Offers and promotions are everywhere but how do brands know what the right offer is? Is there actually a right offer or are there multiple “right” offers? The answers to these questions factor heavily into a brand’s profitability and overall marketing strategy.
For many brands, the right offer is the same offer they’ve been running forever. The offer is put out to the market, more product gets moved off shelves and everyone is happy. Rinse and repeat, right? Well, that’s one way to address the issue but artificial intelligence and machine learning can greatly improve the results.
Machine learning and AI techniques allow us to observe behavior at scale and categorize behavior or patterns more efficiently than ever. Using structures such as Bayesian Networks, probability theory, and other techniques allow us to replay events, determine causality, and accurately measure results based on a certain set of criteria.
Savvy marketers will test offers and see what the ROI of, for example, a 20% discount vs. a 10% discount could be. Simple A/B testing can be effective and is definitely a step in the right direction when compared with rinse and repeat. Taking it one step further, we can utilize AI to vastly improve those existing results. Analyzing behavior with different offers and modeling this data to build graphs allows us to determine, for example, price sensitivity thresholds for types of consumers. Using these models and segments we can predict a consumer’s receptiveness to a given offer and make a decision on the right offer for that consumer. The next consumer may require a different offer to maximize the likelihood of a purchase.
Building models also allows us to perform “what if” analysis. Taking past activity and results, we can replay the activity against a different set of offer values applied to the same model. Replaying past events will reveal additional insights such as the use of previously unused offer values to continue moving product but at a lower cost. These “what if” scenarios can be a very effective planning tool based on actual data and actual past results.
As activity continues across one or more offers, the system continues to observe behavior and learn. This ongoing learning process requires the periodic update of the model to ensure accurate and optimal performance. Updates don’t necessarily need to be strictly periodic and, in fact, can be done in real time for an even more robust system. A very mature model shouldn’t even require real time updates so it really depends on the use case and variety of change. Regardless of how often the updates are done, keeping the model as up to date as possible is an important step that can’t be skipped. These observations simply can’t be done at scale by a human.
So, back to the question – what’s the right offer? The answer depends on the goal but for the vast majority of marketers there won’t be a single right offer. The right offer is the one that motivates the consumer to purchase while providing as little, if any, discount as possible. Often, the right offer may change due to external factors. External factors such as steeper offers from a competitor or economic volatility will have an affect on purchase behavior, which is why keeping the model up to date and utilizing a variety of different data points is critical.
Greg Hansen is the CTO & Co-founder of RevTrax