WordPress AI eshop Optimizer Plugin

Display Preferences

In this section, you control the display of recommendations on the Product page of your WooCommerce site.
Specifically, you can control the display of:

• Up-sells
• Cross-sells
• Related by Woocommerce
• Viewed Recently

A note about Up-sells and Cross-sells: For AI Optimizer, the key difference between these two lies in the “level of confidence” of the recommendations. We believe the terms “Up-Sell” vs “Cross-Sell” are more apt for brick-and-mortar retail where the salesperson can “see” the customers and determine if they can afford a pricier item (i.e., UpSell) or something similar/complementary (i.e., Cross-sell). This distinction might be less relevant online.

In our default configuration:

• UpSells are up to three (3) items that are “extremely highly likely” to be excellent matches for the particular product.
• CrossSells are up to seven (7) items that are ‘highly likely” to be excellent matches for the particular product.

It’s crucial to understand the “up to” in the above definitions. There might be instances when our algorithms can’t provide recommendations that meet our stringent confidence standards. In such situations, the algorithm will either:

• Experiment by suggesting products that have never been sold on your e-shop – or at least, products not reflected in the dataset of past orders you’ve uploaded.
• Opt against suggesting what it deems irrelevant “recommendations.” This is because we understand that people’s attention is invaluable, and our reputation hinges on the pertinence of our recommendations.

Our vision is long-term, and we see ourselves as partners in your journey. You’ll find that with AI Optimizer, your “never sold/number of products in inventory” will diminish significantly over time. In our experimental case studies, an e-shop with 12,000 different items sold only 3,500 items before implementing the AI Optimizer. A year later, this figure had doubled.

Many of our partners initially hesitate to leave their recommendations entirely to AI. They often choose to merge our suggestions with their own. In these cases, we categorize their suggestions under the “Up-Sells” label, while all of our recommendations (ranked by confidence) are labeled “Cross-sells.” Additionally, we provide an option to swap the display order of “Up-sells” and “Cross-sells.”

The charts below offer a gauge of our algorithm’s efficiency when contrasted with the largely arbitrary WooCommerce recommendations:

Privacy First

Our optimization algorithm may carry out semantic/textual analysis on the titles of your products or product variations, and could also consider their pricing when suggesting up-sells and cross-sells.

However, you have the option to exclude product titles and prices from this analysis. Despite these omissions, our robust algorithm can still offer highly relevant recommendations!

Your choices include:

• Include product categories/tags text.
• Incorporate the full text for sentiment analysis.
• Factor in product price data.
• Base calculations on parent products (excluding variations).

Consider Repeat Registered Customers

We track whether a purchase has been made by a guest or a registered user.

We also offer the option to adjust based on the preferences of customers who’ve chosen to register on your e-shop.

You can choose not to provide the user ID, but we’ll still distinguish between guest and registered user purchases.

Account for Seasonality

You have the flexibility to adjust for seasonality by defining an optimal timeframe centered around the current date.

The pro version offers season-specific recommendations based on your selected duration, granted there are enough data points. Furthermore, we aim to provide consistent, automatic updates in the future.

Operational Efficiency Settings

By leveraging Stored Procedures, we can achieve an export rate of approximately 100,000 order items per minute. Ideally, all the boxes should be checked for optimal functionality. However, you have the autonomy to adjust settings; for instance, you might choose to initially process your orders before running aggregate statistics or forgo the more resource-intensive step of rebuilding the product table (which handles about 10,000 variable products per minute).

Your available options are:

• Regenerate the complete product catalog.
• Convert and adapt my orders to the required format.
• Incorporate products that haven’t sold.
• Add aggregate statistics for products.
• Integrate aggregate statistics for customers.
• Include advanced aggregate stats for customer behaviors.
• Generate UUIDs locally for GraphDB nodes and edges.
• Produce the export table or CSV.

Specify Your Dataset’s Starting Point

This allows you to designate a starting point for your order export file, letting you bypass initial or demo orders.

Your options are:

• Set the initial Order ID
• Maximum Number of Records

Upload your order details to our AI Graph DB service

Link to uploading your orders

Download your existing up-sells and cross-sells

Backup your existing up-sells and cross-sells: Secure your current recommendations just in case you wish to revert to them in the unlikely event that our system’s suggestions don’t meet your expectations.

Test Order Data Export on Your Server

If, in the rare circumstance, the uploading of your order data file to our server fails, consider testing the loading of your export file locally. This will help you ascertain if the file is corrupted, contains prohibited characters, or has other issues.