To understand your visitors, you look at four things together: what they do on your site, what they search for, what they say and what they buy. Analytics reports and heatmaps show where they get stuck, and on-site search logs show what they want. Exit surveys, product reviews and chat logs explain why; order history reveals who comes back and who is never seen again. If you start where two sources point to the same problem instead of deciding on the strength of a single report, you are less likely to fix the wrong thing.
When a question comes up, the criteria below will help you decide which report to open first and whether a heatmap, a survey or a purchase report will answer it. Four examples show which change on the product page or in the cart a finding should lead to, and sample messages show how to let it shape what your messages to customers say and who receives them. If you are thinking of replacing a recording tool such as Hotjar, you will also know which criteria to test the new tool against before you switch.
Read behavior from four sources
Even in a small store, part of the data you need to read customer behavior in e-commerce is already in your hands: order history, product reviews and analytics records. Sources you do not have yet, such as an exit survey, need to be set up separately. What is usually missing is putting these sources side by side and knowing which question to ask of each one.
What they do on the site: the product and checkout funnel
GA4 measures a visitor’s path with events. In Google’s e-commerce measurement guide, the main events you will use for the funnel come in this order: viewing item details (view_item), adding to cart (add_to_cart), beginning checkout (begin_checkout), submitting shipping and payment information (add_shipping_info, add_payment_info) and purchasing (purchase). If your e-commerce platform sends these events, look at the share of visitors who move on from each step to the next and at how many people you lose at each step. You do not need to calculate the rates by hand. According to the Google Analytics Help Center, when you add these steps to the “Funnel exploration” template in the Explore section, you see the result of every step on a single screen. Choose “Device category” as the breakdown dimension, and each step is listed separately for desktop and mobile.
In our article on increasing e-commerce sales, we showed how to calculate the rates between steps with sample figures and which step to start with. When you read behavior, go one step beyond the funnel: once you find the step where the losses pile up, open the search, survey and review data for that page. Do not weigh a loss by headcount alone. Some of the people who leave a product page only came to browse, whereas those who start checkout and then give up are the audience closest to buying. Even if there are only a few of them, look for causes at that step, such as a shipping fee they are seeing for the first time, mandatory account creation or a payment error.
What they search for: on-site search logs
The search box is where visitors describe what they need in their own words rather than in your category names. If the “Site search” option in GA4’s “Enhanced measurement” settings is turned on, a view_search_results event is recorded every time a search results page opens, according to the Google Analytics help page; GA4 recognizes the search page by a parameter in the address bar. The parameters it recognizes by default are q, s, search, query and keyword. To see which one your store uses, run a search on your site and look at the address bar: the term immediately to the left of the word you searched for, just before the “=” sign, is the parameter. Any other parameter has to be added in the settings; otherwise the search report will look empty.
Look for three things in the search logs. The first is searches that return no results: if visitors type “keten gömlek” (Turkish for “linen shirt”) but your product names say “linen”, you do carry the product; it just does not come up in search. The second is visitors who leave the site right after searching: the results list may not be meeting their expectations. The third is frequent searches for a product you do not have in stock, which is a direct demand signal for your sourcing decisions. GA4 does not record on its own whether a search returned any results; if your e-commerce platform does not report zero-result searches, check by searching your site for the most frequent terms yourself.
If you find that visitors who use search buy more often, do not take it as proof that search brings in sales; visitors who intend to buy may simply search more.
What they say: exit surveys, reviews and chats
Instead of guessing why visitors leave, you can ask them. Pushouse’s exit-intent popup module puts a short survey in front of a visitor who is getting ready to leave the page: did they find the information they were looking for, and was their decision swayed by price, by delivery or by something else? We covered which pages to show the survey on and how to design it in our article on e-commerce exit surveys.
When you read low-rated product reviews, file each one under a topic: size, fabric, packaging, delivery time, the product turning out different from its photo. If the same topic comes up in several separate reviews within a month, do not write it off as a coincidence; there is probably something to fix on the product page or on the sourcing side. Customers who would not write a review on the product page of their own accord also get a quick way to rate their purchase through a WhatsApp feedback flow that asks for a rating after delivery. They choose their score from ready-made options and can add a short comment if they like.
Questions that reach customer service are the same kind of data. If your site has a chat assistant, its chat logs reveal the information visitors could not find on the product page. Read the logs regularly; if “Does this jacket have a lining?” is asked dozens of times a week, the answer belongs in the product description.
What they buy: order history
On-site behavior also covers visitors who do not buy; order history, on the other hand, shows when customers come back, how much they spend and what they return. Grouping customers by last order date, number of orders and total spend (RFM) lets you divide your customer list into meaningful segments drawn from order records; we explained how to set it up in our customer segmentation article.
