How to analyze product sales in an online store

1. Establish what each column counts.

The Flowpeek product report combines observed views, add-to-carts, purchases and revenue for recognized products. Purchases means distinct transactions that contain the product. If one customer buys three units in a single order, that doesn't automatically count as three purchases in this column. Views, on the other hand, count events, not just distinct people.

CR in this report relates a product's purchases to its observed views. It's not the same as the share of all store visitors who placed an order. Other tools may use different names and denominators: the GA4 documentation distinguishes between event-scoped and item-scoped metrics. Before comparing two tables, read the definitions instead of assuming that a similar label guarantees they match.

Read one row from left to right. The navy-and-beige shirt has 49 views, 28 unique users, 1 purchase and a CR of 2.04%: one purchase divided by 49 views. Dividing by 28 people would give a different metric. The navy-and-cream shirt, in turn, has 28 views, 5 add-to-carts and 0 recorded purchases. It's a good candidate for a closer look at the cart, but not yet an explanation of the abandonment.

2. Check the period, currency and catalog matching.

Pick a completed period with enough observations for your store. Too short a range can favor a product with a single purchase. Too long a range mixes seasons, promotions and varying availability. Start with a few weeks, and when you compare results, note any changes in price, placement and stock.

Read the completeness information. A missing explicit product list in an event, or no unambiguous match to the catalog, limits the report. Flowpeek doesn't guess a product from a similar name. A missing amount or currency stays missing data, not zero. To sort by revenue, pick one currency; judge the brand ranking the same way. The currency filter narrows purchases and revenue, but views and add-to-carts still include observations from all currencies in the selected period. Don't read that CR as the conversion rate of a separate market. PLN and EUR aren't automatically converted to a single value.

3. Change the question, then the sorting.

Ranking by revenue shows the products that build sales value. Ranking by purchases shows those that appear in many transactions. Ranking by views reveals interest that may not yet translate into results. None of these views alone describes a product's full role.

Go through the top few items in each list, then use category, brand or the available metric values to narrow the topic. When looking for product pages to improve, start with products that get noticeable traffic. A product with two views and zero purchases usually gives too little information to judge. Also, don't extend conclusions about one popular category to your whole range.

4. Recognize four typical situations.

Lots of views and few add-to-carts point to the offer: whether the ad matches the product page, the price, description, variants and availability. Check the page on the customer's device and a few recorded journeys. Don't assume straight away that price is the problem; an important variant may just as well be unavailable, or the description may not answer a key question.

Lots of add-to-carts and few purchases suggest checking checkout, shipping costs and whether the product is bought later. Few views with recurring purchases may justify testing more exposure, once you've checked stock. High revenue from a handful of transactions calls for caution: the result may hinge on a few expensive orders. These observations help you choose the next step; they don't yet prove the cause.

5. Example: a popular product page isn't a bestseller yet.

Say product A has 2,000 views, 100 add-to-carts and 20 transactions. Product B has 500 views, 60 add-to-carts and 15 transactions. Purchases relative to views come to 1% and 3% respectively. These are sample numbers to calculate with, not results from any Flowpeek customer.

It doesn't automatically follow that you should move all traffic to B. The products may differ in price, purpose, availability and the source of interest. First check whether A lives up to the promotion's promise and whether B can handle a larger scale. For A, plan a review of the product page and journeys; for B, consider a limited exposure test. Both actions should have a recorded baseline and a review date.

6. Check a few transactions and don't add up people across products.

One order can contain several products, and one person can view many product pages. That's why the sum of purchases from the product rows isn't automatically the store's number of unique orders. Likewise, adding up the unique users of each product doesn't give the number of distinct people in the whole store. Overlapping groups require a different aggregation.

If revenue looks unexpected, pick a specific product and the orders from that period. Check the quantity, line value, currency and the status of the source purchase. The total order amount may include other products or shipping. When comparing with GA4, remember that item revenue has its own definition. If you need further analysis, export the report to CSV; keep the separate currency rows and the notes about missing data.

7. Make one change and come back to the same metrics.

Pick a product with enough observation volume and a clear next step. Write down a hypothesis: "Customers don't see the compatibility information before the buy button." Fix that element instead of changing the price, photos, ad and shipping all at once. Come back to evaluate it over a comparable period.

Alongside CR, write down views, purchases and traffic context. With a larger influx of new visitors, the percentage may drop even as the number of purchases grows. The result after a change isn't proof of its impact on its own; stronger conclusions require a planned comparison. The product report helps you choose a sensible test and follow how it goes, while cost of goods and profitability are something you check in the company's financial data.