Is inept forecasting holding back online fulfillment?
Photo: Walmart

Is inept forecasting holding back online fulfillment?

According to a survey from Peoplevox, purchasing and forecasting is the most common challenge to the e-commerce fulfillment process.

The study found that 24 percent of the 154 retailers surveyed worldwide indicating that purchase and forecasting was the one fulfillment area that required the most improvement in 2017.

In its report, Peoplevox, an e-commerce warehouse management system provider, said suppliers are striving to gain real-time visibility of stock levels and sales volume, both key indicators when planning what to buy. Other challenges in the area of purchasing and forecasting include avoiding buyers working off historical sales data and dealing with a wide SKU base.

“In determining the level of optimum inventory for operation, this will only be achieved when a full range of what’s wanted is actually available and stock doesn’t sit gathering dust and subsequently requires discounting,” said Peoplevox.

Ultimately, retailers may choose to rely on “back to back ordering” — or issuing purchase orders only when customer orders arrive — versus always having all items in stock. Often, back-to-back orders are drop shipped directly from suppliers to customers.

“With increasingly reliable global supply chains, the need to keep everything on sale in stock is diminishing with more retailers electing to sell certain product ranges that they will only purchase on the back of sales order demand,” wrote Peoplevox. “This allows businesses the flexibility of fulfilling a far more extensive range of products to customers while optimizing their up-front purchasing costs and available storage space.”

Other findings from the survey:

  • Only 53 percent of the retailers were happy with their fulfilment and warehouse operations.
  • Sixty-three percent admitted to not always shipping on time, with 34 percent citing the unavailability of stock being a key reason.
  • When it comes to rectifying shipping errors, 80 percent appear accepting of the additional carriage, customer service and warehouse labor expenses as an inevitable cost of order fulfilment.

BrainTrust

"Direct-to-consumer and allowing customers to return products to the location of their choice has created a lot of uncertainty."

Paula Rosenblum

Co-founder, RSR Research


"Forecasting and fulfillment are challenges from an analog system. These issues would disappear if they were addressed through digital lenses."

Adrian Weidmann

Managing Director, StoreStream Metrics, LLC


"I will stick with my same old story. Item level inventory visibility is still the root of success! And that’s my 2 cents."

Lee Kent

Principal, Your Retail Authority, LLC


Discussion Questions

DISCUSSION QUESTIONS: What do you see as the main hurdles to improving forecasting for online fulfillment? Do you see back-to-back orders or drop shipments as viable solutions that retailers should explore?

Poll

19 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Paula Rosenblum
Noble Member
6 years ago

The main hurdle to improving forecasting for online fulfillment is return rates. What do you do when you know 25 percent of the product is coming back and you don’t know what state that product is going to be in? Do you buy more? That’s a big over-buy. It’s a problem, and it’s not going away any time soon.

As for drop ship, our data tells us that retailers are unwilling even as manufacturers ARE willing. Makes sense, really. No retailer wants to be dis-intermediated or to give out their customers’ names and addresses to a frenemy. More than 50 percent of vendors in surveys we have performed tell us they are also selling direct to consumers. That makes them frenemies.

This is a really tough problem. I spent a lot of time trying to understand why apparel retailers in particular, who always had a very good handle on their inventory when I was a practitioner, suddenly had very poor visibility into it. The systems didn’t get worse, after all. Finally the light dawned and I realized that the combination of direct-to-consumer and allowing customers to return products to the location of their choice (including stores) created a lot of uncertainty around what they really own at any given point in time. I’d love to stroll through some retailers’ back rooms and look at the returns from online sales waiting to be processed.

I’m not sure how to fix it beyond trying to drive traffic back to stores.

Roa G
Roa G
Reply to  Paula Rosenblum
6 years ago

“25 percent of the product is coming back … I’m not sure how to fix it beyond trying to drive traffic back to stores.”

Would not an omnichannel software approach greatly alleviate this problem? It’s a tall order for a company to take on, but it’s the most effective technology we have today for, at the very least, the retail/fashion industries.

Neil V Cooper
Reply to  Roa G
6 years ago

What sort of things do you sell? Why are you getting back 25% of dispatches? Fashion can get high returns, especially if sizing is wrong, but non fashion retail you should be getting under 4%! Have a look at what other fashion retailers do to try and reduce returns?