If you would rather not rebuild the groups by hand from an order export every month, you can describe the audience in natural language with Website AI Reporting in the Pushouse panel. A request such as “customers who bought a coat in the last 90 days but no boots” is turned into a rule applied to purchase history; you read the rule on screen and correct it if needed. For visitors who have never placed an order, trace their path through the site with funnel and survey data instead.
Heatmaps, surveys and purchase reports answer different questions
You can sort the tools that measure customer behavior into three groups by the question they answer; some tools fall into more than one group. Decide which one deserves your money or time by looking at what kind of question you have.
The analytics funnel shows at which step you lose visitors, while heatmap and session recording tools show where on the page you lose them; together they answer the “where” question. These tools show you visually where visitors click, how far down they scroll and at what point they leave. Microsoft Clarity presents itself as a heatmap and session recording tool. Hotjar, for its part, has become part of Contentsquare: Contentsquare’s Hotjar page says new Hotjar accounts can no longer be created, and it also states that Contentsquare offers heatmaps, session recordings and surveys on its own free plan.
Their weak points are clear too. Recordings show what a visitor did; why they did it is usually left to your interpretation. They give their clearest answers on technical problems: a button that does not respond on phones or a form that throws an error gives itself away within a few recordings. Watching hundreds of sessions also takes time, and if you do not start watching with a question in mind, you will spend hours and come away with observations that are interesting but useless.
Survey and feedback tools answer the “why” question. An exit survey, a post-delivery rating or a single question on the order confirmation page lets visitors explain their decision in their own words. Their limitation is that you hear only from the people who respond, and a survey with poorly written options does no more than confirm your own guess.
Reports based on purchase data answer the “who” question: who buys again, who buys once and disappears, who returns items often. They cannot see how people move around the site, but they give you the audience for your messages directly.
To make the choice easier, write your question down in a single sentence. “Does anyone notice the shipping information under the price on the product page?” is a heatmap question. “Why do people add items to the cart and then not buy?” is a survey question. “How many of the customers who bought a coat in winter came back in spring?” is a purchase report question.
When choosing a Hotjar alternative, test the candidates against the same criteria
Stores searching for a “Hotjar alternative” may be after one of two different things. The first group wants another tool of the same type, one that offers heatmaps and session recordings: some want to leave Hotjar over the price, a traffic limit or the interface, while others are looking elsewhere because they cannot create a Hotjar account. The second group is really looking for the answer to a different question: they have watched the recordings and seen where visitors leave, but they still do not know why. If you are in the second group, a tool that asks visitors directly may give you an answer faster than a new recording tool would.
If you are in the first group, Microsoft Clarity’s homepage says the tool is free forever and puts no limit on traffic, so you could try it first among the candidates. If you are in the second group and the tool you already use has a survey feature, switch that on before moving to a new tool. Whichever group you are in, compare candidate tools on these criteria:
- Does it work just as accurately for mobile visitors? Check your traffic’s device split in your analytics report; if a large share comes from phones, try the tool on a phone first.
- Can it mask form fields such as card numbers, passwords, phone numbers and addresses in recordings?
- Does it work with your cookie consent, so that it does not start measuring until the visitor gives consent?
- In which country does it store recordings, and where is the company behind the tool based?
- Can you write the survey wording in Turkish and in your store’s own voice?
- Can you turn what you find into action, for example by filtering responses by device and page and exporting them, or by sharing the results with your team?
A hypothetical example: a furniture store with a high exit rate on its product pages watches session recordings for a few days and sees that visitors spend a long time zooming in on the sofa photos. The recordings do not say whether they are curious about the fabric, the dimensions or the color. When the store spends two weeks asking “What stopped you from buying?” on the same page, with dimensions, color, price, delivery, “I was just looking” and “Other” as the answer options, it turns out that most respondents could not be sure the sofa would fit in their living room. Instead of reshooting the photos, the store adds a drawing with the sofa’s dimensions to the product page and, two weeks later, compares the product-page-to-cart rate with the previous two weeks.
Consent and privacy limits when collecting data
Most tools that measure visitor behavior use cookies. Whether a cookie needs the visitor’s explicit consent is assessed separately for each cookie, according to its purpose, in the Guide on Cookie Practices (Çerez Uygulamaları Hakkında Rehber) published by KVKK, Türkiye’s Personal Data Protection Authority; we covered the consent and notice steps in our article on cookie policies.
Decide whether the heatmap, session recording or survey tool you use requires explicit consent on the basis of the guide and your legal adviser’s opinion. Tools that require consent should not run until the visitor gives it; the guide does not treat simply entering a site as consent to cookies, and it considers it important that these cookies are switched off when the preference panel first appears. Even with a tool that does not require consent, you need to inform visitors no later than the moment the data is collected; if the company providing the tool is based abroad, the guide also requires the transfer made through cookies to comply with Article 9 of Law No. 6698 on the Protection of Personal Data and with the regulation on transferring personal data abroad.
