Leapfrog to Predictive and Profitable Promising with AI

While aggressive and accurate estimated delivery dates (EDDs) are a top priority for today’s retailers seeking to boost customer experience and conversions, many have not prioritized it. This hesitation stems from two main concerns: the cost of central fulfillment and the complexities of leveraging stores, including labor constraints, picking efficiency, and potential order splits. Additionally, some retailers lack the robust analytics needed to create and monitor promise execution.

Retailers know that providing an EDD has become an expectation—the question becomes how aggressive it should be and at what cost?

Buy Online, Pick Up In-Store (BOPIS) was originally created as a speed “solve” when shipping promises were typically 5-8 days. It also helped drive retail traffic by showing inventory availability. This was the answer to Amazon in the early days. “You have Prime, we have BOPIS.”

But as online shopping continued to grow in overall % of retail and Amazon expanded to nearly 50% of all eCommerce volume, some omnichannel retailers started playing the speed game with Amazon. This is seen in the overall delivery speeds across the industry dropping in just the last year from 5.6 to 4.4 days .

Today’s promising strategies are a mixed bag. Some retailers avoid making a fast promise. Others pad their estimates to protect for the lack of precise visibility and the dials to set it. Some try to make a fast promise but suffer for it on the cost side. Then there are the lucky few with the product margins high enough to not care.

The cost components are not just speed related but also driven from order splits that spike when stores are leveraged to achieve that speed. We’ve also seen for years retailers take brute force methods to stop unprofitable orders. Some retailers resort to restrictive measures like shutting down online sales for certain items (SKUs), limiting nodes that can ship the items, and setting quantity limits to avoid order splits.

But every single order is different. These methods don’t respect the nuances. To make matters worse, these methods may also influence offline sales negatively.

Today’s promise engines have gotten smarter, crunching data on transit time, carrier rates, and inventory location to make accurate and faster shipping promises. All of this data unlocks a new level of control. Imagine a promise engine that can dynamically adjust shipping costs, strategically slow down a promise when necessary, or even recommend not making one at all.

This nuanced approach allows you to solve multiple goals simultaneously, ultimately impacting your bottom line.

This could be the case for change for the many retailers who have yet to prioritize promising because of the cost and complexity of executing a fast and accurate promise. The great irony is that a promise engine has the capacity to mitigate the very reason retailers are leery of investing in one in the first place.

Take this example:A consumer adds four different items to their cart. Using a smarter promising engine, the retailer has enough data to know this order is going to be shipped from three different nodes. The dreaded split. It also knows the shipping costs will exceed profitability. In today’s world, that order is taken, and the damage is done. But should it?

These are the realities, yet promising and sourcing engines out there are still pushing features like candy to retailers. If you can set a dial—then the problem is solved! But if you have too many dials it becomes unmanageable and even unexplainable in some cases.

The work we’ve done with leading retailers with this ML model has shown the prediction is accurate above 99% of the time. More impressive, it typically shows 2 to 3 days improvement against your current promise. There is even more benefit in the areas of drop shipping where you are likely making very conservative promises. For one retailer we worked with, we saw a 6 day improvement in the predictive promise vs. the static one based on vendor contracts. Even more impressive though, it comes WITHOUT the heavy data analysis and setup required by most promising engines.

Using this process eliminates one of the large barriers in the market today to profitable promising and starts you down the journey of making an accurate promise. But of course, this predictive promise is a representation of current performance and doesn’t consider the profitability goals you have across your channels.

From here you’ll want to leverage the rest of our promise features to layer in these strategies.

In our opinion this should be done with a combination of features and the necessary explain ability and testability with them.