Another issue is that all major problems were solved a long time ago by the team of professionals, no reason to mention that it's a very developed product. That is why I was trying to find minor problems, that can bring more effect in the future. As I mentioned, data sampling meant that I need to work with Excel a lot. Creating a segment with screen resolutions and some specific words in page links brought me to ability to see only two day range at a time. That meant that if I want to look at one month period, I have to calculate bounce rate or other indicator manually from 15 different requests to GA.

So, walking through numbers I found a couple of interesting things.

I have been comparing items in filter of sightseeing. In result I found that number of page views from pages with filter near airport and train station is very high in cohort range with another items of sightseeing. After I was think, that train station is present in every cities of Ukraine and I can compare pages with cities and active filter "near train station" from cohort. And I saw, that in the chart a large segment none tourism cities. And I saw that on the chart are large segments of cities, which usually we can't name the cities of tourism. These are industrial cities, such as Kryvyi Rih, Zaporizhia, Nikolaev and others.

I looked through the user flow by their ID. And I noticed that a large segment of these users targeted cards of hostels and hotels. I thought that it's can be a people in business trip. Because the people in business trip need a fixed bill about paying the cost for the trip. And trains usually arrive very early in the morning or late in the evening, when public transport don't work. Taxi services don't provide the fixed bill.

The simpler way to use this segment is to convert them to the hostel and hotel that provides a fixed expense.
I was decide to meet with users to find more insights of their experience. At in-depth interview I found insight pattern of users in business trip and has prioritized their on pains&gain by Kano model.


___________________

Kano Model is a tool team use to make design decisions. It enables to plan design better by prioritising features based on their expected impact on customer satisfaction. In consequence, it helps understand whether a given feature will bring delight or frustration.
The graph of the model presented below may be helpful when defining each of the listed categories. The y-axis is customer satisfaction, and the x-axis is feature implementation.
___________________

In small towns has a problem that there haven't places to have lunch or kitchen their is bad.

I was looking for the bounce rate and page views on the filter of price ranges: less than 250 UAH., From 250 to 350, from 350 to 500, from 500 to 1000, from 1000 to 1500 and more than 1500 UAH. And I saw it in major tourist cities, such as Odessa, Lviv, Kiev, at prices hovering less than 350 UAH. - The bounce rate is very high.
Later, I compared this data with less well-known tourist cities by season. And I found a pattern that the quality of the apartments on a cheap time interval varies with the seasons and geo.

And the bounce rate also changes, because when users search for a cheap apartment in a famous city in the hot season, they will get a bad and small search result. And the cheap range is different for cities such as Odessa and Ternopil. In the next photo you can see that in Odessa at a price of less than 250 UAH you will find only a bed in a hostel, but in Ternopil at a price of less than 250 UAH. You will find a private apartment with a comfortable bed.

So, how we can help users to find apartment in different cities faster and better? We can show him how much proposition he will get on the different price range in the city. We will show him what is cheap and what is expensive for this city at this season.