Test your setup yourself. Browse a few pages of your site twice: once without making any choice in the cookie banner, and once after rejecting cookies. Use your browser’s developer tools to check that services requiring consent do not set cookies, and your measurement tool’s dashboard to check that these visits were not recorded. To prepare the banner’s explanations and preference controls based on the measurement tools you use on your site, take a look at our cookie management page. Also verify that form fields are masked in session recordings by placing a test order on your site and watching the recording; afterward, cancel the test order and filter it out of your reports.
Turning a finding into a decision: four examples
Follow two rules when you turn a finding into a change. First, wait where possible until two sources show the same problem; if the funnel shows losses on the product page and the survey on that page also puts delivery questions at the top, you usually have enough evidence to act. Second, write down what you will measure before you make the change, and compare the result with the preceding period of the same length, such as the previous two weeks; comparing a campaign week with a quiet week will mislead you. Before comparing, take your own team’s visits and test orders out of the figures; in a store with a few dozen visitors a day, even two test orders shift the conversion rate.
Searches return no results
Your list of zero-result searches may contain several words for the same need (“evening gown”, “formal dress”, “prom dress”). Work them naturally into your product titles and descriptions, and if your e-commerce platform has a synonym setting, add them there too. If you see a product group that people search for often but you do not sell, share it with your sourcing team. A month later, look again at whether the same words return results and at the search-to-cart rate.
Delivery is the main question on the product page
Suppose “I couldn’t find the delivery time” is the top answer in your exit survey. First check where that information sits on the page. If it is at the bottom, a heatmap shows whether visitors scroll down that far; if they do not, the information might as well not be there. Move the estimated delivery date and the shipping fee directly below the price, then compare the product-page-to-cart rate for the two weeks after the change with the two weeks before it.
They leave when they see an extra charge in the cart
When the funnel shows losses between the cart and checkout and the survey says “Shipping costs more than I expected”, handing out a discount code may be the first solution that comes to mind. Try cheaper options first: show the shipping fee or the free-shipping threshold on the product page, and tell shoppers in the cart how far they are from that threshold. We worked through the numbers on when a code still leaves you a profit in our article on discount codes for customers who abandon their cart.
They do not come back after the first order
Order history may show that customers who shop once and then disappear are concentrated in a particular category. In that case, look at post-delivery feedback for that category specifically. If the ratings are low, look at the product or the delivery first; a promotional message will not solve that problem. If the ratings are good and customers still do not return, take the category’s natural purchase cycle into account. Someone who buys a duvet or a winter coat is not expected to shop again within a few months; the product may also meet a one-off need. When the cycle has passed and they still have not returned, one possible reason is that your store has slipped their mind; to test that possibility, you can send a short message about a relevant new product to customers who bought from that category and have opted in to your messages.
Tie customer-centric marketing to behavioral data
Customer-centric marketing means deciding what to tell whom, when and in which words based on what customers do and say rather than on what the store wants to sell. Once you tie that definition to the sources above, three concrete habits follow.
Build your message content from your customers’ own words. If “Does it run small?” comes up often in surveys and chat logs, write “runs small, we recommend ordering one size up” in your campaign text instead of “slim fit”. If “gift wrapping” is a frequent term in your search logs, put that information in the first sentence of your special-occasion campaign.
Choose the audience by behavior. If you are going to run a WhatsApp marketing campaign for one category, send the message first to people who have already shopped in that category, and keep frequent returners out of discount campaigns. For a detailed look at which groups to separate and how the wording changes for each one, see our article on dividing customers into groups before a campaign.
Show customers that what they told you made a difference. If you changed something based on feedback, announcing it to the customers who complained or asked about that issue can build trust, and it gives them a reason to answer your next survey too. Here is a sample announcement for a customer who asked about the fit and has opted in to your messages:
Hello {{customer name}}, last month you asked us about the fit of our products. You can now find the model’s measurements and a size chart on every product page. Take a look at the new season’s {{category name}} products here: {{category link}}
A customer who gave a low rating in your post-delivery survey, on the other hand, should get a short reply aimed at solving the problem instead of a campaign message:
Hello {{customer name}}, we saw the rating you gave for order {{order number}}. We would like to sort out the problem you had; just reply to this message whenever it suits you, and our team will get back to you.
Decide in advance who will get back to customers who give a low rating, and within how many days; if the reply is delayed for days, the customer may feel their problem has been put on hold. After each low rating, note in the same list whether the problem came from the product, the shipping or the communication; after a few months, that list becomes a source you can use to test the reasons you have read in reviews and surveys. If you would like to see how an exit survey, or customer groups drawn from your order history, would work in your store, you are welcome to write to our team.
