Skip to main content

Explore and Exploit

Cherry's artificial intelligence (AI) uses reinforcement learning to increase your revenue. Essentially, it tests Promotions out and allocates more customers to Promotions with higher benefit .

Behind the scenes the AI balances two modes, Explore and Exploit.

Explore

In this mode, the AI is allocating Promotions to customers to maximise knowledge. It is allocating a significant amount of customers to new and novel Promotions to determine if customers will take them up, and if so, what is the revenue uplift.

Exploit

In this mode, the AI is allocating Promotions to customers to maximise value. It is allocating most customers to Promotions that deliver most revenue to you.

Explore then exploit

Cherry's AI allocates a mix of traffic to both Explore and Exploit.

current

Upon deployment, the AI is primarily focused on learning. It then utilises this learning to allocate more and more traffice to the most valuable offers (e.g. Exploit).

This allocation maximises the revenue uplift 1-3 months after starting.

It's a balance

As seen in the picture above, the AI doesn't allocate 100% of customers to either Explore or Exploit. When the system is starting, it's important to allocate some traffic to Promotions you know have significant value (e.g. the pre-AI control).

After the AI has learnt a lot and knows which Promotions are best, it still keeps a small amount of traffic on Explore. Any changes or drifts in customer behaviour are then picked up quickly and adjusted to.