Although we might wish we could – consumers don’t tend to impulse buy holidays!
NB: This is an article by Johnny Francis, Head of Data Science and Advanced Analytics at Forward3D
For marketers this presents a challenge, as the lead time between the initial engagement and the final decision can take place over the period of several months, whereas for more impulse-driven sectors like retail it is usually up to 30 days. In fact, when comparing travel to retail, everything from retention to measurement is different, and getting into the mind of consumers in order to understand their journey is vital.
When looking to improve performance within travel, marketers need to consider a series of key aspects: thinking from a multichannel perspective, considering the many different data sources available, the importance of testing and, crucially, how to tie all of these parts together.
Think Multichannel
Whether it’s PPC or Social, all teams have their own targets and taking a siloed approach hugely restricts the ability to consider marketing activity as a whole. Typically this can lead to using the dreaded last-click attribution methodology, which – given what we know about travel lead-in times – is unlikely to give marketers meaningful insight. It is only when a mix of all channels come together that marketers are really able to fully optimise results.
For travel brands, there must be a clear optimisation strategy in place built from a detailed analysis of the full customer journey. During the decision making process, consumers can engage with a variety of different touchpoints before making the final purchase.
With retail the first interaction can often be when the purchase is made, with travel the customers will generally interact with the site several times before they make a final decision while doing their research. This can include online browsing, visiting a high street travel agency, looking on social media and even calling the call centre before arriving at their conclusion.
Virgin Holidays has recently worked with Forward3D to look at one particular touchpoint. They placed a number of ads within Instagram Stories to promote Las Vegas, San Francisco and NYC. The videos were in real time and were designed to encourage consumers to book their dream holidays. The creative complemented their other online advertising and allowed them to engage the consumer over various touchpoints. Obviously consumers wouldn’t be purchasing through Instagram, but Virgin Holidays understood that this was a vital part of the purchasing journey.
Educating the various internal teams about the benefits of working together is important. Some businesses go as far to change the way that budgets are structured internally to encourage collaboration.
Consider Atypical Data
Some brands constrain themselves by making optimisation decisions based on a handful of data sources that are isolated from one another. As an industry we now have access to more data than ever before, collected from a variety of sources, and we should be using that information to collectively understand customer motivation and interaction points.
By utilising additional data such as purchase information and linking together multiple user sessions, brands can ensure that they are creating an accurate picture of the customer journey which is then used to inform the strategy when we discover what works.
More often than not this helps marketers understand that it’s not about constantly bombarding the consumer across multiple channels, as the optimum media mix will be much more subtle than that. Utilising these data sources helps marketers understand where and how the consumer needs to be encouraged to engage with the brand during the decision making process.
Test, Test, Test!
Joining together data sources where there is a deterministic link (such as creating a single customer view using a unique user ID to join different device and order IDs) is one thing, but it becomes much more difficult when we are trying to understand the link between a customer’s offline activity and the conversion online.
For example, how do we evaluate the causal relationship of a customer watching a TV advert and then purchasing a holiday online? We have some data for online sales and information regarding the nature and viewership of the advert but in this case there isn’t any precise way of matching the two data sets together.
Statistically investigating a relationship between the two is possible, and it would allow us to estimate the impact of TV activity on online sales. To do this we first set strong testing conditions to give us the best chance of obtaining significant results, then create a variety of models to most accurately fit the data and use these predictions along with the actual data obtained during the test to estimate the impact of the campaign.
This methodology is robust and flexible enough to answer a range of other questions such as:
- What would be the impact of turning off brand PPC?
- How does our OOH advertising drive online sales?
- What is the impact of running ads on mobile devices on desktop conversions?
Answering these questions continues to provide marketers with context around how our online media is performing. We have successfully used this methodology with our clients in the travel industry many times with great results.
In the case of a popular airline, we ran a test to calculate the impact of turning off branded PPC on overall revenue. In their case we found that this would have no impact, and as a result, the budget used for brand terms was then allocated to other campaigns where it had the potential to create a more efficient ROI.
Applications in the real world…
In real life, most businesses (even those who pride themselves on being “strong” data-wise) will eventually become aware of imperfections in their data. This could stem from the inability to link together different sources, missing or corrupted values or a break in data integrity.
When starting out, the most important thing for a brand to do is utilise the best data they do have in simple but effective ways to achieve the largest impact on your campaigns, rather than trying to create an “all-singing, all-dancing” set of models which are then too complex to use effectively.
The key is to start off small and improve incrementally, while achieving milestones along the way and gaining additional understanding at each point.
By beginning with some small steps, brands can evaluate the outcomes and use your learnings (both positive and negative) to inform your strategy going forward. This helps to build trust in the process without staking too much, and eventually each piece will end up adding to an outcome which is greater than the sum of its parts.
