DTC Fresh Meat Makes a Great Case for Automation

The chicken drumstick is a family favorite, and there is tremendous excitement when a plate lands up on the table - glistening red spicy exterior, charred in some places, a nice aluminum foil wrapped leg and some sweet-sour sauce. Before you know it, the plate is clean. In most cases, we appreciate the chef or our mom and move on. But sometimes it is fun to ask what goes on behind the scenes to put such a delicious plate on the table? We will not talk about the making process (unfortunately!) but talk a little bit about the supply chain.
Buying meat, fish is not everyone’s cup of tea, particularly in India. The butchery (or the fish market) may not be a very hygienic place, you may see cats lurking around, and there is a strong smell you can’t shake off quickly. Yes, many love this process, they have their techniques to check if the produce is fresh etc. but for an average person on the road, it’s a hassle.
Access to good meat is hard, and it has become a little harder these days. Why? As consumers, we are not only interested in the product but also want to know the source of the produce, ensure it is chemical-free and of course, ensure it is fresh. Yes, the consumers are demanding, and the market is complicated. The good news is that many brands are trying to solve this problem, see Licious, FreshToHome, Fipola, Tendercuts and many more - they deliver a pack of a ready-to-cook chicken drumstick on the same day or the next day.
These direct-to-home (DTC) fresh meat brands are full-stack operators; they manage the customer acquisition, delivery (optional) and operate an integrated intensive cold supply chain. The brand has to get quality and consistency (for product and fulfilment experience) right every day while also ensuring the operations are capital-efficient (working capital).

Direct-to-Consumer Meat Hub-and-Spoke Model
The hub-and-spoke supply chain of DTC meat

The supply chain is some variation of a hub-and-spoke model. Every day, supply chain operators make thousands of decisions, and all these have to work in harmony at all the nodes. It is hard. Let’s see why it is so and why meat DTC brands have to automate for the future.

Product Complexity

First up is the shelf life of the produce. The short (actually very short) shelf life makes the fresh meat supply chain very challenging - fresh meat, seafood has a shelf life of two days, ready-to-cook three days, and cold cuts four days. The supply chain planners have to make accurate predictions every day; otherwise, there will be a lot of wastage. In the case of meat, there is no way to salvage the waste (it is not edible there might be other uses). Out of the 150-160 SKUs, brands approximately keep 60-70% available for same-day delivery (these are generally fast movers) while the balance is made-to-order (more exotic preparations).

Approximate SKU distribution of DTC Meat brand
Approximate SKU distribution of DTC Meat brand

The SKUs level predictions from the spokes/fulfilment centre (let’s call the fulfilment centre a store for easy reference) are aggregated at the hub and herein lies the next complexity. On average, a single bird/animal can yield 9-12 different SKUs (2-3 SKUs in case of seafood). There is no way to use half of a bird/animal; any mismatch in raw material and SKU predictions can result in a lot of wastage. Optimizing the raw material quantity to meet the demand is a complex problem; this further impacts the sourcing plans. So getting SKU level predictions right is crucial for the fresh meat supply chain. One way brands handle the problem of excess raw material is to develop longer shelf-life products such as pickles, spreads, but these tend to be long-tail products and not core revenue drivers.

Market Complexity

That of the market matches the complexity of the product. Many factors influence demand; here are some sample cases:

  • Day of the week due to varying cultural preference, e.g. some avoid meat on Tuesday’s, some on Thursdays etc.
  • Week of the month, e.g. consumption increases on month-end or beginning
  • The month of the year, e.g. festivals across different states - demand for meat falls during Christmas, through New Year, surprising?
  • Predictable events - holidays, which might be culture related or just national/state holidays
  • Unpredictable events - rains

The demand variation due to above factors changes across cities, stores for the brands - the demand for the same SKU might have different trajectories across two stores in the same town. So the stocking of stores has to be done by estimating demand for that store and adding buffer stock.

