So, when it comes to advertising, travel brands need to do whatever they can to improve revenue margins and get the edge over their rivals.
British travel brand Secret Escapes, which sells heavily discounted deals on luxury travel, managed to do exactly that when it set out to use machine learning to optimise its Google Ads, make its bidding setup more efficient, and improve Cost Per Lead.
Rumyana Miteva, Head of Search for Secret Escapes, spoke at Blueclaw’s Leaders in Travel: Digital Summit last month about how Secret Escapes succeeded in reducing CPL by 38% and boosting its ad click-through rate by 23% using campaign drafts and experiments on Google Ads, and adopting Smart Bidding.
She also shared the lessons her team learned from the campaign, as well as some general observations about employing machine learning in marketing campaigns.
Machine learning in travel: A short summary
Many of the most useful practical applications of artificial intelligence (of which machine learning is a subset) in marketing at present involve processing and analysing data.
With the arrival of the internet, the amount of data we can collect and access has increased exponentially – far beyond the ability of humans to manage or organise. Nowhere is this more true than in the travel industry, which in Miteva’s words, has a “staggering” amount of available data, with an infinite number of customer touchpoints.
It’s little surprise, therefore, that the travel industry has been quick to latch onto the possibilities offered by machine learning. As Miteva put it, any given task might take a team of professionals between 15 and 20 minutes, whereas a machine only needs 30 seconds.
Some examples of machine learning in action in the travel industry include:
- Recommendation engines
- Flight fare and hotel price forecasting
- Optimised disruption management
- Data-driven UX personalisation
- Intelligent travel assistants
- Dynamic pricing
- Fraud detection
When it comes to travel marketing, however, Miteva noted that because machine learning algorithms are “black boxes” – it’s impossible to know exactly how they work or what goes into them – marketers tend to avoid them out of uncertainty in what they’re investing in.
This is a shame, as the Secret Escapes case study proved exactly how beneficial the application of machine learning in advertising can be to a business in a competitive industry. Miteva urged the event’s audience of travel marketers to be willing to give machine learning algorithms a try.
“If you give them a chance, they learn and they improve,” she said.
The case study: How Secret Escapes reduced CPL by 38%
Because Secret Escapes is a members-only travel club that has built its brand and reputation around exclusivity, its growth and income rely on new users signing up for membership via the Secret Escapes website.
Despite running a number of marketing initiatives to encourage sign-ups, however, Secret Escapes was struggling to hit its target CPL (Cost Per Lead) goal with its existing bidding solution.
“We decided to put our existing bidding solution to the test, and see if we could identify a more efficient bidding setup,” said Miteva.
Using campaign drafts and experiments in Google Ads, which let marketers propose and test changes to their Search and Display Network campaigns, Secret Escapes was able to conduct tests across multiple territories simultaneously and compare outcomes.
The brand also adopted a Google Ads Smart Bidding approach called tCPA, or Target Cost Per Acquisition. This is a dynamic, automated approach to bidding that uses advanced machine learning to optimise bids automatically, and tailor bids to each auction.
Essentially, it aims to keep the cost per conversion equal to the target CPA set by the brand (whatever that might be), and to get as many conversions as possible at the target cost.
The results of the test, said Miteva, were “staggering”. On average, using tCPA Smart Bidding improved click-through rate for Secret Escapes by 23%, and produced 65% more conversions at a 38% lower cost per lead than the brand’s previous bidding setup.
As a brand, Secret Escapes has an incredibly complex bidding setup with thousands of possible combinations – so it was important to get things right, and brought immense benefits for the brand. Secret Escapes took its time running the tests first, before transitioning gradually over from the old solution to the new one once the tests had returned a significant, positive result.
Other machine learning tools for marketers
Ad bidding isn’t the only opportunity for marketers to integrate machine learning into their work. Miteva gave an overview of a number of other useful solutions on the market that use machine learning, and their benefits for marketers.
Dynamic Search Ads (DSAs): In Miteva’s words, these are the easiest way to set up paid search campaigns, and ideal for advertisers with a well-developed website or large inventory.
Data-Driven Attribution: This feature is only available to Analytics 360 customers, and generates a custom model for assigning conversion credit to marketing touchpoints.
A company’s business model is always evolving, said Miteva, and therefore “we can’t just live in a world of first click or last click.”
Universal App Campaigns: These types of ads are designed automatically based on your Google app listing. Google will decide which ads are performing best, and serve the top performing combination.
Similar Audiences: This feature, part of Google’s Display Network, finds new potential customers with similar search behaviour to people in your marketing list. In travel, said Miteva, it’s hard to hit users at exactly the right time – but using a tool like Similar Audiences expands the pool of relevant prospects.
Google Translate: Although it might sound basic – even silly – to suggest that travel brands make use of Google Translate, Miteva noted that neural machine translation (NMT) has greatly improved the accuracy of Google Translate, and urged travel marketers to give it a try.
Finally, Miteva shared the lessons that she and her team had learned from their successful foray into machine learning, in the form of some key takeaways.
- Bring ‘real-time’ data to life
- Target the right users at the right time
- Improve business efficiency wherever you can – it’s worth it!
- Deliver relevance at scale
- Remember: machine learning is still learning! Make sure you allow time for it to adapt, improve, and take effect.