The project was configured with very deep analytics. Throughout the entire period of cooperation, we did split testing, built funnels both for orders through the website and for orders by phone. We constantly improved and refined some details in order to increase the conversion. We paid a lot of attention to SEO.
You need to select the suppliers. The supplier must ensure he has a product in stock. The supplier must have a responsible person — the one who will be responsible to the platform for delivery times, etc. (if we list for example a TV on the website, then if someone buys it — we know exactly who to contact, who will pick up the phone, send an invoice, the TV itself, etc.).
Those suppliers must be aggregated into one price list (different suppliers can have the same TV). With the help of a special parser, collect together the articles of all positions and prices from different suppliers, compare them with the average market price, make calculations and make the most advantageous offer for both parties. Then “enrich” each position in the catalog with characteristics and pictures.
After the client chose the desired TV and clicked the “Buy now for a $1,000” button, the leasing company (which came up with this whole project) goes to the supplier and buys this TV from him for $900, returns to the client to get the agreed $1000 to, accordingly, earn itself $100.
Conducted a scoring. After filling out the questionnaire, clients were contacted from the call center, the company received additional information, and checked them in the database, in which all information from the tax office, banks, etc. was aggregated. We calculated your credit history and made a decision: to allow the purchase or refuse. After that, all further installment purchases took place according to a simplified system.
In your personal account, you could see all the necessary information on purchases and payment balances.
First, with scoring according to the principle described above. Secondly, with a parser of prices and characteristics of products from suppliers. Thirdly, minimally with a recommendation service and a service for exchanging product reviews.
Everything happened gradually. Only in those cases when the client understood which process “hurts”. For example, first we structured the client statuses so that no orders were lost or forgotten. Then we set up integration with suppliers to receive all the information from them automatically and not “manually”.
In the case with the same TV, while you arrange an installment plan with the call center operator, this unit must be reserved from the supplier so that he does not sell it to anyone else.
Also we daily asked suppliers for an invoice for products purchased by users during the day (and there were 300 such items per day). With those who had integration on the website, everything was simple — they received a notification and could automatically send us the necessary documents. And for those who did not have integration, we generated a message that was later sent to their messenger, where we also indicated all the articles and the quantity.
TV costs $1000. How can I buy it in installments:
It tracked in which case, what type of installment to offer. For example, if the markup is 20%, then this is 15 payments. Supplier, brand, margin — could also influence the type of installment plan.
Products with slightly better characteristics, but more expensive. This idea is not ours, it is taken from Amazon. But the conversion she did was unreal. At least 10% of users clicked and bought from this block.
Products that are bought along with this. We did not integrate with any service. We came up with our own constructor, which made it possible to recommend the products to consumers in a specific way to reduce the burden on logistics. Means that, if you are looking for a music center and a TV and have already chosen the first product, the second one will be selected for you either from the same supplier or from another, with similar delivery conditions. So that you do not wait for your purchases on different days.
Search on the website was able to do banal things, such as recognizing an unswitched keyboard layout or single-root words. But besides this, it could recognize products not only by the name of the category, but also by the brand, and in the future - by the characteristics. That is, you could simply enter “Samsung TV with Wi-Fi” and get all the current models.
Let's make a reservation right away that a menu item is not a category. Thanks to the constructor, we could add to the menu, for example, the item “Bath products”. Washing machines, dryers, etc. Or we could create “Samsung TVs”. All this helped a lot in terms of SEO. There were auto-generated headers and a templating engine that wrote meta- for products.
At the stage of purchase, the number of fields to be filled in was reduced to a minimum: a phone number for authorization, a delivery address, and a promo code. But after authorization via SMS, it showed a form that the buyer had to fill out with the call center operator, and offered to enter the data online in order to spend less time on the phone.
Freezing the price during the day for the product added to the cart.
Thanks to deep analytics, we saw that at authorization stage, many users entered their home phone number where sms couldn’t be sent, and they simply called back. With the help of hints, we explained that they needed a mobile phone, and the conversion increased.
Logistics. We divided the whole country into zones so when a person indicated a settlement and an address, we could immediately show him free time slots, because we knew that a car would be in this zone during this period.
We had a formula, according to which we counted all the products every evening, so that the next day a person could look at products with a “by popularity” filter.
For each category, you could select and configure your own filters. After all, for laptops this is, for example, diagonal, weight, type of processor. And for washing machines: loading, maximum number of revolutions, etc. And, by the way, we provided that some filters are bullets, others are a range, and still others are a list. Separately, it was possible to manage sorting in the filter constructor. For example, brands by number of products, and processors by alphabet.
Collections of products for the holidays. “Ideas of products for Christmas”. It was brought to the main page and decorated to distinguish it from others.
After the purchase was made, positions were offered much cheaper, but according to various parameters suitable for the order.
Personal recommendations work somewhere on the basis of the RetailRocket service, somewhere on the basis of rules and complex constructors, and somewhere on the basis of automatic proprietary algorithms.
