About Mike Doherty

Mike Doherty

Compliments from a CEO

It’s always nice and also encouraging to hear positive comments from a retail CEO about Flowcasting and supply chain management in general. Below are some nice compliments from Ken Larson, President & CEO of Princess Auto Ltd, about Flowcasting, supply chain and the team (both internal and external) that helped make it happen. Thanks for the kind words Ken!

Working backwards

You’d likely agree that Amazon is a very innovative company.  AWS, Prime, Kindle, One-click-shopping are examples of innovations they’ve delivered through a process called Working Backwards.  Working Backwards is a systemic way to surface and vet ideas that sometimes lead to new products or services.  The key tenant of the approach is to begin by defining the product’s (or service) arrival, and then work backwards from that point until the teams achieve clarity of thought around what they will deliver.  

The main tool used to accomplish this clarity of thought is the PR/FAQ document – short for Press Release / Frequently Asked Questions.  The document is written from the perspective of the customer and is dated into the future, when they believe the innovation will arrive in the marketplace.  What’s simple, yet brilliant, about this approach is that it keeps teams focused on delivering something significant, valuable and meaningful to customers – rather than many approaches that are essentially based on the idea of incrementing-your-way-to-greatness.

You may not know, but Flowcasting is also based on the concept of Working Backwards.

Loyal disciples understand that the Flowcasting process starts with a forecast of consumer demand, by product and selling location (e.g., store, etc.).  Conceptually, what’s a forecast of consumer demand?  It’s when and where the customer wants to acquire the product – i.e., when and where it needs to arrive in their hands.

Flowcasting integrates the supply chain from this forecast of consumer demand by working backwards using an approach called arrival based planning.  For any product at any location, arrival based planning calculates planned shipments of when and how much inventory needs to arrive – scheduling a shipment so that the arrival date and quantity ensures the projected inventory doesn’t drop below safety stock.  

The basic logic is simple:

For each item/location:
1. Figure out what you expect to sell (or calculate what you’ll ship) 
2. Figure out what you have and what your constraints are
3. Figure out when you will need more product to arrive
4. Use the arrival based plan to figure out when you need to ship

The concept of planned shipments and arrival based planning is foreign to most retail demand and supply planners. After we do an educational session with planners, we often break people into teams and ask them to calculate a shipment based plan, using arrival based logic, with an example something like this:

Without exception, every retail client we’ve worked with struggles with this simple ask. Everyone has been so conditioned to focus on “ordering” that they struggle to understand that ordering (or committing the planned shipment) is the last decision, not the first. They are focused on “do we need to order” in week 1, rather than planning arrivals and the corresponding ship dates (i.e., planned shipments) first. 

Planned shipments are the foundational construct of Flowcasting and how the process integrates the retail supply chain to ensure the fundamental principle is achieved – that is, a valid simulation of reality.  Every planned shipment consists of an arrival date (when the product will arrive at the destination) and a ship date (when the product will ship from the supplying location).

Flowcasting plans shipments over the entire planning horizon and shares these projections with suppliers and other stakeholders in order to help them plan and improve productivity.  For suppliers we often refer to what is shared with them as a supplier schedule – that is, the projection of planned quantities and their associated ship dates by product and origin/destination.  

Ship dates are the key element to anchor the supplier schedule on.  That’s because ship dates are very specific from a supply chain and work perspective.  It’s when the inventory needs to be available and also when transportation needs to be scheduled to initiate the delivery.

The supplier schedule is, in most cases, the culmination of Working Backwards for a retailer who is using Flowcasting.  From the forecast of consumer demand, we work backwards to calculate when product needs to arrive and ship, through the entire distribution network, right back to when the factory needs to ship the product – integrating the entire supply chain via a series of dependent, cascaded planned shipments.

When you first hear the phrase Working Backwards it sounds like a dumb idea and implies you’re heading in the wrong direction.  Turns out, it’s been a proven and brilliant way to drive innovation.  

It also turns out to be the best way to plan inventory flows for the retail supply chain.

