Use case #4: Occupancy rate impact on conversion

CategoryPerformance |Solution: DataMa Compare | Type : Ad hoc | Client: Web travel | Extension: None
Tags: #Occupancy #Stock #Conversion

 

Context

Client is a major hospitality player. 

Web product team has an ambitious target of conversion uplift this year vs. previous year. However, we’re in the middle of the year and this target seems hard to reach. While traffic and average booking value are going up, the drop in conversion is suspected to be related to an issue of availability of hotels

Question is to prove and estimate the impact of occupancy shortage on conversion rate

Approach

Market equation

While first thing would be to check the effect of traffic volume and average booking value on conversion using a classic market equation (see default demo of DataMa), we will focus for now on conversion

The idea is to split the conversion rate into pieces so that we make the occupancy rate factor appear. However, the occupancy rate is generally an offline concept, not directly related to online conversion.

The complexity in occupancy rate or stock analysis in travel is that the occupancy of a given day (the number of rooms available divided by the total number of rooms) is different depending of the booking window, or lead time, ie. the distance between the day of the search and the day of the booking.

Luckily enough, the client has a Custom event in its web analytics tool (Google Analytics) that gives for each hotel page viewed the availability of that hotel (Available/ Partially Available/ Not available)

This market equation comes as follow, and we have in fact two options to treat our problem from a dataset stand point

Dataset

In that use case, the data comes from Google Analytics, stored in Google BigQuery. 

From there, we can create the SQL logic to classify each session by availability, at session level, so that we don’t duplicate the data.
There are essentially 4 types of availability status for a session:
1/ Not reached = Don’t reach a Hotel Page
2/ Not available = Reach only unavailable hotel pages
3/ Partially available = Reach a partially available hotel page, but not fully available
4/ Available = Reach at least one available hotel page

As we also want to be able to breakdown results and impact by market and device, we will also add those two dimensions. And of course, since we want understand a drop between two periods, we grab some kind of date dimension.

Once done, we have two options to perform our analysis in DataMa

Option 1: Keep the data as is and visualize impact of availability as a mix effect of availability status on Conversion – see dataset

Option 2: Unpivot data to transform dimension to metrics by getting one metric for session reaching hotel page, and then one other for sessions reaching an available hotel page – see dataset

We ingest the data using DataMa Prep connector with BigQuery (and unpivot for option 2 can be done there)

Takeaways

Straight in DataMa Compare, the analysis allows to quantify the impact of availability on conversion.

Option 1 dataset shows a clear mix effect on status availability.  The share of sessions reaching only unavailable hotel page results has increased significantly, and they obviously convert really badly, so this has a clear impact on conversion. The advantage of this view is that you can easily communicate on the evolution of each type of sessions, including partially available ones, which is a bit of a grey zone in terms of impact

1:The DataMa Compare Waterfall opened on conversion showing a clear mix effect on Availability dimension

2: In the Moves slide, we can see available hotels decreasing in the mix and in performance

Option 2 dataset allows to visualize the impact through the share of sessions reaching at least 1 fully available hotel page vs. the number of sessions seeing a hotel page. The advantage of this view is that you can then easily split availability the impact by another dimension, like market or device

3: Waterfall having a first step in market equation on Available sessions

4 With this step you can easily split by other dimension, like market here

We both cases, we see that more than half of the drop in conversion comes from availability issue

Outcomes

The learnings have been shared at C-level and country level, has allowed to review the target of conversion to take into account the limitation of occupancy and has initiated a large thinking on increasing availability and offer by reducing the stock dedicated to OTAs and building new hotels in most searched destinations