Chris Petersen, PhD.
Member
6 years ago

When customer expectation goes to shipping/receipt within 24 hours or less, there is precious little back stock for fulfillment.

Retailers have historically used stores for buffer stock, but that is increasingly dangerous and inaccurate as more customers shop online.

Back-to-back ordering essentially resulting in drop shipments is one collaborative strategy where distributors can help fill the void. However as online shopping and volumes become more volatile, it is becoming increasing difficult to forecast demand for hot products with a short fulfillment window.

While not perfect, one has to ask how Amazon does so well with Prime. There are at least five keys to their success:

  • More reliance on predictive AI versus just historical sales data;
  • Reliance on Amazon Marketplace to help fill in the void for critical products;
  • Creating very accurate forecasts for the consumer as to when the product would arrive (not everything is required in one or two days);
  • The subscription model creates and incredible forward view of demand curves;
  • One of the most sophisticated rapid response delivery systems in the world.
Bob Amster
Trusted Member
6 years ago

Forecasting for online fulfillment varies by product category. It is not easy to forecast fast fashion. Razor blades can probably be forecast to within 95 percent accuracy. Seasonality, history, and a pinch of human touch, each assigned a weight factor, are the main ingredients. The rate of non-sellable returns disrupts that formula.

Ben Ball
Member
6 years ago

“Forecasting” (aka — guessing future demand based on history) is and always will be imperfect at best. The solution is the link of AI and IoT in every location and fixture where an item is warehoused, stocked or merchandised. When a retailer’s inventory management system is being told directly and in real time that a unit of SKU 2378 just got removed from Store 54 or Bin 36 or whatever — and the same system is using AI to create constantly adaptive restocking processes — it is possible to envision a world of zero out-of-stocks. At least those that are forecast error-related. This applies equally to online, brick-and-mortar and omnichannel retailers, it doesn’t matter. Toss in block chain technology to eliminate the transaction errors and the retail operations world starts to look like a much better place.

Harley Feldman
Harley Feldman
6 years ago

The number one main hurdle to online fulfillment (or in-store fulfillment for that matter) is inventory visibility. If there were 100 percent accuracy in knowing the level of SKUs available and where they were located, fulfillment would be straightforward and replenishment could be executed with a high degree of accuracy. The best way to attain this inventory accuracy is using Internet of Things technology to collect the inventory data remotely, accurately and using technology instead of humans. The method for tagging the items to be to used for IoT data collection is RFID. This combination of IoT and RFID is becoming less expensive and, with many new technology solutions from smart shelves to Warehouse Management Systems, a better bet on the objective of 100 percent inventory visibility.

Adrian Weidmann
Member
6 years ago

Forecasting and fulfillment are challenges from an analog system. These issues would disappear if they were addressed through digital lenses. Redesigning the existing workflows and established processes using new digital technologies like RFID, AI and data management along with revamped business processes like consigned inventory and brand-based shop-within-a-shop concepts can make supply-chain management and forecasting challenges a thing of the past. Retailers are still using a crew of twenty-something-year-old category teams that track spreadsheets and conduct business via specific analog business rules and constraints.

Lee Kent
Lee Kent
Member
6 years ago

Since only 24 percent said that purchasing and forecasting require the most improvement, I will stick with my same old story. Item level inventory visibility is still the root of success! And that’s my 2 cents.

Jackie Breen
6 years ago

There are two major components required to improve fulfillment forecasting: real-time inventory visibility and the right technology to leverage that information. Once retailers have overcome the hurdle of gathering all the information, they then need the ability to define workflows, processes and trends using automation and AI technology. The fact that so many retailers today are using antiquated systems and managing all this data through spreadsheets and manual processes makes it unsurprising that 47 percent of retailers were unhappy with their fulfillment operations.

Ralph Jacobson
Member
6 years ago

The quest to continually shorten supply chain transaction times has created “demand” for improved process (e.g., “Back-to-Back,” etc.) and more capable technologies. What we’re talking about here is the myriad internal (to the retailer) and external forces affecting forecasting accuracy. I can tell you that major retailers are right now leveraging data from sources like local store location events (sporting events, local news, etc.), weather patterns, social sentiment, internal call center data, etc. to better match forecasting curves to their actual product demand curves. We are seeing some significant improvements in forecasting using all these sources of data and the numbers are as high as a 17% improvement in forecasting. This is being done with the help of machine learning technologies that consume vast amounts of data and generate increasingly accurate forecasts. These technologies, along with process improvements, like those mentioned in the article are ways to address the online forecast challenge.

