The Enemy Of My Enemy Is My Friend

Good Morning,

First, as always, thank you for joining.

Two pieces today, both inspired by different conversations and projects. Both are a little bit more technical, but help you understand what goes on behind the curtain and the impacts that seemingly small decisions can have.

If you have been enjoying the newsletter and think someone in your network would too - please forward them this week’s edition and they can sign up here for themselves!

Here’s what this issue brings:

• Rumour is that Amazon is opening up its “Key For Business” platform to third-parties. I’m sharing my thoughts and what had me take a 180 on my initial position
  

• Ever wonder how you should start analyzing your SKU mix to understand if you have the right items in the right bins at your warehouse? I’m sharing an intro to approaching that problem
  

Would Selling Access To Other Last Mile Providers Be The Right Strategic Move For Amazon?

Last week, Jonathan Hessney made this post about Amazon.

In it, he shares some (insider) information that Veho will seemingly be the first external carrier to get access to the Amazon Key For Business network.

I shared some thoughts in the comments of his post, but you’ll get the full reasoning of my position here (which won’t be shared anywhere else) and why I haven’t changed my mind.

Background

What is Amazon Key For Business anyways?

Amazon Key For Business is a service designed to streamline package deliveries for multi-unit residential buildings and commercial properties.

It works by integrating with a building's existing electronic access control system to grant delivery drivers immediate access to the property.

Here’s how the system works to make package delivery faster:

• Amazon installs a smart fob device that connects to the building's existing access control system

• When an Amazon delivery driver arrives, they request one-time access via their handheld device app

• Amazon's system verifies the driver's ID, route, GPS location, and time of request

• If verified, the system grants the driver temporary, time-limited access to enter the building

• The driver enters, delivers the packages to a secure location (e.g., lobby or package room), and exits

• After the delivery is complete, the driver cannot re-enter using the same credentials

Since drivers don’t get stuck waiting for residents (or doormen) to give them access to the building, they are able to make the delivery faster and more reliably.

Amazon Key for Business is available in many areas across the United States (although Amazon does not publish an actual coverage map).

For eligible properties (typically those with 10 or more units), Amazon provides free installation, hardware, and maintenance.

So you can see, this program has been a strategic infrastructure investment for Amazon since it was first introduced in January of 2019.

Why Is Immediate Access A Benefit For Last Mile Delivery

When it comes to eCommerce last mile deliveries, your biggest constraint is time.

Time is what determines your available capacity.

The more capacity you have, the more money you can make.

You might think then that the easiest way to make more money in last mile is to simply have people work more hours to get more deliveries done.

While (kind of) true, there are a host of service and safety issues that come into play the longer your shift goes.

Safety is the main reason that commercial drivers are not allowed to drive more than 13 hours per day.

When it comes to consumer last mile activity (other than the safety element), you are also faced with performance decreases and realistic delivery ranges if you try to push too far past 10 hours a day.

This is why the majority of last mile service providers work 10 hour shifts, with the smaller carriers or gigwork networks having people work 8 hours or less.

Appreciating that time is your capacity constraint, there are three ways to increase the amount of value-add activity (i.e. delivery time) that you have in your system.

1. Reduce as much as possible administrative activities. This happens primarily at the start and end of a driver’s day

2. Ensure your route optimization is solid and that you have the right profiles and information by customer & delivery activity

3. Decrease your average delivery time

Reducing your administration time is always important but usually ends up having a pretty short runway when it comes to the contribution it can make.

Route optimization (what most people thinks makes the biggest difference) can have a solid impact however can get pretty limited based on the service geography and density - there really are only so many ways to make deliveries to dense pockets of activity when you have multiple customers on the same streets and neighborhoods.

Decreasing your service time is often the best way to dramatically improve your costs (with the caveat that you need to have the additional volume to be able to push back into the routes that you have now created more capacity in). This is often easier to achieve in a B2B delivery structure than it is in B2C in North America since our B2C deliveries are (for the most part) drop and go.

This however is where a system like Key For Business can have a huge impact however - as it eliminates the non value-added wait time that occurs at these types of multi-resident locations.

Below are three graphs to illustrate the benefit someone like Amazon gets from this service.

For each of these scenarios, you have the same core variables:

• Same service area

• Same driver

• Same truck

• 10h shift

You can see that as you decrease the average delivery time across the route (which is exactly what Key For Business does), you end up making more money for the same investment over the same period of time.

This chart shows the impact of all 3 scenarios when it comes to the relationship between the stops and the additional contribution revenue.

Programs like this all contribute to why Amazon can consistently offer high service levels while also having industry leading costs.

But, these solutions didn’t come out of thin air.

Amazon has invested tons of money for these types of infrastructure advantages.

Is This A Smart Strategy For Amazon?

If you take a look at Jonathan’s post, you’ll see that my initial reaction to this information was that it doesn’t make sense for Amazon to give up the benefit and take on the risk that could happen from other carriers.

Having thought about it more, I’ve changed my mind.

I think that if Amazon gets what it would need for this to make sense (and I don’t mean money), it’s a pretty brilliant move forward.

Here’s why:

• Data
  Amazon Key requires Amazon to be able to validate the driver’s route, activity and delivery requirement. This means that Amazon would receive delivery plans from any other carrier that would look to want to use the service.
  
