How we launched advertising for a client—a large company that sells airline tickets online

Finding the optimal campaign structure

Also, depending on the number of search queries, you can restructure your campaigns. For example, there is a super popular query: cheap flights. On the one hand, you can leave it in the general campaign. But on the other hand, it can take up to half of your budget. Therefore, sometimes it is worth moving such words to a separate campaign in order to monitor their effectiveness separately.

Competition for such keywords is very high, and you need to carefully choose your bidding strategy. You can use “maximum conversions” or “target cost per conversion.” And sometimes it's better to just use the “always at the top of search” strategy so as not to lose impressions.

Sometimes, smart bidding strategies underestimate bids because competitors raise them too high. As a result, the ad is completely removed from the impressions. This is especially critical during the high season, when it is essential to be at the top of the search results. In this case, you either have to adapt your strategy or temporarily switch to manual management models.

Sometimes it is useful to divide even within a single query. For example, a separate group only for the exact variant “cheap flights,” and all variations and typos in another. At the same time, negative keywords are added to both groups so that they do not overlap. This helps to more clearly control the distribution of the budget and see which options work better.

Next, you can divide campaigns by device. Google Search allows you to separate desktops, tablets, and mobile devices. Effectiveness can vary greatly there. For example, competitors may overheat search results on mobile devices, in which case we choose the “always at the top of search” strategy. On desktop, things are calmer, there is less competition, and it makes more sense to stick with the “target cost per conversion” strategy.

Thus, we had several types of campaigns at once:

1. Campaigns for super-popular queries such as “cheap flights.”
2. Campaigns for destinations that have historically shown good results.
3. Experimental campaigns for new destinations where there are search queries but the client does not yet have sales statistics.

Third-party bid management services and offline conversion uploads

In addition, we tested third-party bid management services such as Kenshoo and Marin Software.

It was difficult to evaluate their effectiveness, as everything depends heavily on seasonality. But overall, when switching to these platforms, we saw a 10-15% increase in ROAS with the same budget.

Whether this was a result of these particular services or simply a coincidence within the six-month season remains an open question.

Given that a 10% increase is within the margin of statistical error, we decided to abandon third-party optimization services and return to Google's intelligent strategies. In our case, we tested “Maximize Conversions” and “Target Cost per Conversion.” According to our calculations, “Target Cost per Conversion” works slightly better in the long run.

However, there is an important nuance to this strategy: for it to work correctly, you need to have a well-structured campaign. If you have internal data on the average check depending, for example, on the flight destination, you can divide campaigns into categories: some destinations bring in more revenue per booking, others less. For each group of campaigns, you can set your own CPA targets corresponding to profitability.

But here's the catch: when booking airline tickets, we don't know the exact average check at the time of purchase. Google Analytics records the “successful booking” event, but after that, there are often additional sales. These can include seat selection, insurance purchases, car rentals, or hotel reservations—all of these services are purchased after the main booking. When the data is transferred to Google Ads, we only see the fact of the booking, but we don't know the actual total revenue.

To solve this problem, we started using the offline conversion upload tool. Each transaction has a unique ID, and after it is completed, we can clarify the actual revenue. To do this, we export the data to Google Sheets or CSV, calculate the actual revenue for each transaction, and upload it back to Google Ads, changing the conversion value.

In theory, this allows us to switch to value bidding and optimize based on actual revenue rather than a conditional CPA.

We tested this strategy. From a technical point of view, everything worked correctly: revenues were uploaded, the data matched, and the reports showed the correct figures. But in practice, it turned out that Google Ads algorithms do not fully take into account adjustments to the value of offline conversions. It seems that the system “sees” the changes but does not use them to their full extent for training.

As a result, we returned to the old proven model — “Target cost per conversion.” At the same time, our campaigns remained divided by average check (above average, average, below average), and we set different CPA targets for each category. This structure proved to be the most stable.

The query “cheap flights” remained a separate story. We divided it into as much detail as possible: within the campaigns, we created separate groups for the exact match of the word and for all its variations and typos. We set our own target bids for each group. We monitored the average position, number of impressions, and CTR on a daily basis. If we noticed any anomalies and saw that competitors were completely knocking us out of the search results, we temporarily changed our strategy to “Target position in search,” regaining our leading positions.

Overall, it was this tactic — constant monitoring and flexible management of bidding strategies — that allowed us to maintain stable performance even on the most competitive keywords.

Key points

In conclusion, the online travel agency market is highly competitive. Moreover, the main competitor is not other agencies, but Google itself. Google Flights is effectively becoming a monopoly, and it is becoming increasingly difficult to compete with it.

To be competitive, you need to:

1. Fine-tune your campaigns and their structure as much as possible.
2. Expand your list of keywords — don't limit yourself to popular destinations.
3. Constantly check new airport and destination pairs.

Even if you already have stable data for certain routes, no one knows which destinations will “take off” in 3-4 months. It is better to keep all words that are not marked as “low demand” active.

It is also useful to reload and test keywords once a quarter to track changes in demand. Destinations that were unpopular yesterday may be at the top today, and it is on them that you can collect cheap traffic and get additional bookings.