The Arrival Based Plan
For those students, here is the arrival based plan for the example outlined above:

What’s important to understand is that the planned arrivals that are calculated above would be the same, regardless of supply source.  They indicate the quantity and timing of a shipments arrival and are not dependent on the supply source.  The planned shipment, of course, is dependent on the supplying source and the transit lead time.

This make arrival based planning a powerful approach since it enables you to easily plan a change of source well into the future and have all the cascaded dependent demand reflect a valid simulation of reality.

Trust the process

Nick Saban is a brilliant college football coach and widely heralded as a football genius. At time of writing, his coaching record stands at 261 wins, 65 losses and 1 tie. He’s won 17 Bowl games along with 7 national titles (the most ever) and counting.

If you’re a fan of the Alabama Crimson Tide, in a state where college football is king, Saban is arguably more popular than Jesus. Fanatics think he is Jesus.

When asked about his unrivaled success, Saban offers a counter-intuitive philosophy that’s guided his coaching career, anchored on two fundamental principles:

  1. Don’t focus on the outcome
  2. Trust the process

Incredibly, Saban doesn’t focus on the result – which, for college football – like most sports – is whether you win or lose. Sure, make no mistake about it, Saban wants to win, it’s just he believes that the path to long term success is by trusting the coaching process.

His belief is that by coaching his team, repeatedly and without fail, so that they can execute their designed plays on offense and coaching schemes on defense, he’ll maximize his odds of winning. Losses are important to this process as it provides the team the opportunity to review the film and see where the process failed – either in execution or, sometimes, in coaching and design. Which leads to more coaching, practice and trust in the process.

Trusting the process is a philosophy that has served Saban and the Crimson Tide incredibly well.

We can learn a lot from Nick’s nuggets of wisdom.

Given that we’re often referred to as the Salty Old Sea Dogs of Flowcasting, it’s fair to say that we’ve been around the block a few times and, over time, changed our thinking often. No more than so than in developing, managing and measuring the retail forecasting/demand planning process.

For most retailers, the number of item/store products that sell less than 26 units a year (about 1 every two weeks) can be pretty significant – often comprising 30-50% of a retailer’s assortment. You wouldn’t expect to be as accurate in determining a forecast for these types of products, since there is a fairly large element of probability involved – as an example, based on history, you can feel confident that 1 unit sells every month but you’re not sure when it will be.

While it’s tempting to aggregate the lower level item/store forecasts up to a higher level and assess forecast performance, that’s of little use to anyone – after all, customers buy products in stores, or online, to be acquired or delivered at their preferred location. They don’t buy them at some aggregate level. Not to mention that item/store replenishment plans are driven from these forecasts and the dependent demand is cascaded throughout the supply network.

Like Saban, we work with our clients to help them understand and hopefully instill the idea of assessing the forecasting process by determining something called forecast reasonableness.

So then, what’s a reasonable forecast?

The following diagram outlines, conceptually, what we’re talking about:

The idea is to assess the reasonableness of the forecasts based on a sliding scale determined by selling rate. As an example, you wouldn’t expect to have as accurate a forecast for a product that sold 12 units a year as you would for something that sold 1200 units a year. Of course, you need to determine what’s a reasonable tolerance and sliding scale but from experience that’s not too difficult.

If the item/store forecast is within tolerance then the process/solution is producing a reasonable forecast. The beautiful thing is that the planner spends no time chasing these forecasts since, in all likelihood very little can be done to improve the process/outcome for these.

In practice, forecast reasonableness is an exception condition for the demand planners to action. For the item/store forecasts that are outside of tolerance, the planner can investigate these, see if the same item is outside of tolerance for a number of stores and determine if anything is systemically happening or could be improved in the process to bring them within tolerance.

To determine how well the forecasting process is working is simple. What percentage of the item/store forecasts is within tolerance?

We believe that, in retail, forecast reasonableness should replace traditional measures of forecast accuracy (like MAPE, WMAPE, etc., that were developed for more continuous demand streams in manufacturing and distribution).

Now before you think we’re completely mental, here’s something to ponder on. One of our retail clients, who plan using the Flowcasting process, does not measure baseline forecast accuracy.