  Even with highly anonymized data, Amazon would get access to all kinds of insights it doesn’t have today (how many deliveries these carriers make, how long did it take them between Key access points, what are their min and max levels for activity … the list goes on)
  

• Own The Door
  With Amazon being a widespread provider, installing Key becomes even more attractive for property owners who can buy into one system for all their delivery access and don’t have to manage multiple systems (there are “competitors” on the market today")

But isn’t it a risk for Amazon to give up this strategic advantage?

I don’t think so.

Amazon knows that outside of USPS, FedEx and UPS, most regional (and especially) gig networks don’t have the levels of density that having access to buildings would dramatically increase their profitability.

(Remember, creating capacity that you then can’t fill doesn’t improve your revenue and profit streams)

In addition, Amazon would start collecting passive revenue from their “competitors” who once committed to the system would have a hard time pulling away and losing the access that it providers (because it would immediately push their costs back up).

Kuddos to Jonathan for getting the inside scoop on this - I believe we will be seeing more on this in 2025.

An Overview Of SKU Profiling For Inventory Slotting

With ever expanding sales SKUS, it can be hard to know if you have the best layout in your warehouse.

Today I’ll share some thoughts, guidelines and the general approaches to understand and classify all of the different SKUs in your portfolio and create a data analysis that can help you effectively communicate with different parts of your organization what is happening based on your sales or shipment data.

Core Data

Before getting into any analysis, you want to make sure of two things.

1. You need to get a large dataset of 1-2 years at minimum at a daily level

2. Ensure that you have enough “dimensions” to each SKU to be able to classify and segment them in different ways

Examples of Product Dimensions

• Size

• Weight

• Season

• Category

• Type

• Department

• Color

• Fabric

• MSRP

While some of these are generic and broad that would apply to a lot of businesses, find out if there are any particular properties that are important or unique to your product that should be included.

The general rule of data dimensions is to always take more than you think you might need (dropping columns after the fact is way easier and will save you time versus realizing you would have like to have had something later and having to re-run your data or join it back in).

Data Cleaning and Prep

Most people want to jump right into doing their analysis.

If your data is clean and ready to go, then by all means, jump in.

What happens more often than not however is that you need to make sure you have everything you need and it has been cleaned so that you will be able to get into a proper analysis later.

This is especially important if you are going to join in data from other sources later. You want to make sure the fields you will use to relate each data set are of the same type and structure (or if you had a situation like I just did - creating fields in one dataset in order to be able to join it to the other).

One thing I ALWAYS add in at this step is making sure to have all of the different date parts.

When people run data from their systems, they often just pull out a complete date (or date&time) field.

Even if I am not sure if I’ll need to use it, I make sure to extract or create the following date elements:

• Year

• Quarter

• Month (name)

• Month (number)

• Week Number

• Day of Week (name)

• Day of Week (number)

Creating Your Own Data Fields

The best stuff (or at least, things that will make your analysis easier) don’t often come out of canned systems.

The basics will, but the nuance that will take an analysis from good to great usually happen from adding additional elements to your data

(Protip … adding doesn’t not mean REMOVING or CHANGING core data. Adding means adding additional columns to work with, but never removing elements that link back or can balance to the host system)

Adding a few key elements early in the analysis to be able to keep things clean often saves you a lot of time. Here are examples of things you may want to add to your dataset:

1. Add “types” to transactions that had missing values. This makes it easy to understand later why a sku falls into that category without having to double check if it’s a mistake or why it didn’t go to somewhere else

2. Since you are working with data over time, create a few different types of “period keys” so that you can segment and summarize data in intervals and to ensure that data elements from one year to the next don’t aggregate
   
   Example:
   Month Key: YYYY-MM
   Week Key: YYYY-WW

3. Add a sales season flag to quickly align transactions for certain periods to seasonal buckets

4. Add a “source” column that can be used to filter the transactions by multiple types. This allows you to have fewer “master groups” and helps with overall readability of your data for someone else after the fact (when you need to explain it!)

Aggregating Data

Your ultimate goal when you are forecasting or doing a slotting analysis is to be able to get to a final table where each SKU is only represented once.

All of your unit figures (total sales, average sales, total orders, average orders, etc) should end up as column data so that you can get an understanding of how that product performs, when and how.

To do this, you need to think about what metrics or calculations make sense at a transaction level when you have all of your data, and when you need to start rolling up elements in order to make your calculations at the next level up.

Since this is a slotting analysis, first roll everything up by the weekly period key and then by SKU. This gives totals for units, orders and the average quantity for each SKU by order for that period.

The next important element is to understand how to determine your SKU classes. SKU classes help you understand the velocity of your sales and which products have what impact.

For a lot of product analyses, you can often get away with needing 3 classes:

• A - Your 80/20

• B - Sales from 80-95%

• C - Sales from 95-100%

This is done by aggregating all of you sales units, sorting them in a descending order, calculating their contribution percentage and then running a rolling sum of those percentages.

Make adjustments as required by your activity and product offering.

Now that you have your products aggregated to a point where you can see their impact over time, and know which SKUs contribute in which way to your overall sales activity, you are ready to start looking at where to put what SKUs and why.

(I’ll cover how I would approach that in a future newsletter)

That’s it for this week. Thanks for being here.