Cate Trotter
Member
6 years ago

A lot of improvements in this area can be made through better inventory data. Some retailers are already moving towards this by using RFID tech to tag and check items. As this tech keeps getting cheaper and better there could be a larger uptake.

Consistent, up-to-date, accurate inventory data could help retailers to use their stock better, for example shipping products from nearby stores that have the item wanted available, which may reduce over-ordering/waste.

Forecasting is a challenge as it can be impacted by so many things (the weather for example if increasingly hard to predict, before you even consider return rates, fashion trends etc). But the better the inventory and sales data the better this can also get.

At the end of the day customers aren’t too worried about the mechanics of their order (whether it’s back-to-back ordering for example) as long as it arrives when they expect. Usually that’s within 24 hours or less these days so improving inventory data is a must!

Doug Garnett
Active Member
6 years ago

Fascinating set of answers here. And I’m reminded that perfect inventory management only happens by limiting sales (and your store’s future).

For a robust and strong retail future, there is no perfection in inventory management — only better and worse. And that’s true for all retailers (online or offline).

A digital inventory management system can make it smoother, but won’t fix it all. Inventory management problems don’t start at the DC — they start months earlier in the supply chain. The costs of shortening the lead times are very high if you’re moving volume. You might be able to foist those costs back onto the manufacturer, but the costs are still there, only more hidden.

It’s probably a good thing that the statistics show unhappiness on the part of executives. That suggests generally robust businesses driven by the market. If online retailers are doing well, then they will have more, not fewer, inventory management problems. What they need to eliminate are those bottlenecks and shortages that limit their ability to be financially healthy.

Neil V Cooper
Reply to  Doug Garnett
6 years ago

Like a store, not all space is equal; different presentations and positions, not forgetting price create different demands! It’s possible to test and roll out multiplying by the known uplifts of the different treatments!

Doug Garnett
Active Member
Reply to  Neil V Cooper
6 years ago

That’s all true … to a point. However, a robust market makes testing more complicated than it seems it should be. Also making conservative assumptions on every test and every inventory situation won’t create success.

That’s why I suggest that a level of inventory problems needs to exist or a retailer isn’t aggressive enough to succeed. And I’ve worked with major retail organizations where inventory problems are the primary focus rather than customer opportunities. Those retailers don’t thrive.

I am probably describing it wrong.

Neil V Cooper
Reply to  Doug Garnett
6 years ago

I had a very robust market, I never had testing problems or inventory problems. I tested, tested and continually monitored every sku, line by line. Roll out forecasts (more accurate than assumptions!) were always pretty accurate and I carried no stock that I didn’t want more than a few months. The great thing with transactional web sites and bricks and mortar is that you can’t oversell with a real-time stock position. And it’s not hard to work out from that position forward orders and demand.

Neil V Cooper
6 years ago

It’s unbelievable that retailers are still debating these basic issues after 22 years of the Internet. It’s like the back end is in the too hard basket! Let’s job it out to a third party warehouse and fulfilment centre, we won’t pay too much attention to stock and forecasting as we are firstly bricks and mortar and the internet is, well, just something nobody in our organisation really understands or wants to take responsibility for! Unless you are in fashion, 25% returns? Really? You are doing something majorly wrong! You are looking at the whole thing the wrong way around! Blank sheet of paper time!

Franklin Chu
6 years ago

The ability to see the stock level in real time is very important to planning e-commerce sales events. When customers come to an e-commerce site from different channels, the last thing they want to see is that the products have been sold out or – even worse – receive an email saying their order’s been cancelled.

Being able to forecast accurately is a core capability for running a successful e-commerce business in China. Retailers would know when and where to market, and allocate a sufficient budget for each promotion. This capability can be only built upon trusted data analytics, as well as a good understanding of both marketing channels and their own supply chain capacity.

Neil V Cooper
Reply to  Franklin Chu
6 years ago

I agree Franklin Chu.