Case Study: How to Get +20% Email Revenue Every Month with Product Recommendations

Company

Antoshka is one of the biggest Ukrainian retailers of children's goods. Here you can choose, compare and buy everything that a child may need: items for newborns and babies, trending toys, clothes and shoes.

Tasks

The main purpose of cooperation with our platform was to increase sales with automated personalized campaigns.

So, we need to set up triggered emails with personalized recommendations based on user behavior on the site:

• Abandoned carts;
• Abandoned views;
• Price drop for previously viewed products or for products in the cart;
• Reactivation;
• New items.

Solution

Antoshka's marketers and developers from our platform have jointly developed several algorithms for product recommendations that can be used not only in emails but also on any website page. As a result, the user sees cards of those products that he’s most likely to buy.

Below is a brief description of these algorithms.

Recommendations algorithms

1. You will also like

The neural network predicts the most likely click based on the users' browsing history, taking into account the price range of the viewed items.

2. Discounted items

The principle is the same, but only discounted products are selected for display.

New users without browsing history doesn't will see discounted bestsellers in their recommendations.

3. Items from the same category

The neural network analyzes user’s views and offers products from the same category and price range. It is also possible to recommend only new items from the corresponding category.

4. Cross-selling

After the customer makes a purchase, they are offered accessories and related items.

5. Recommendations for out-of-stock products

If the items that the user was interested in are out of stock, they’ll see similar products from the same categories.

Triggered campaigns

Recommendations in all Antoshka's triggered emails are based on the age and gender of the child.

Сampaign priority and the number of triggered emails for a contact (per day and per week) are determined by segment conditions, workflow conditions, and Annoyance level.

1. Abandoned cart

This triggered email contains up to 3 items added to the cart and up to 3 recommended items.

It sends no earlier than an hour after the user has added a product to the cart if they didn't make a purchase.

This trigger has the highest priority of the behavioral triggers, as it is the most converting one. Abandoned cart email generates about 50% of the income of all the behavioral triggers although it accounts for only 14% of all delivered personalized triggers.

2. Abandoned browse

This is the second highest priority triggered email. It’s sent no earlier than an hour after the user has viewed the items, and nothing has been purchased or added to the cart.

The email contains up to 3 viewed and up to 3 recommended products.

It is the most frequently delivered trigger, accounting for 33% of the total delivered behavioral triggers. It ranks second in conversion to purchase.

3. Price reduction

This group of behavioral triggers includes the Price drop for items in the cart, Price drop for viewed items, and Price drop for similar items triggers.

The first two emails contain only products that were previously added to the cart or viewed and their cost has decreased by more than 5%.

The triggered email Price drop for similar products contains up to 6 recommended items with reduced prices.

All these messages are sent if within 7 days after viewing or adding to the cart the user still hasn’t bought anything.

4. New items

This email is sent to those who viewed products but didn't buy anything. It recommends new items from viewed categories. The message is sent once a week or less and contains up to 6 recommendations for new products with no more than 3 from the same category.

In terms of conversion to purchase, this trigger is about the same level as price drop triggers. As a rule, there are no discounts for new items.

5. Reactivation

An email with personalized recommendations is sent once every 3 months if the user hasn’t visited the site for 30 days.

Email contains up to 6 viewed but not purchased items and up to 6 recommended items but no more than 2 from the same category.

This trigger is the least converting of all.

Personalized recommendation algorithms are also used on Antoshka’s website:

• in the Great deals block on the main page,
• in the Similar items block on product pages.

Similar items generate more than 80% of the website recommendations revenue.

Special Request Inline

Results

1. Due to the implementation of behavioral triggers, the monthly income from the email channel has grown by an average of 20%.
2. All performance indicators of personalized emails are higher than bulk ones: open rate and CTOR are more than 2 times higher, the transaction rate is higher up to 70%.
3. Dynamic content accounts for almost 86% of all conversions from triggered emails.
4. Recommendations in the Great deals and Similar items blocks increase the time users spent on the site and the number of pages they viewed. They also help potential buyers to make a choice in the category of goods they are interested in.

Next Steps

1. To increase the share of personalized emails due to the implementation of new behavioral triggers Next purchase and Regular demand.
2. To improve the effectiveness of existing triggered emails through A/B testing.
3. To add personalized content to bulk emails.