Instead, they use a forecast reasonableness exception to evaluate the process, honing in on any forecasts that are not within tolerance to see if there is anything that can explain this and, potentially, how they might “improve the process”.
During the time they’ve not measured forecast accuracy they improved daily in-stock from an average of 91.7% to an average of 97.7%, while also improving inventory performance.

For a customer, in-stock is everything. They couldn’t care less about forecast accuracy.

Maybe you shouldn’t either.

Wrong becomes right

Phil Tetlock spent almost two decades determining people’s ability to forecast specific events – things like elections and the outcomes of potential geopolitical decisions. Unfortunately, the results were not impressive. Most people were about as accurate as a well-fed, dart-throwing chimpanzee.

There were a small number of notable exceptions. A small group of people consistently provided quite accurate forecasts – people he aptly named “super-forecasters”.

In a subsequent forecasting tournament organized by the Intelligence Advanced Research Projects Activity (IARPA) – a focused branch of the United States intelligence community, the super-forecasters trounced teams of professors and “forecasting experts” by wide margins.

What made the super-forecasters so super?

It wasn’t intelligence or that they had more experience than others. In fact, in many cases, they were mostly amateurs yet they outperformed the CIA’s best and brightest (who also had the advantage of years of experience and classified information). Armed with only Google, the super-forecasters beat the CIA, on average, by 30%.

What made them great at being right was they were great at being wrong!

The difference in their ability to forecast was simple, yet crucial. The super-forecasters changed their minds – a lot.

Not a huge, 180-degree shift, but subtle revisions to their predictions as they learned new information. As an example, one of the consistently top super-forecasters would routinely change his mind at least a dozen times on a prediction and, sometimes, as often as forty or fifty times.

Importantly, most viewed a revised forecast based on new information not as changing an initially wrong forecast but rather as updating it. Turns out, updating is the secret to being a great, or super, forecaster.

The concept of updating is important in Flowcasting as well.

As loyal Flowcasters know, an important component of the process is the sharing of what we call a supplier schedule – that is, a projection, by item and delivery location of how many units are needed to ship over a long time horizon (typically 52+ weeks).

If the schedule indicates that, 39 weeks from now, the supplier will need to ship 10440 units of a product to a location, what’s the chance that this projection (i.e., forecast) is perfectly accurate? Pretty low, right? And it doesn’t need to be – it just needs to be reasonable.

A week later, and guess what? The supplier schedule has been re-calculated and updated to indicate that 10400 units are needed to ship in that particular week (perhaps even on a different day). The updated forecast is more right than the previous one. This process of minor revisions continues as the projections are updated until it’s actually time to ship (i.e., when the planned shipment has reached the agreed-upon order release horizon).

In retail, supplier scheduling is the super-forecaster for suppliers – recalibrating and subtly updating the forward looking projections, based on the latest information until…

Wrong becomes right.

Flipping your thinking

From toddler to teenager, most of us had a fairly similar educational experience: The teacher would stand at the front of the room and spew out information that the students were to absorb. Then they would assign homework that would, theoretically at least, test how well you absorbed the content. The homework assignments would then be scored by the teacher and – if you were lucky – you might find some scrawled notes in the margin to give you a clue as to where you may have gone astray.

All this assumes, of course, that you didn’t just copy your homework from a smart (if gullible) friend, thereby completely circumventing the ability of the homework assignment to test knowledge. Not to say that this approach was completely ineffective. Somehow, most of us did learn what we needed to know to become productive members of society. That said, when the best praise you can muster is that it’s “not completely ineffective”, then you are basically admitting that there is ample room for improvement. 

Karl Fish is a veteran teacher with 20 years of experience. He teaches high school math in a town just south of Denver, Colorado.  Karl thought he could do better for his students than “not completely ineffective”, so he decided to flip the traditional thinking on its head.

Instead of using class-time to “teach” in the traditional sense, Karl tapes his lessons and uploads them to YouTube. Classroom time is used for application and practice.  His students are required to watch the lecture whenever and wherever it is most convenient for them.  What would be traditionally considered “homework” is actually done during in-class time.

