Everyone at the company loves the developer's website. What about buyers?

The website of a large international developer has plenty of caring relatives. The brand manager wants a beautiful video, sales need a prominent button, PR has its own goals, HR has its own, and the owner's wife wants a GIF of a kitten on a pink background. Each of them is right in their own way. Only the apartment buyer comes to the website with their own wish list, and in our meetings there is usually nobody to read it out.

When I worked at a large Kazakhstani construction company with international projects, more than eight teams worked on the website. Among them were CVM, brand managers, PR, HR and the sales department. The suggestions were very different: move a block, put the video higher, add a striking animation. All these tasks had to be prioritised.

We decided to ask users. Together with the contractor's team we launched a CSI survey. Over two and a half months, about 700 people took it. Their answers helped us revise our priorities and focus on what was stopping people from choosing an apartment.

Here is how it went. The figures and the status of improvements below refer to 2025.

Why we chose CSI

To give a sense of scale: at the time the website had about 500,000 users a month. The average deal was 30–40 million tenge, and a sale took about 60 days. The core audience was aged 25–50, and in our communications we focused especially on people aged 25–35.

The company operated in several foreign markets and worked in residential and industrial construction. At this scale, requests for the website naturally multiply. But sorting them out simply by gathering everyone interested in a meeting room becomes harder.

We already had NPS surveys. But they were taken by buyers who had completed a purchase. We wanted to hear from those who were still choosing too. A person may be annoyed by the apartment search, leave the website and never reach a post-purchase survey.

CSI, the Customer Satisfaction Index, suited this task. For each parameter we asked only two questions: how important a particular parameter is to the person and how satisfied they are with how we have implemented it.

Say a person cares a lot about loading speed but rates our website low on that point. That is a task. If design gets good ratings but matters less than speed, that is a reason to take a closer look at the next idea to add heavy animation.

With a single quality rating, it is harder to draw such a conclusion. You want to fix everything that scored low. A pair of ratings helps you understand what users care about more.

How we set up the survey

The quiz had 11 questions. Five pairs about importance and satisfaction: overall website quality, loading speed, design, information content and ease of buying online. The last question was open-ended: what could be improved?

The survey was shown in a pop-up. But first we let people settle in and really explore our website. To see the survey, a person had to visit the site more than twice, view more than three pages and spend more than 30 seconds on it. People who had already answered were not asked again.

If a visitor did not take the survey, the next invitation could appear 14 days later, once they met the display conditions again. It hardly makes sense to measure satisfaction while annoying people with the same window on every page.

This way we collected answers from people who had already got to know the website. There is a limitation to keep in mind, though: this mechanism does not reach visitors who leave right away. Studying them requires other tools. I would not draw conclusions about all visitors from these answers.

Responses were automatically sent to a spreadsheet via a webhook. It stored the ratings, with comments in a separate column. Another tab calculated the averages and the overall CSI.

We did not need to build a separate survey system for the launch. We used a ready-made service, and the contractor's team helped with the setup. You can set up such scenarios yourself; a partner's help is useful when you don't have the resources to learn the tool.

How we read the results

In our CSI calculation, each parameter's rating was weighted by its importance. The formula looked like this:

CSI = Σ(importance × satisfaction) / Σ(importance × 10) × 100%

The denominator is the highest possible score for the same importance of parameters. So the result shows how close you are to the maximum.

Our target was 80%. In different months the index ranged from 73% to 84%.

The case used this scale:

• 0–40%: very low satisfaction;

• 41–60%: low;

• 61–75%: average;

• 76–90%: good;

• 91–100%: very high.

These are guidelines of the scale we adopted. There is no need to turn them into a mandatory standard for every business.

The overall percentage itself should also be read with care. You can get a decent average and miss an inconvenient part of the website. That is why I would advise looking at the ratings of individual parameters and reading the comments. That is where it becomes clear that a person had to download a file, or why they got confused by the offers.

What we had to reconsider

There was an idea to move the video above the main content. Based on the survey results, we concluded that almost nobody watches the video and decided not to spend resources on that change.

A caveat is useful here: people's answers describe their perception and what they report about themselves. Actual video views should be checked with events in analytics. The survey helps you understand attitudes towards the video, while analytics shows how many times it was actually played and watched to the end.

Another idea was an element that was visually and content-wise impressive but heavy in terms of loading speed. Users showed that speed matters more to them than looks. We focused on making the website faster.

I think this is a useful outcome of the research. You can reconsider a task that looks attractive inside the company before development even starts. A website for an apartment buyer and a website shown on a big screen during a presentation are used in somewhat different conditions. A person waiting for a floor plan to load on their phone may simply not need our animation.

What we changed on the website

The comments gave us a very concrete list of improvements.

Search and filters. Users complained that it was hard to find the right option. We made the filters more flexible and convenient. At the time the case was published, work in this area was still ongoing.

3D unit selector. To view an apartment's details, people had to download a file. They lacked an understanding of which way the windows faced and where the apartment was in relation to the lift. For some properties we launched an MVP of a 3D unit selector showing apartments on the floor plan. It let you view the building, choose the entrance and floor, see available and reserved apartments and the apartment's position on the landing (how close it is to the lifts and stairs). The view from the window and changing the time of day to see the sun's position were still in the plans at that point.

City selection was supplemented with tiles and a map.

Promotions were personalised by location: after choosing a city, a person saw offers only for their region. We wanted to help people make sense of banners with different terms.

Neighbourhood. In response to questions about the future surroundings of the building, we added a section with development plans for the area. When buying an apartment, it is important to understand what will be beyond the beautiful courtyard.

Apartment comparison. Users no longer had to keep switching between pages. In the comparison you could leave only the differences: if the apartments are in the same residential complex, the repeated name and address are hidden, and the differing characteristics stay in view.

The result was a task list where the origin of every item was clear. In one place a person could not find an apartment, in another they could not figure out its location, in yet another they saw too many irrelevant promotions. With requests like these, you can go to designers and developers.

What I would advise those who want to repeat this

Start with the decision you intend to make based on the survey. In our case we needed to prioritise website improvements. We chose the parameters and set up the collection of answers for that task.

Keep an open question. Scores will show the problem area, and a comment can explain what exactly gets in the way. The phrase "it's inconvenient to choose an apartment" allows many interpretations. A remark about downloading a file already gives you a direction for work.

After making changes, it is worth repeating the measurement and checking behaviour in analytics. Satisfaction, usability and commercial results are related, but you cannot automatically equate them. In this case there is no data that would allow us to claim by what percentage the CSI survey increased sales. What there is: 700 responses and specific improvements made based on the feedback.

Keep collecting after the first wave. At the time the case was published we were still collecting feedback: the website evolves, new tasks appear, and visitors' expectations change.

I would start the discussion of every next improvement with a simple question: which buyer difficulty are we removing right now? The answer helps both to set the priority and to check the result later. Personal preferences about the button colour can wait a little. And if the owner's wife still wants a kitten, show her the research results.