From 51 million domestic passengers in 2005 to 126.7 million in 2018, the airline business in India has grown exponentially in the last decade. To cater to this growth airlines have undertaken several technology initiatives.
The idea is to use technology as an enabler. There is a rush towards automation and the use of algorithms. Automation increases efficiency and reduces variability thereby providing more certainty on the experience. Algorithms build on patterns gleamed from data and ideally should predict the nature and quantum of demand. Yet automation and algorithms often overlook unique market dynamics. So what works well in the west does not necessarily work as well in developing markets like India. Why so?
Market growth includes several variables and the past is not a predictor of the future
Passenger demand for air travel has grown exponentially. The country saw 51 million domestic passengers in FY11, which grew to 61 million in FY14 and 137 million in FY19. This at a consolidated level.
Digging deeper, the cities that are growing fastest have very different characteristics. Take for instance, Surat and Chennai.
Both cities are projected to quadruple GDP in the next twenty years. Yet the bases and the determinants are very different. While one city has a population of 6.8 million, several small business units and consolidated price elastic demand, the other has a population of 10.7 million, several large manufacturing facilities and segmented demand. Growth in demand is a consequence of a variety of factors and forecasting algorithms simply cannot take into account qualitative factors.
If that wasn't challenging enough the transportation ecosystem as a whole is changing. There is an exponential improvement in rail and road services and consumer behaviour has changed. For instance, in the context of Bengaluru, Hyderabad and Chennai one clearly sees that in these cities demand for bus services has taken off and this is a consequence of the location of airports, the improvement of road infrastructure, the nature of demand, the intent of travel and the city-centre to city-centre connections.
The pricing differential between competing modes is 2X or more and a segment of travel is choosing buses overflights. As rail services improve additional demand will shift to rail as well – as has already started in several parts of the country.
Algorithms assume a static system and this is just not the case. The past is not a predictor of the future.
Pure data mining does not reveal true nature of segmentation “Data is the new oil” is the latest catchphrase and data mining, machine-learning and predictive-modelling are buzzwords often thrown around in the industry. Almost every other vendor has a data-mining solution that promises to fully-automate functions and reveals trends that can be applied towards extracting more cash (read value) from customers. Yet some of the data mining solutions fail even on the most basic level.
For instance, airlines are keen to understand the nature of demand. And indeed all demand is not created equal. A high percentage of corporate demand is ideal as it yields more and is price-inelastic. But data mining can only reveal so much.
The demand can be identified with a variety of tools including the channel, unique access codes and even day of travel and time of booking. However, digging deeper there are clear trends where corporate demand is booking via patterns that are not aligned. In fact, as budget constraints grow and as companies look at travel costs, a sizeable amount of corporate demand is flowing through non-corporate channels often booking in patterns where it can be easily confused with leisure demand. In this case, the automation not only provides incorrect data it leads to incorrect decision making.
Calendars, holidays and changing patterns
Forecasting algorithms are traditionally built on western calendars. In these calendars, holidays are fixed thereby making automation easier. For instance, Christmas and Thanksgiving always fall on the same day thus algorithm will tell you to increase airfares around that travel period. Attempt this with Indian holidays and the algorithms go awry.
Both because the majority of the Indian holidays are based on a lunar calendar and also because the demand patterns keep changing. The change in demand patterns is an aspect that cannot be looked at purely quantitatively. Whether it is flooding in Kerala, over-tourism in Himachal or activity in Bagdogra the shifts are interesting and indeterminate.
There is also the unique nature of travel. Certain cities in the North see a spike in demand for mountain regions and hillside destinations in the summer whereas a similar trend is not seen in Southern cities; demand for beach destinations from southern cities is less compared to those from the North; cities like Goa that earlier had extreme variability of demand have seen demand curves flattening by virtue of pricing and promotion decisions – not by airlines but by hotels.
Overall while demand is growing by leaps and bounds, there is also a clear shift in demand patterns. The drivers are many including demographic changes, behavioural patterns, technology trends, socio-economic changes and pure supply-demand dynamics. Understanding these requires complementarity where managers not only understand and take action on algorithmic outputs but also understand the architecture of the algorithms themselves.
As automation increases and algorithms become even more pervasive, airlines cannot simply rely on machine outputs for demand predictions. The challenge is to find informed and focused managers who are able to integrate diverse inputs and combine it with quantitative methods towards understanding segments of demand and patterns therein. An intuitive understanding of the market coupled with an understanding of technology is required. In a rush to automation and algorithms, airlines cannot forget unique market dynamics. The challenge of forecasting demand patterns continues to be a key focus area for airlines.
Satyendra Pandey is the former head of strategy at a fast-growing airline and a partner at the research and advisory firm Siyasha. Previously, he was with the Centre for Aviation (CAPA) where he led the advisory and research teams.