Configure AI recommendations
AI is the capability of machines (particularly computer systems) to learn, reason and self-correct. The ability to acquire information and rules that can later be formulated into conclusions is an attempt at imitating human intelligence. This highly-advanced technology is becoming particularly useful in marketing. It enables the analysis of large portions of data and proposes solutions that make marketing activities more efficient due to the proposed solutions by this technology. For marketing, this means the implementation of algorithms on the website where the data is acquired and analyzed. Apart from page views, the module takes into consideration transaction data, including historical data, and product feeds. This way the system is able to put forward optimal strategies and adequate solutions.
|Minimum data requirements||Recommended|
|- Product feed, history of page views (at least 3 months)||- History of page views (at least 6 months)|
|- Transaction history (at least 3 months)||- Transaction history (at least 6 months, including different channels, if applicable)|
|- Periodic transaction updates (1 week) or online data stream|
In order to be able to recommend products, Synerise AI needs to know what products you are selling. That is why you need to provide this in the form of an XML file written in RSS or atom standard.
The product feed should be built according to Google Merchant Center Help recommendations. Synerise supports the tags described there.
Two namespaces should be defined in the XML Prologue:
All tags that are defined by Google Merchant Center Help, should be preceded with a “g:” prefix. The only tags that do not use the prefix are: title, name and link (written in angle brackets < >), and your custom tags that are not defined by Google Merchant Center Help.
Obligatory tags (written in angle brackets):
Schedule of AI implementation
|Implementing product feed||1 day|
|Integration of events via AI tracking or API||1-2 days|
|Batch update of historical transactional data||7 days|
|Algorithm training||0,5 day|
|Campaign setup||0,5 day|
AI Recommendation statistics
Synerise offers a high quality recommendation module that lets you create product recommendations boosted by the power of an AI engine. You just need to choose the recommendation type scenario and the AI engine will generate product offers adjusted to the scenario type based on the gathered data. When it comes to scenarios, there are plenty of options to choose from. You can offer your customers similar products (in terms of brand or category) to those they have seen earlier or products which are visually similar. If you want to encourage visitors to a website to make a purchase, you can prepare recommendation frame including top products in your store or you can adjust the recommendations to individual preferences.
Whichever option you choose, it’s still important to monitor the statistics it generates and optimize as needed. The recommendations will surely make impact on the users, which will be reflected in the revenue, CTR, the number of purchased products, etc. The statistics provide you with an insight into the recommendation frames results, so you learn how many recommendations are generated and how many of them are actually viewed.
Location in the system
To reach the statistics of AI Recommendations, follow the path:
Campaigns -> AI Recommendations -> Global stats
The first step you need to take in order to start the analysis is choosing the period you want to examine. To do so, click the date picker (1).
- Date picker – Choose the period you want to analyze.
- Clicks – The number of unique clicks on a recommendation frame. Each time the user clicks the product, a parameter is added to the URL. That is how we know that the clicked product comes from a particular campaign. These statistics are updated every hour.
- CTR – (Click Through Rate) It is the number of unique clicks on a recommendation frame divided by the number of views of the recommendation frame.
- Click Through Rate – The chart presents the proportion of unique clicks to the views of the recommendation frame in a given period.
- Recommendation viewed vs. generated – It is a contrast between two situations: when a recommendation was actually viewed (e.g. when a user scrolled down to the product recommendation and saw it actually) and when the user entered the website, the system generated product recommendations, however, the user failed to reach them on the website.
- Recommendation seen – The total number of views of the recommendation frame in the chosen period.
- Revenue – Revenue generated by the recommendations. It is counted when a customer buys a product within 24 hours after seeing the product in our recommendation.
- Products bought – it is the total number of the purchased items during the selected period.
- Revenue – it is a graphic representation of the gained revenue out of the product purchases. The bar below the chart lets you extend or shrink the time period of the statistics generated on the chart.
- Clicks in Recommendation – It is the graphic representation of the total amount of clicks on the recommendation frames during the selected period. For each chart included in the statistics you can extend or shrink the period length.
- Conversion Rate – It is the graphic representation of the results you achieve with regard to conversion. To count the conversion, the system analyzes the purchases in a given time to the total number of people who visited your website.
- Views – It is the total number of unique recommendation frames a visitor has seen.
- Visits – It is the total number of generated recommendation frames.
- CTR – It is the number of recommendation frame views divided by the number of generated recommendation frames.
- Clicks Products – It is the number of unique products which were clicked by the users.
- Views Products – It is the total number of times a user viewed the product in the recommendation frame in the specific time period.
- CTR Products – It is the number of times unique products were clicked divided by the total number of the product views
- Charges – It is the number of times a product was part of transaction (e.g. a certain product was a part of 10 different transactions).
- Quantity – It is the number of all unique bought products.
- Transactions – It is the total number of orders your site generated.
- Charges – It is a graphical representation of the products from our recommendation frames which were part of transaction.
- Quantity – It is a graphical representation of the number of all bought products.