The result of this flip in thinking has been significantly improved understanding of the content. Working through examples and case studies not only improves the students understanding, but also improves their ability to collaborate with fellow students and Karl himself.

Contrast this approach with doing your homework in the evenings (maybe even the wee hours of the morning). If you get stuck, there are few alternatives other than to become increasingly frustrated and demoralized.

In Karl’s class, you can pose your question to a group (many of whom are likely struggling with the same question) and work together to solve the problem. What better skill and habits are there for someone to learn in high school?

Retail supply chain planning is also in need of a flip in thinking.

Traditional thinking in retail has ingrained into people’s heads that ordering is the key decision a supply chain planner needs to make. Day in and day out the retail supply chain planner only has 2 questions:

1) Should I order today?

2) How much should I order?

All anyone can talk about nowadays is “demand driven supply chains” that are super-responsive to consumer demand – yet the entire planning approach is geared toward figuring out when to place an order upstream with, at best, an indirect link to the actual customer demand.

Flowcasting flips that thinking completely. In a nutshell, the decision to place an order is a mere after-effect of the planning process (not the one and only decision) and, generally, it should be performed by a computer, not a human being.

Instead of asking “When and how much should I order”, maybe the first question should be “When does more stock need to arrive?” – Isn’t that what’s really important to your ability to stay in stock and serve a customer?

By focusing on this question first, new thinking emerges. Once you understand when product needs to arrive, then you can calculate when you need product to ship (based on where it’s coming from) and when you need to order.

Of course, to answer this question you’ll need a system to project inventory and product arrival dates well into the future.  And to calculate the corresponding ship dates and order dates. The basic change in philosophy is simple – to really know when product needs to be ordered, you must first know when it needs to arrive at the destination.

While this may sound quite simple and logical, in practice it requires a great deal of education and understanding to make such a flip – particularly if people have been “ordering” for a long time.

In our experience, most retailers continue to subscribe to the “order first, ask questions later” philosophy and flipping to an arrival based planning approach like Flowcasting will not be a slam dunk, regardless of how logical it sounds.

If you’re struggling with the concept, we’d be happy to schedule some homework time with you and your team.

Principles

Folks that know us well and have worked with us, know we’re what you might call principles freaks.  A decent chunk of time and effort we spend helping retailers implement Flowcasting is oriented to instilling a set of principles, which guide our thinking and, hopefully over time, our clients.

Julia Galef, in her brilliant book, The Scout Mindset, beautifully outlines the paradox of principles in a chapter aptly named, “How To Be Wrong”…

Many principles sound obvious and that you know them already.  But “knowing” a principle, in the sense that you read it and say, “Yes, I know that,” is different from having internalized it in a way that actually changes how you think.

Times Ten

It’s hard to believe and counter-intuitive but it’s often easier to make something 10 times better than it is to make it 10 percent better.

Yes, really.

That’s because when you’re working to make things a bit better – like 10 percent better – you always start with the existing tools, assumptions and paradigms, and focus on tweaking an existing approach/solution that’s become the accepted norm.

This kind of progress is driven by extra effort, extra money, and extra resources – attempting to squeeze just a wee bit more from the current system. While making a minor improvement is generally a good thing, often we find ourselves stuck in the same old paradigms.

But when you aim for a 10 times improvement, you need to lean in and focus on being brave and creative – the kind of thinking that, literally and metaphorically, can put a man on Mars.

Retail Flowcasting solutions have been emerging for a while with some technology companies claiming that Flowcasting is in the DNA of their solution. More often than not that’s debatable but what’s not is that a retail Flowcasting solution needs to be at least 10 times more scalable and simpler than similar solutions that have been used in distribution and manufacturing – because in retail we’re dealing with 10’s of millions of item/location combinations.

We’ve been lucky enough to implement a purpose-built, times-ten retail Flowcasting solution. In our experience, it’s 10 times faster, 10 times simpler, and costs about 10 times less to implement compared to others who claim to have Flowcasting native solutions.

Over the course of the implementation, the chief architect shared with us how he was able to build something that’s a times-ten solution.

