probably going to be disappointed. There is no real breakthrough in decision technology for pricing just around the corner. It has been suggested that ticket auctions on the Internet (an option with which American Airlines is experimenting) could provide useful data on price elasticity. But it could also offer a first class opportunity for manipulation by the competition.
So must airlines simply resign themselves to these imperfections? Not at all. The greatest potential for improvement in airline pricing comes not from investing in more sophisticated decision technology, but from speeding up internal pricing processes. Simply, the faster the pricing processes the more often they can be repeated. The more often they are repeated, the more data can be generated for analysis. The more the results are analysed, the more the organisation can learn from experience - and experience is the alpha and omega of pricing-to-market.
European airlines especially stand to gain from improved pricing processes because they do not have the benefit of years of experience in deregulated markets. These airlines would be well advised to exhaust the potential waiting to be realised in their pricing processes before they begin investing huge sums in new pricing decision technology.
Finally, there is the challenge of forecasting passenger demand more accurately. It is no exaggeration to say that accurate demand forecasts are critical for airlines in competitive markets. The formula is simple - the better an airline's forecasts can identify overfilled flights, the less revenue will be lost through selling below value. The better the forecast data can pinpoint no-show rates, the more aggressively the airline can overbook and eliminate revenue loss due to empty seats. Recognising this logic, the best airline revenue management departments have hired mathematicians to work out a way to improve their use of historical data and sales expectations to come up with accurate forecasts. Several airlines already boast of forecast errors of only one or two per cent.
The problem with these statistics is that they are subject to the law of averages. Even extremely small cumulative forecast errors can disguise substantial mismatches on individual flights, as anyone who has ever experienced "involuntary denied boarding" can attest.
The historical flight profiles and future expectations typically used to forecast passenger numbers are undermined by the sheer unpredictability of human behaviour. Here, airlines will not be able to achieve a real breakthrough in forecast accuracy until they begin to examine passenger behaviour at a wholly different level. Instead of trying to squeeze the marginal improvement out of their flight profiles, revenue management departments will have to look at the behaviour of individual passengers.
As a first step towards improving no-show predictions, the passengers booked on a particular flight could be analysed according to criteria such as type of journey, fare type, and the risk of missed connections. Airlines could use these data to develop models predicting no-show rates for particular passenger profiles.
The next step to improving passenger forecasts - one which is admittedly some way down the road in the technological future - would be to predict the behaviour of individual passengers. Airlines are already sitting on a mountain of information on passenger behaviour that has not yet been exploited for revenue management purposes, like data from the check-in and revenue accounting departments or frequent flyer programmes - the US alone has some 32 million FFP members.
We have estimated that improvements in forecast accuracy alone can bring a major carrier, depending on its specific circumstances, benefits equal to around 0.6 per cent of revenues by reducing involuntary denied boardings and no-show losses.
Diminishing returns
But even gains of this magnitude have to be weighed against the costs of developing and maintaining a new generation of forecast systems. The question of cost-benefit analysis brings up the more general problem that, as with every investment, improvements in revenue management are subject to the law of diminishing returns. As revenue management nears its goal of collecting from every passenger the maximum price he would be willing to pay, capturing the marginal dollar of revenue becomes increasingly difficult and expensive - a lesson learned by more than one airline.
But it is important to remember that what makes little economic sense today could well make sense next year, or the year after. As the price of a megabyte of memory continues to halve every 18 months or so, ever more data-intensive applications cross the threshold of profitability. The economics of revenue management systems are also highly sensitive to economies of scale. As airlines grow and come together in alliances, the opportunities for joint development or joint use of new-generation systems increases.
All these issues of revenue management ought to command the attention of the top level of airline management. Just as politics are too important to be left to politicians, revenue management is too important to be left to revenue managers.
Source: Airline Business
