In my current role at Dunelm, I spend some of my time producing a daily report to monitor performance across our ‘Direct To Customer’ vendors and benchmark against our central fulfillment operations. In this post I’d like to give an overview of supply chain reporting and explore why a rounded set of metrics is important. 

Why Supply Chain Analytics?

To be considered an outstanding retailer, a set of expectations has to be consistently met. These expectations have changed over time, and are always rising. 

There was a time when success was driven by store estate coverage, and the availability and range of stock held within your stores. As the web began to mature, browsing experience became important. Now, excellent delivery capability sits firmly in customer expectations. Supply Chain Analytics helps businesses understand their post-purchase operations and improve the customer experience.  

No doubt there are exceptions to this generic narrative (some large retailers are digital pure play), but the underlying principle is solid: get your customers what they want, when they want it, for the best possible price, and with the best possible service.

What to measure? 

We can think of the supply chain as a series of touchpoints (web order placed, product picked, product packed etc) each with an associated time stamp. Measuring the time between touchpoints gives us an array of speed metrics. Top level, we can track the time between customer checkout and customer delivery, but operationally, it is useful to understand the speed of a pick, or the time taken to load a lorry. 

All of these speeds should be compared to the various ‘promises’ made in the background – the expected delivery date announced to customers on site, or the service level agreements made with Hermes and other carriers. It’s within customer expectations for a made to measure sofa to take longer to arrive than a cushion, but once we have advised the arrival date, we should strive for a 100% delivery to promise.

As operations (both internal and with external DSV vendors) become busier, lead times increase and late dispatches become more frequent. Therefore, it is sensible to track volume flowing through the network: order counts, line counts, unit counts, parcel counts, both inbound and outbound. The number of orders waiting to be picked and dispatched (the order well) is another useful indicator of whether a vendor is falling behind.

If volumes begin to challenge warehouse or vendor capacity, then action may be taken to reduce inbound orders on specific skus – toning down PPC spend / removal from the PLA feed, dynamically sequencing these products further down the page, or temporary removal from the site.

Late orders more frequently trigger a contact to the call centre, with customers wanting reassurance that their order is on the way. Self-serving this info, through parcel-tracking portal for example, is helpful to this end. Understanding patterns in refund levels can suggest disservice – the longer the customer has been waiting, the more likely they are to cancel their order. Refund volumes inherently record volume heading through the network in the opposite direction.

Taking a step back from a bewildering array of metrics, the notion of a ‘perfect order’ is a helpful one. Aggregating the key components can provide a single KPI of performance, suitable for a quick glance sense check of yesterdays performance. The definition is subject to business constraints, and some common criteria may include:

• Delivered in Promise Time
• No Call Centre Case
• Not Returned
• Delivered in perfect condition

At the end of the day once the parcel has been delivered, the customer NPS (Net Promoter Score) serves to record customer feedback. A high NPS can signal propensity for the customer to return or recommend us, which can feed into Machine learning models to predict customer churn. The features which drive NPS would be sensible criteria for the perfect order metric described above, drawing some metaphorical lines in the sand where customer expectations currently lie.

A wealth of Information, A poverty of Attention

Having the above for each and every order line gives us a wealth of information. But, as the saying goes – a wealth of information creates a poverty of attention. It is important to plan what information will be presented, and to whom: The presentation of the weekly trade pack for the exec should have different content than a daily vendor deep dives for the carrier compliance team.

Dashboarding tools, such as PowerBI can serve well for some purposes, but an automated python script dumping daily summaries into an excel workbook can serve better for focusing follow-up work.

Goodhart’s Law – Metric Diversity

In the early 20th century, the Vietnamese city of Hanoi had a rat problem, following the installation of a new sewer network. The solution, the governors proposed was to offer a small bounty for every rat tail submitted. Surely, this was the way to encourage the populous to capture and kill the rats – problem solved?! 

It started well, but as time passed, the rat population seemed to be on a worrying rise. Rats were spotted, alive and well, just without their tails. Reports of rat farms to harvest extra tails led the government to rescind the deal.

What has this story of rats got to do with supply chain reporting? The lesson is to be careful with the measures you track, for too much concentration on a single KPI can bring unintended consequences. This is commonly known as Goodhart’s Law.

Tracking a balanced mix of metrics is critical to avoid falling into this trap. Setting the number of parcels as a standalone target for example, could lead to every item being boxed separately to increase the parcel count. Long term metric tracking should be given weight alongside short term metrics too.

I’ve been learning a lot about supply chain operations, and how to share of view of performance. Understanding these measures is foundational to the other projects I am working on.

Until next time,

Scott