Does the following scenario sound familiar to you? You publish a promo code for a specific purpose, then one of your site visitors figures out how to get that promo code and shares it with the rest of the world. Suddenly it’s all over Instagram or X, and within hours, your promo budget is spent without realizing your goal.
At best, you saw an increase in sales, but at worst your publicized promo has resulted in reduced margins with no topline impact, a customer service headache, an unexpected stockout, and an all-around embarrassment.
The frustrating thing about this scenario is that you put so much work into the promo (advertising, creative, financial analysis, budget approvals) and all that effort seems wasted. Meanwhile you have to report how your promo performed against projections.
This understandable concern is probably why we’re so frequently asked the question, “How does Session AI prevent shoppers from gaming the system?” The answer is that in-session marketing is a way to focus promotions on the visitors who need them with low risk of improper usage. Let’s explore why this is the case.
Early purchase predictions
In-session marketing uses early purchase predictions to decide which visitors receive a promotion. Early purchase predictions leverage clickstream data and AI to analyze visitor micro-behaviors and identify purchase intent. Since they don’t require any personally identifiable information (PII) or other historical context, early purchase predictions can be used for every session — including for anonymous visitors.
When the early purchase prediction indicates that a visitor is on-the-fence — that is, influenceable — it can trigger a real-time offer for that visitor.
A real-time offer is a on-site event that appears to the visitor. It’s “real-time” because the purchase prediction and generation of the offer happen in microseconds. The real-time offer will usually include a nudge to encourage the influenceable visitor to complete the purchase, such as:
- Time-limited coupon: A coupon with a ticking clock will convert more visitors.
- Free shipping: 53% of consumers say they shop online because of free shipping, outperforming discounts by 12%.
- Cart threshold discounts: Greater discounts for larger purchases can increase conversion rates, as well as AOV.
Real-time offers use early purchase predictions to segment shoppers and show promos only to a targeted group of visitors when the promos are likely to make a difference in conversion.
Gaming the promo system
Ecommerce promotions and promo codes are powerful tools to drive sales and attract customers. However, promo code abuse can have detrimental effects on your business.
Here’s a representative example of how shoppers could game the promo system:
“Here’s how you can get new boots on sale. Close your browser, remove all cookies, clear cache, then search Google for “Summer Hiking Gear” to see the landing page of abcd-sports.com for a 20% discount offer.”
These are the type of instructions that clever shoppers will share in order to help others game the system. This type of sharing is difficult to prevent as it can be hard to make your targeting rules sufficiently complex to prevent savvy users from finding patterns that always work.
Single-use promo codes can be a solution to gaming, but not a foolproof one. When creating promo codes, it is important to follow best practices to ensure they are effective and secure. Set clear terms and conditions for each promo code, specifying its usage limitations and expiration date. But best practices alone won’t prevent shoppers from gaming your promotions.
Single-use promo codes can’t go viral, but instruction sharing is a more difficult problem. A representative scenario of this sharing can look like the following:
“Here’s how you can get a discount code for new boots. Log in to the abcd-sports.com site. Click on the hiking boots banner on the front page, select a product, and leave it in your cart for at least two days. You’ll receive a 20% off single-use code in your email.”
Once a visitor discovers a reliable formula it can be tough to prevent the instructions from spreading.
In-session marketing: no games
The good news is that Session AI provides strong protection from discount gaming using early purchase predictions and real-time offers. Our AI includes multiple machine learning (ML) models that predict session outcomes, making it very difficult to game the discounts.
How can we be sure? There are 4 major reasons ML-driven in-session marketing is much harder to abuse than traditional sitewide coupons and single-use promo codes.
- The ML models are continuously retrained with the latest behavioral data, so the model that scores your visitors today might not be the same as the model that scores your visitors tomorrow. The model is optimized for the outcome you want – higher conversion – not for the outcome the promo-hunting visitor is looking for.
- ML models include timing components. For example, the amount of time between clicks, the time of day, and day of week that the session takes place are all included as inputs that help the model predict the outcome. A visitor is very unlikely to detect a pattern in these signals, and even if they could, the patterns that predict behavior may have changed (see above).
- We A/B test scoring models, which means visitors coming to the site at the same time may be scored by different models for different real-time offers. The selection of which visitor is considered for which offer is random. A visitor can tell another person that they received a certain offer, but even if that other person behaved similarly, they might not get the same offer.
- We also A/B test experiences, so a user who is scored by an early purchase prediction model may not receive a real-time offer because they are in the control group. If a talkative visitor shares their promo publicly, it’s certain that any group of people following their instructions to seek the promo will include some who never get it.
These features were all designed for the purpose of making the most accurate predictions and finding the best action to nudge those who need it to complete their purchase. And as a happy side effect of good ML, it is virtually impossible for site visitors to “game the system” as they used to with older approaches to promos.
Session AI has been running real-time offers based on our early purchase prediction models for some of the largest retail brands for almost a decade. We haven’t found a single instance of a user successfully gaming our promos.
Convert more of your shoppers with the maximum margin. That’s just what in-session marketing does!
Want to learn more? Contact us to speak with a Session AI expert.