E-commerce Reverse Logistics

Optimising for profit

Last updated: 2 November 2021 · 6 min read
E-commerce Reverse Logistics


Price discrimination is an effective means of optimising revenue and profitability. By forward selling returned goods, retailers can improve cash flow and sell through rates. Statistical modelling and other tools give retailers the business intelligence needed to forecast returns. This enables forward selling of inventory.

The true cost of returns

Why are reverse logistics costly? Merchants offering free returns obviously bear the costs of postage or courier fees. For merchants who don’t offer free returns, the process is still costly. Firstly there is the labour cost associated with inspecting and repacking the goods for resale. In many cases this cost exceeds the gross profit of the product itself and merchants simply write off returns.

For those merchants who re-stock returned items, the true cost lies in the fragmented nature of reverse logistics. Put simply, having goods in the wrong place at the wrong time. The problem is especially acute at the time of writing this article (Autumn 2021). COVID-19 and Brexit related supply issues are causing severe disruption for many merchants. Running short of stock, whilst experiencing high return rates is particularly galling!

Selling forward

Let’s think about Christmas and Black Friday for a moment. Merchants know this will be a busy time. They ensure they have inventory to meet the expected demand. This may be easier said than done this year, given the ongoing supply issues. Following the holiday period the merchant is inundated with returns. It’s a pattern that’s been repeated for decades.

In the “old days” merchants could clear unused stock in the January sales but this is no longer a viable option. Flash sales and continual discounting have dented the effectiveness of sales. Savvy shoppers take advantage of consumer laws covering online sales. They simply wait for the January sales to return and re-purchase goods at a lower price. They get goods today, at tomorrow’s prices. There’s nothing online retailers can do about it.

What if we turn the situation on it’s head? Instead of waiting for stock to come back then trying to resell it. We sell it now, for future delivery. The concept of a futures contract is nothing new, but it’s not common in the retail space. Let’s look at how this might work.

E-commerce merchants have a wealth of data available to them, everything from product return rates through to individual buying behaviours. This data is an untapped gold mine that can be used to build powerful predictive models. Along with the merchant’s own data, we can use freely available external data e.g. weather data to improve the accuracy of predictive models. How many heavy jackets are returned following an unusually warm spell of weather?

This isn’t intended to be a technical article, so I won’t go into the detail of machine learning and statistical modelling. Suffice it to say, we can build models that predict what will be returned, and when. They’re not 100% accurate, but they don’t have to be. I used the term futures contract earlier, but this isn’t what I’m actually advocating. I propose retailers offer goods for pre-order. Retailers take a pre-authorisation but only take payment as and when goods come back into stock.

Regression analysis

For each product/sku we want to predict how many items will be returned and by when. “when” can be represented in terms of “how many” days after purchase. It’s classic regression analysis, specifically multivariate regression (we have two dependent variables). We can simplify things by fixing one of the dependent variables i.e. how many items will br returned within 14 days. This is known as multiple regression.

Several techniques can be used for regression analysis, from simple linear regression through to random forests & neural networks. My advice - KISS. Complexity bias has killed many an IT project. In my own experience, success will depend on:

  • The quality of data
  • The quantity of data
  • The relationship between the business and data scientists

This modelling allows merchants to offer products today, for delivery in a few weeks. Why would a consumer be willing to pre-order? Firstly the merchant may actually be out of stock today, a scenario more likely today than it was 12 months ago. Merchants can also afford to discount or otherwise incentivize shoppers to buy for future delivery.

Sell-through is improved. By selling forward, retailers carry lower inventories, and avoid writing off returns. Cash flow is also improved. Retailers see a lower initial outflow due to reduced inventories. When returns do come back, one customer is refunded and the next is immediately charged.

This strategy is also a great conversion optimisation tool. Traditionally, shoppers know prices will be lower in January than December. There is a strong incentive to wait. As I previously mentioned, even if shoppers do buy in December they retain the option to return and repurchase in January if prices drop. By selling forward, retailers can tackle this issue head on. The only way to obtain goods at a lower price is by buying today for future delivery, waiting is not an option.

Incoming returns

The strategy just proposed is based around on medium term estimates - what is coming back over the next few weeks. We can tweak the strategy slightly to work with near term estimates - what’s coming back in the next couple of days. Before explaining how to we can make these forecasts, let’s look at how we would use this data.

Are shoppers willing to pay more for the immediate delivery of items? Are they willing to wait if they can obtain a discount? Yes they will. Most shoppers do not pay extra for premium, next day delivery. Price discrimination is a proven, successful strategy for e-commerce merchants. Retailers can offer discounts or other inducements to shoppers who are willing to wait a couple of days.

Retailers already employ price discrimination

Of course merchants can’t forward sell products unless they are pretty confident they will have the inventory to make good. The merchant can say with some certainty what is coming back in two or three days time. That’s why retailers need to know what’s coming back …

For the near term we don’t rely on forecasts and statistical modelling, we have more accurate data available. By asking shoppers to book a return online (via a website, chatbot or other means), merchants gain forward visibility. They can track which goods are coming back, when they are due to arrive, which depot they will arrive in etc. Of course the data can’t be treated as gospel. Items will be lost or damaged in transit, or they may not be in a saleable condition. The majority of goods will arrive back in a saleable condition though.

This strategy allows all retailers, even those placing long forward orders, to incorporate a just in time supply chain into their operations. The strategy is self-balancing - the higher the rate of return, the more the retailer can reduce forward orders and inventory.


Most retailers regard returns as a costly, but necessary aspect of online trading. Along with courier and labour costs associated with repacking and restocking goods, there are also hidden costs. Reverse logistics often result in inventory being in the wrong place at the wrong time, compounding supply issues. Statistical modelling can be used to predict returns into the medium term. Parcel shops and couriers give retailers forward visibility of inventory changes in the near term. This business intelligence can be used to forward sell products. This is especially beneficial around the holiday season.


Do you have questions or comments about this article? Drop me a line:

Try Viko

Want to see Viko in action? Try our interactive demo