Discounts are an effective, but costly way to entice prospects to buy something in an online shop, the higher the discount the more likely the sale. Not everyone needs to be enticed though and those who would buy something without discount make every discount campaign an expensive undertaking, where you lose money on the sales, you would make anyway vs. the additional gains of sales, that would not have happened without the discount.
The perfect solution would be to only offer a discount to those, that would not buy without it and not interfere with potential new customers in the checkout process, who have already made up their mind. Webtrekk's artificial intelligence approach of predicting the conversion probability of a user is a good approximation of this perfect world, at least in terms of selecting whom to offer a discount. The predictions are usually available based on historical data, taking into consideration all behavior of a user in the past, but Webtrekk also provides real-time predictions based on the current session data, enabling the classification of first time visitors. All you need to do then, is to define thresholds based upon the conversion probability, e.g. everyone below 60% probability, and present them with a discount, whereas users above this threshold won't see any discounts, thereby reducing the cost of the discount campaign.
Additional factors, that could be taken into consideration to decide upon the discount, are the expected customer lifetime values of users, offering a bigger discount for high value customers or just free shipping for less promising customers.
The right time to show the discount is shortly before the user is leaving the shop. The real-time behaviors in Marketing Automation offer exit intent as trigger to display a layer shortly before the user leaves and these can be combined with thresholds for the conversion probability: The user wants to leave and has a conversion probability for the next 30 days lower than 60% then show the layer with the discount offer.
Naturally, besides discounts other kind of campaigns, e.g. free shipping or longer return periods, can be combined with Webtrekk's predictions.