Demand Probability and Cost Cure for each Store-SKU combination
Determine the optimum stock for a Store-SKU combination, balance business objectives

The demand planners have to factor not only the seasonality, trend but also factors many other factors in their forecast for the next and next week at the store level. What more? These predictions have to in harmony across stores at the hub. Forecast more and stock more; you might end up with a lot of wastage, loss of significant working capital; Forecast less and stock less, you will lose sales, result in poor customer experience. The further away you are in the supply chain, the impact is that much severe.

Marketing Interventions

Marketing intervention spice things a little more. Customer acquisition cost accounts of a sizable part of DTC cost structure. Same applies to the fresh meat companies as well. The product and marketing teams work on several initiates to acquire and retain customers, for example:

  • Acquisition - flat promotional discount for SKUs to nudge purchase, free delivery within a specific area, the amount of discount will vary by region, hub, store etc.
  • Retention - targeted discounts to avoid customer churn, e.g. higher discount on frequently ordered SKU
  • Clearance - reduce the inventory by offering a higher discount
  • Many more to engage the customer and keep them hooked.

The interventions listed above influence the demand and have an impact on the forecast and demand plans for the next day, week. Each promotion can impact sales differently and needs to be evaluated in-depth. The marketing team has to collaborate with supply chain to assess the impact of demand shaping initiatives at the granular level, e.g. Store-SKU, Hub-SKU to get better ROI.

Decisions, Data and Automation

The factors listed above barely scratch the surface; there are more variables at play, and a lot of this knowledge and patterns are hidden in the existing data. Now image a team two or three planners trying to pull this data from their ERP and trying to build models in spreadsheets. It is tedious, spreadsheet slow down tremendously with data and cross sheet linkages. The planners have to now make those thousand predictions/forecasts to keep the operations ticking every day.

Handling such a diverse dataset manually, and building the forecasts by hand can be time-consuming, and error-prone. Errors that go unnoticed can result in significant over-stocking or under-stocking, and either is not suitable for business. The tribal knowledge that exists in the team works well for a particular scale beyond which it needs to be digitized. It might be better to automate the process let the planners focus on more strategic initiatives and insights.

Things are changing, new techniques such as machine learning can generate granular SKU-Store level forecast, accommodate a range of other factors such as promotions, events, holidays and anomalies. All this without requiring an army of analysts or experts looking at data. The algorithms will crunch the data and understand the hidden patterns that were not apparent to humans previously.

Advanced analytics enable the planners to segment the portfolio and unearth opportunities in sub-categories, customer sub-segments and localities. Usually, 20-25 SKUs drive 80% of the sales, so there are might many hidden opportunities to further penetrate the market with other SKUs.

For fresh meat DTC, the frequency of decision-making is so high and product so complex that automation is the only way forward. Of course, there will always be human-in-loop to spot check the predictions and make necessary adjustments. But that should be an exception and not the norm. Automation enables planners to work on more customer-centric initiatives than data munging.

Build vs Buy the Technology

Not an easy choice. Out of the gate, a brand would want to build such capabilities in-house. If technology is core operations, and there are enough internal champions who can build a team and product around data, then why not? By all means, build.

But for most brands, technology is an enabler and not the core. The secret sauce is focusing on sourcing partnerships, improving the quality of the product, refine customer acquisition and ensure fast delivery. So, it may be prudent to buy technology and focus on the core. The idea is to start small, build the data warehouse, maybe roll out automated planning for a couple of stores, or categories and gradually expand the scope to cover operations. SaaS solutions such as Algoshelf allows you to follow that path, and there are no upfront IT investments - pay-as-you-go based on the SKU breadth.

In Closing

The DTC meat brands are passionate about their product and have to put in a lot of hard work, data and technology to put a plate of chicken on the table. The brands balancing the business objectives and running an efficient supply chain need a robust technology backbone. Investments in automation will enable them to grow sustainably and serve the customers better.

Love to hear your comments and feedback, please write to raghav[at] Happy to speak with planners and analysts working on supply chain machine learning.

Image Credits:  Atharva Tulsi on Unsplash