Revenue Management, or the sophisticated dynamic pricing algorithms used by airlines, is considered a science. RM taps into large historic databases, it builds regression-based forecast models, and it calculates probabilities across demand levels to optimize the number of seats allocated to each group of fares. Analytically, it determines how many $99 tickets to sell and how many seats to save for $599 passengers.
Of course, despite how sophisticated the models are, RM still relies completely on inputs to the model – a critical input is average fares for each range of fares: $49 - $99, $100 - $139, $140 - $175, and so on. The model, for example, calculates average fares in each range to determine the potential upside of one fare range versus another. For this, it requires a constantly updated view of the fares. The analytically derived optimum allocation of seats changes if the average value of the $49 - $99 fare range is $55 or $65 or $95. If the average fare is only $55, demand within this fare range is considered of relatively low value -- there is tremendous upside to holding seats for higher fare demand, $100 and up. If the average fare is closer to $95, however, the upside to the next fare may be very low -- driving little incentive to “hold out” for higher fare demand.
The valuation of each fare bucket comes from Revenue Accounting. Revenue Accounting calculates the value of each fare and thus each group of fares. If Revenue Accounting is inaccurate, or out of date, the Revenue Management model can’t properly allocate seats across the spectrum of fares. If the value of the $49-$99 range is actually $65, not $55, the upside of holding a seat for the next higher fare is $10 less and the model will therefore tend to hold fewer seats for the higher fare. The model will instead sell another seat, or more, in the $49-$99 bucket.
In the dynamic travel marketplace, the “value” of each bucket can change on a real-time basis. Matching a competitive low fare, for example, will drive the value of the lowest valued “bucket” down. Implementing a fare increase on all non-promotional fares will drive up the value of all higher “buckets”. An increased mix of certain passenger types – more 7 day advanced purchase or more connecting passengers from Frankfurt or more passengers qualifying for a large corporate discount or a currency swing – will result in changes in the value of one “bucket” versus another. Another complicating factor is interline traffic: interline requires proration of a fare published and sold by another carrier. In most cases, the interline relationship specifies how a fare will be divided but the other carrier actually sets the fare; this means revenue accounting needs to communicate with the other carrier and jointly calculate the portion of the fare the operating carrier is due. Interline can both complicate and delay determination of the average fare in a fare grouping.
Such changes can, over time, be large enough to require a complete re-ordering or re-classification of fares. “Fare inversion” occurs when a fare level that was defined to be higher than another turns out to have a lower average value. In this case, the RM model treats the two as one larger category and does not set aside seats for what is now calculated to be no real upside.
If the system can accurately and quickly identify fare inversions it can still determine an optimal allocation of seats across the remaining spectrum of fares. But if the system can’t accurately, or quickly, update the average fares it will continue to use an outdated value; it will allocate seats on wrong information. If there is fare inversion but the system doesn’t know it, the system will save seats for the “higher” fare even though there is no difference in actual value. The system could actually turn away a “lower” fare passenger even though he is worth as much as the potential new passenger – who may or may not actually show up.
An outdated calculation of the average fare in a “bucket” can completely undermine the value of the RM system. Therefore, ideally an RM system has a direct link to revenue accounting and revenue accounting in turn is updated on a real-time basis. Since, in many cases, neither of these conditions are met completely, many Revenue Management systems promise updates based on as much information as is available at any point in time. All RM systems must strive for such timely and accurate updates in order that these key inputs to the RM model remain valid; the value of any RM system always relies on the accuracy of the inputs.