Whenever I join a company, the first order of business is to learn what we’re doing today. This means sitting in on meetings, observing engineer workflows, understanding business processes and most importantly, sifting through large amounts of quantitative and qualitative data wherever possible. Bridging the gap between internal procedures and customer-oriented data analysis will help understand pain points and future opportunities.

When I joined ID90 Travel back in 2015, they did not have any analytics strategy in place. Understanding which pages customers were interacting with the most or which channel was most fruitful was entirely unknown. However, after about 4-weeks, we were able to put into place a solid analytics program. With analytics in place, we could begin uncovering some extremely interesting trends and valuable opportunities.

My first “Aha!” moment

This may sound silly, but for a long time, the majority of ID90 Travel customers didn’t know we sold hotels. Even though the homepage had a hotel search form and accommodations language, customers were completely overlooking a key value proposition.

When we surveyed customers, we found there was a complete disconnect. They thought of our brand as a tool for only booking non-revenue flights. To be fair, it makes sense when you consider we only sold non-revenue flights for nearly 6 years prior. So, we had to dive deeper. We needed to find a way to better juxtapose our hotel product with our active flight users.

This is when I learned we were booking 10x the amount of flights a day compared to hotels. What’s more, only 4% of flight bookings had a relevant hotel booking (i.e., an attached hotel). This is also when we noticed we were not merchandising any hotels post-flight booking. We knew where the customer was flying and when. So, we got to work on a novel hypothesis. Could we improve hotel adoption by bridging the gap between hotel and flight data to better merchandise relevant content?

Data Analysis into Marketing Opportunity

Our team began reviewing everything we knew about flight bookings. We had to normalize a lot of data so we could align it with our hotel data structures. Once we were done, we began feeding flight booking data into a new micro-service we named Need a Room. We also began feeding hotel booking data into the same service to join the two products.

The Need a Room service acted as an intermediary messaging bus for our customers. The service read flight booking data and would determine if the same customer had a relevant hotel booking. If the customer did not place a booking, Need a Room would generate a message that we could reuse across the customer’s experience.

Our most notable example of this was our email program. The email would trigger one hour after booking a flight if the customer didn’t book a hotel. The email simply educated the customer that we offer great hotel deals in the area they’re flying to. What’s more, we positioned on our Watchlist feature as a way for customers to simply monitor hotel prices overtime.

The email program was a success. It is by far the most high-volume and high-touch point marketing program we have in place to date. We saw more and more customers interacting with the email and booking hotels. But we didn’t stop here.

Extending data to multiple use cases

We began leveraging Need a Room to create dynamic merchandising widgets others could use external to our application. Any partner could pass a few parameters to our API and our service would return pure HTML with highly relevant content. This was great for post-flight bookings on some of our partners like United Airlines. In most cases, we would surface our most popular 3+ star hotels in the area the customer was flying to.

This Need a Room widget turned out to be one of the most valuable pieces of merchandising software we built. Once we implemented it on Partner websites, we saw an explosion in hotel bookings. In a way, we hacked our growth by integrating our technology into an already established ecosystem. And this process is highly repeatable too!

Lastly, we created a push notification series that would send a message to the customer 24-hours post-flight booking. Similar to the email program, this would only send if the customer hasn’t booked a hotel. However, we also put in place other limitation mechanisms to avoid bombarding customers. For instance, customers could only receive a maximum of one Need a Room push notifications every 14 days. In addition, push notifications only triggered for flights with a departure greater than 72-hours from the time of booking. This led to an even more engagement and by virtue, further adoption.

So we built an email, push notification and dynamic merchandising program. We had to ensure we could measure the success of each respectively and independently of one another.

So we put into place the proper UTM parameters to track Need a Room service-related traffic from EmailInternal Cross Promotion, External Cross Promotion, and Push Notification. Once we had all of these measures in place, we could understand how each marketing touch point worked. With that, we could focus on the messaging that was most rewarding and tweak others.

Use the 80/20 rule and double down on what you have

You’ll notice during this time, we only relied on the Need a Room service for all automated marketing decisions. Need a Room became the central messaging bus, the cerebral cortex of the entire operation. This is important. Instead of building several pieces of software to help market to our customers, we relied on a single source. When we established this new source of truth, it was easy to extend the data to future use cases.

It’s also important to note that we shipped various versions of Need a Room. The MVP was the data service and email program. The email only had an image, some educational text and a call toa action. We later made the email more dynamic and optimized the subject line and body text for optimal open & click-through rates. Speed trumps everything in software development. It is better to collect data on a subpar product early than it is to collect data on a fully refined product. In the former instance, you can redefine your direction early based on quantative and qualitative data you capture. In the latter instance, you’ve already committed way too many resources into a hypothesis that may be incorrect.

You’ll also notice that we didn’t simply move onto the next project once we finished the email. We doubled down on exposing the Need a Room data to other mediums. We built a dynamic merchandising widget. We tested different formats like a carousel or full screen modal. We translated the content for international customers. We then extended the product messaging to push notifications.

If you find that something is working, use the 80/20 rule. 80% percent of the time, your team should double down on the channel or feature that is working. The remaining 20% of the time should be reviewing other long-term opportunities you’ve discovered from the shift you’ve imposed from your core efforts.

Nearly Tripling Our Hotel Attachment Rate

At the end of everything, we went from 4% to approximately 11.5% in just over a year. Imagine a 300% increase in hotel bookings all from a highly scalable, highly targeted programmatic marketing service. Today this service continues to run without much maintenance. It’s become a core retention and activation loop for our airline partners. From here, we built several other services with the same framework in mind. We built services that monitored search, product views and checkout behavior. Each of these services have their own use case and data sets that can be highly extensible.


Take the time to understand your quantitative and qualitative data. Ensure you have a tagging strategy in place to understand what your customers are clicking on, what they’re searching for, where they’re going on the site. Leverage surveys to gather further insight and answer questions that can’t be solved with behavioral tracking. Bring in customers to have conversations about their needs, their goals and their ambitions. Take the time to speak with your engineers and understand what data you have available, what limitations exist today and how you can put in place a plan to leverage the data for marketing purposes.

Today’s marketers have more data at their disposal than any time before. The question is not if we can measure, but what to measure and what questions we need answers to.