Last updated May 15, 2024
Sometimes it feels like the last two years have been a long drone of writing on the death of SaaS. It makes sense in the context of b2b SaaS saturation, corrected interest rates, and subsequent budget consolidation. There seems to be a more concerted focus on moving into “unsexy and undigitized” fields like supply chain.
Inherent in this claim is the assumption that supply chain is antiquated, and because it is unsexy, it has not received the attention of entrepreneurs that would lead to a wave of digitization. Given general consensus that SaaS has consumed the low hanging b2b problems to solve, it’s only a matter of time before the focus shifts to a field like supply chain.
I helped standup the supply chain at Bowery Farming, and then launched Mammoth to help food & beverage companies optimize their supply chain decision making, and what has become evident to me is that supply chain is most definitely unsexy, but the notion that it is undigitized is absolutely false, and even further, any apparent lack of SaaS penetration in the space is not for lack of trying.
Supply chain isn’t “undigitized” because it’s unloved. It’s not digitized because it’s extremely hard to digitize. More specifically, supply chain broadly suffers from a lack of data composability and structure. There is software for nearly every possible workflow you can imagine in supply chain, but most of these tools lack the ability to seamlessly pull and organize data from the myriad of other critical supply chain solutions. As result, there exist a plethora of balkanized software point solutions, that depending on the company, function as a half used system of record, that is then most often supplemented by excel.
Before going deeper, my main caveat here is that the below is influenced by my corner of supply chain, which is that of consumer packaged goods (CPG), and even more specifically food & beverage and agriculture. There are specific characteristics to food and beverage, such as shelf life and the dominance of grocery as a distribution channel, that make it somewhat unique, but I’ve found that almost every supply chain has its own quirks (part of what makes data standardization so challenging), but that despite these many distinct characteristics, the high level activities one must do to create a functioning supply chain are fairly common.
The below focuses on software solutions for companies that sell or produce physical goods, often leveraging a network of partners to either supplier ingredients, manufacture the products, distribute the products, and sell the products.
Depending on the vertical and relative complexity, supply chains can be fairly unique, but most can be understood by a few processes and activities:
Demand Plan: Every company needs to forecast what they believe demand for their products will be. Most companies will do this on a yearly basis, adjust quarterly, and even monthly depending on production timelines. A demand plan factors in historical sales, expected new customers or channels, promotion schedules, seasonality, product launches, and several other more nuanced factors. A good demand plan is broken down by the product (SKU) level and by the customer (e.g. Walmart) and channel (e.g. grocery). An accurate demand plan is the tip of the supply chain spear, and without a predictable margin of error, all other elements of supply chain planning become tenuous. A demand plan typically comes from the sales & marketing team.
Supply Plan: Based on the demand plan, a supply chain team will need to create a supply plan. A supply plan asks the question, in order to fulfill 100% of my demand plan:
Making this supply plan is often a large undertaking, and requires a significant orchestration of information across the organization, and depending on the supply chain, outside of the organization as well.
Inventory Tracking: Inventory is simply the snapshot of the quantity and cost of a company’s finished products and raw materials by location (and in food, shelf life is also a significant factor). However, tracking inventory is not simple. At its most vertically integrated, inventory tracking is a question of converting on the ground items into information in a database. At its least vertically integrated, it requires collecting data from a variety of partners, such as co-manufacturers, co-packers, third party logistics providers, and even suppliers if they are holding raw material inventory for you. Understanding inventory is a key input to the development of the supply plan. Knowing when and where a raw material becomes or will become a finished product significantly impacts an operator's understanding of supply chain resiliency, efficiency, and cost.
Manufacturing Plan: A ****manufacturing plan can be rather complex, but put simply, it is an understanding or commitment of average throughput for each product over a period of time. This is dependent on inputs such as staffing levels, machinery downtime, maintenance, packing rates, and most importantly yield (quantity of input necessary for 1 unit of output). For companies that work with co-manufacturers and co-packers, this is fairly straightforward, as throughput is essentially a contract term. Regardless of the arrangement, every supply plan has built in assumptions about how much of a product can be produced by when and with what materials.
Material Resources Plan: Based on inventory and the manufacturing plan, which materials do you need to procure, and when do you need to order those materials so that they arrive on time? This is the final step of the supply plan, but is the first step in actioning the supply plan. Depending on how a supply chain is structured, a Material Resource Plan may involve optimizing which suppliers to order from so as to lower cost, or it may involve expanding the supplier or contract base to ensure coverage of the necessary raw material supply.
