The Power of Dynamic Pricing: The Benefits of AI-Powered Pricing Compared to A/B Testing

If you hear dynamic pricing and think ‘Taylor Swift and Ticketmaster lawsuit’ you’re not alone, but when you finish this blog we’re hoping you have a new frame of reference, and a renewed desire to explore dynamic pricing for your business.

You know that your pricing strategy is one of the most important decisions a business makes. Get it right and you will maximize your profits and grow like gangbusters. Get it wrong and you lose customers, revenue and market share.

Given what is at stake, you may be thinking pricing strategy does not seem like fertile ground for risk-taking, and you’d be right. However, this doesn’t mean it’s not ripe for innovation.

Dynamic Pricing vs. A/B Price Testing

Starting with the basics, what is dynamic pricing? Dynamic pricing is the practice of adapting product prices automatically based on customer behavior insights. The most effective dynamic pricing solutions use AI and machine learning to inform their models, thus AI-powered pricing.

Many believe AI-powered pricing sounds complex (and expensive) and is therefore reserved for the largest retailers. At Spresso, we assure you this is not the case. In fact, with solutions like Price Optimization, a retailer can leverage an out-of-the-box SaaS solution making, AI-powered dynamic pricing available to the masses!

What is the alternative to dynamic pricing? Historically, ecommerce retailers have run A/B price tests to assess the optimal price point for various products. A/B tests have their roots in statistical analysis and are widely used to measure the impact of a hypothesis statement by comparing two versions of something to see which performs better. In a pricing scenario, this would be comparing two different price points, an A and a B, for a particular product and then measuring conversion to determine which price performs best.

While this can be a beneficial strategy, there are drawbacks to price testing in a purely A/B capacity. In order to better understand why Price Optimization is a superior pricing strategy, it’s helpful to review how it came to be.

How Price Optimization Came To Be, A Brief History Lesson:

The journey to Price Optimization wasn’t an overnight revelation. Rather, the retailer from which Price Optimization originated ran the gamut of price-testing strategies before building their own solution.

This retailer’s first foray into price testing began with testing one price, measuring results, and then testing another price. Challenges with this approach were abundant as it was difficult to isolate the role price paid in tracking against a goal. Did a product convert better because of the new price point or because marketing sent out a push notification that highlighted that category? It was impossible to isolate and that is why this stage of price testing did not last long.

Next, the retailer explored price scraping, looking at competitors’ sites to gauge prices, and then adjusting their own prices. While this approach yielded slightly better results, it remained insufficient, specifically for the retailer’s unique product assortment of both private-label products and bulk sizes. Ultimately, the retailer realized that its customer base and brand were unique and may not even be shopping the competitors' websites, so basing prices on competitors’ pricing was never going to be an effective strategy.

Next was A/B price testing, and while there was incremental improvement the strategy, it still proved fundamentally ineffective. The primary reason is that in an A/B test, the “wrong” price was shown 50% of the time for the duration of the experiment - and that assumes one of the prices being tested was the optimal price. Showing the wrong price for the duration of an experiment is a concept called “regret” and we’ll elaborate on it shortly.

Thus after years of experimentation, the retailer arrived at the conclusion that there is no solution available that does exactly what our business needs – we will build it. And Price Optimization was born.

Price Optimization And The Fascinating World of MAB Algorithms

The foundation of Price Optimization is a multi-armed bandit algorithm (MAB). MABs are at the core of some of the most successful, innovative technology - it’s what powers Spotify and Netflix recommendations (even what art is shown to viewers!), as well as ecommerce powerhouses like StitchFix. MABs are designed to efficiently allocate resources among multiple options to maximize rewards. In the context of Price Optimization, the MAB dynamically allocates website traffic to different price points, learns from the customers’ behavior, and adapts over time. This real-time adaptation, or dynamic optimization, is where the magic happens, allowing businesses to consistently find the optimal price point as the solution optimizes for conversion, profit, or a combination thereof.

At the crux of MABs are two principles: explore and exploit. Understanding their definitions helps articulate the benefits of MABs, especially when compared to A/B tests

• Exploring - beginning the experiment by trying new options, the model is acquiring new knowledge, recommendations are given with an unknown outcome
• Exploiting - go with the known favorite, the model makes optimizations on existing knowledge, recommendations are given with a known outcome

There is another important definition when comparing A/B tests with MABs. Regret is defined as the difference between the reward you achieve and the reward you could have achieved if you’d made perfect decisions. With Price Optimization exploration and exploitation come together to minimize regret, in other words, the “wrong” price is shown significantly less frequently than with A/B price testing.

Comparing and Contrasting A/B Price Testing vs. MABs

While we are of the mind that MABs are superior to A/B price tests, each has pros and cons.

A/B Price Testing:

Benefits

• Easy to begin and implement
• Reliable for isolated testing

Drawbacks

• Fails to adapt to consumer behaviors in real-time
• Regret is present as the “wrong” price is served for the duration of the test
• Lack of ongoing exploration once test is completed
• Can be expensive to run multiple tests

MAB & Dynamic Pricing:

Benefits

• Minimize regret during testing period because traffic is dynamically shifted to best performing price during experiment
• Ongoing exploration means you’re always hedged - if another price starts performing better, your site traffic with adjust accordingly
• Maximizes for business goal, able to optimize for conversion and/or profit depending on the goal of the business

Drawbacks

• While easy to set up, greater implementation effort than A/B test
• Website traffic must be significant enough for the MAB to explore, and then exploit

For more detail about the benefits of dynamic pricing check out our recent blog about dynamic pricing for DTC.

Why Dynamic Pricing is Superior to A/B Price Testing

Dynamic MAB-driven pricing solutions offer a myriad of advantages over A/B price testing:

• Maximize Returns: Dynamic pricing optimizes for multiple objectives simultaneously, enabling the seller to optimize for conversion, profitability, or both
• Real-Time Optimization: MABs continuously learn and adapt in real-time, ensuring the price shown to the consumer is the optimal price given market conditions and customer behaviors
• Always-On: in a MAB there is always exploration even during exploitation, meaning the solution is able to pick up on market shifts and adjust accordingly

Dynamic Pricing: A Must-Have for Online Sales

As business’ roadmaps reflect the shift towards improving profitability, the investment in revenue-generating technology becomes an obvious one -with pricing being at the epicenter. The utilization of MABs for online pricing offers an adaptable and efficient method to find the optimal price point. While A/B price testing has its merits, it falls short in comparison to the responsive nature of dynamic pricing, like that which is offered with Price Optimization. As businesses strive to stay ahead in the competitive landscape, harnessing the power of AI-driven Price Optimization is the strategic choice to increase revenue, improve margins, and boost customer satisfaction.