The above is a significant simplification of the core processes and actions in supply chain planning, and does not even scratch the surface as to the specific data that goes into these plans. Data inputs such as Bill of Materials (the recipe for each product, including packaging), minimum order quantities (MOQ), shelf life, lead time, shipping terms, truck load quantity, case load quantity, and more, all factor into the analysis required above.
All of this is orchestrated by the Sales & Operations Planning (S&OP) process. This process typically looks like:
In essence, the perfect S&OP process is able to create a single representation of a company’s supply chain end to end. This process is a constant negotiation and trading of data, with each team acting in accordance with its own incentives and trying to underpromise so that they can overdeliver. For example, the sales team often pads forecasts to ensure enough supply exists but not so much that they underperform closing the sale. Supply chain teams can underestimate production capacity in order to justify holding more inventory to ensure they hit the demand plan. Finance teams push for running a leaner inventory cycle to reduce costs and minimize cash outlays. Even suppliers partake in this dance, given that they face many of the same production challenges. And lastly, retailers (end customers) pad their estimates as well, so as to minimize out of stocks, and back charge their customers (cpg brands) for their own overestimates.
This is all to say that supply chain planning is about trying to predict the future as accurately as possible, assessing the validity of the inputs as best as one can, executing on an agreed upon plan, and then measuring the resulting performance against the expectations set forth by the supply plan. In order to do this properly, an immense amount of data must be handled from a wide variety of sources.
This is not only a problem for the Fortune 100, but almost any company making a physical product that has not completely outsourced their production and sourcing to a turnkey solution provider.
At every single point of the S&OP processes, there is internal company data and external third party data (e.g. inventory from a 3PL, velocity data from Nielsen), and the efficacy of the supply chain is dependent on the wrangling of this data.
So where does this data live? Surprisingly, a lot of it actually does live in software. There is plenty of software intended to meet the needs of the processes above.
The issue is that the software solutions rarely communicate with each other, and are often so customized to each customer that building robust and scalable integrations is not worth the effort.
One only has to look at the most successful supply chain software solutions to find the evidence. In an industry where software fails to communicate seamlessly, the natural result is a solution set largely built around services. When software solutions are heavily services based, the customer base naturally migrates toward a set of companies where massive and expensive integration and data cleaning periods make economic sense.
The most successful supply chain software solutions - SCM (o9, Kinaxis), ERP (SAP, Netsuite) WMS (Manhattan Associates, Deposco) - all have robust, 3rd party integration service ecosystems (consulting companies) built around them. It is no wonder that these products are extremely expensive ($250k-$1M ACV), and typically service large international enterprises.
It is also no surprise that this model is untenable for most companies that need supply chain software. Even those operating in the $100M+ revenue range still resort to begrudgingly installing an ERP at some point but running most S&OP activities in excel.
As a user of many supply chain software platforms and a builder of one as well, it is evident that so much of the substandard user experience of current offerings stems from the fact that almost all solutions require significant manual data cleansing and structuring while also attempting to become core systems of record themselves.
Supply chain needs its own universal API, one that will connect the myriad of software solutions across each supply chain vertical, and allow for the conversion of core company data into key real world supply chain concepts.
A universal supply chain API will allow current solutions to better deliver on their customer promise. Similar to how Plaid provided the foundation for the explosion in consumer fintech applications, the universal API for supply chain should also enable a host of better, verticalized, point SaaS solutions at more competitive prices for the long tail of medium sized businesses that produce physical goods and power our economy.
This will not be easy. Many of the dominant software solutions are old, and can be highly customized. Further, most supply chains have a few unique data elements to them, such that sometimes the software used to store this data is not used for its intended purpose, making extracting and standardizing the data difficult.
The key, I believe, is starting with one type of supply chain software solution, saturating that market, and moving to the next vertical. To understand this, we have to lay out the landscape of supply chain software in more detail to clarify how vast the flow of data and level of connectivity really must be. ****
There are 15+ discrete categories of supply chain software. Many of the largest players move across categories, and many of the start-ups in this space are picking off bits and pieces of each category and sometimes inventing new ones. If you’re very familiar with the supply chain software landscape, skip this part. If not:
https://lh7-us.googleusercontent.com/Wde9QKNlnXKPMtCUqEpoBGGh1suExv92sjMEO-61PX3eDbP8sALrEKe7lroPlT6tl27aLEdpq3KTUCtDLAE-LZqKudDyZOEUVpcJ3u1P8Ii8NOIIgz4Y1QJa7Admzn3U5ZQ-8OS4HtD3zazsD91WHSc