His secret:

  1. Start from scratch
  2. Solve the hard problems first

Most existing solutions are too heavy and burdensome – the result of solution providers not knowing how to say “no” to a specific ask – usually from a paying client that doesn’t know better, but also sometimes from competitive peer pressure. Checklist Charlie said our solution needs 11 different ways to handle safety stock, so we’ll make it so. The result is a system that cannot be the foundational starting point to build a times-ten solution.

Our architectural hero said the reason he started from scratch is that only what was necessary would go into the solution and not a drop more. A bit like Einstein once said, “everything should be made as simple as possible, but no simpler”.

In tandem with making things as simple as possible, our colleague believes that another reason he was successful is that he also solved the hard problems first.

To wit, he worked, tested and developed stunningly intuitive and simple ways to handle a variety of problems that have plagued existing providers from morphing their current solutions to a retail-focused Flowcasting solution, notably:

  1. Extreme scalability
  2. Handling slow selling items properly (both the forecast and the replenishment plans)
  3. Planning for product phase out (to provide a valid simulation of reality for any item at any location)
  4. Managing seasonal items (to maximize sales and minimize inventory carryover)

I won’t bore you with the specifics on how our architectural and design colleague solved each of these chronic retail planning challenges. He did say that he’d spoken to a number of grassroots folks in retail about each of these challenges and then – from a blank sheet of paper and some fundamental principles – developed, tested, refined and eventually delivered a times-ten solution.

Once the hard problems were solved, he believed that the rest of the planning challenges would be chicken-shit to deal with. Turns out, he was right.

The lesson here is simple, yet profound. If you’re looking for a quantum leap improvement – a times-ten solution if you will – then you need to start from scratch and solve the hard stuff first.

Subtraction

“What gets left out is as important as what gets put in.”

 Steve Jobs philosophy

Subtraction

I have a good friend named April, who has the coolest son, Lincoln.  Lincoln is in grade one and looking forward to grade two.  When I asked him his favorite subject, he declared “math”.  So, probing further, I asked him what he liked about math…

“Subtraction”.

“I like taking things away and seeing what’s left”.

I was startled. I’d never heard anyone claim their love of subtraction.  Or taking things away from something and seeing what’s left.  Of course, it got me thinking.

Flowcasting, as a concept and business process, is gaining lots of momentum.  People are really beginning to understand that, when it comes to supply chain planning, most companies have been planning looking in the rear view mirror.  Most are forecasting what should be calculated.

With Flowcasting, you only need to forecast at the item/store level and calculate everything else – all demand, supply, inventory, capacity, financial projections can be determined using this forecast.

The legendary Steve Jobs is a disciple of “subtraction thinking” – his colleagues claim that Steve was more interested in what gets left out versus what gets put in when designing a new product.

Flowcasting systems are systems cut from the same cloth. 

Think about today’s planning systems and all their features.  All kinds of bells and whistles – you know the drill, a different algorithm for each type of demand stream…special systems to help you decide on the algorithm…multiple forecasts and automatically picking the best one…a different planning system for slow selling items…a special system for allocation……etc… On and on it goes.

For retailers, these systems were built for a world that will someday no longer exist.  A world where we were forecasting at the wrong level – DC’s, plants, etc. – anywhere other than the point of final consumption.

Technology companies are also beginning to understand and embrace the principles of Flowcasting.

Hopefully they’ll embrace the concept of “subtraction”.  Instead of adding to their systems, they should be subtracting. 

For the retail supply chain, forecasting is at the store level and only there.  Simple approaches are all that’s required to forecast at store level and further remind yourself that all other demands can and should be calculated.

Are there systems available that support the Flowcasting process?  Absolutely.  They are elegant, simple and were built with “subtraction thinking” in mind.

It’s amazing what you can learn talking with a first-grader.

I plan on talking to Lincoln again soon. 

Changing the game

In 1972, for my 10th birthday, my Mom would buy me a wooden chess set and a chess book to teach me the basics of the game.  Shortly after, I’d become hooked and the timing was perfect as it coincided with Bobby Fischer’s ascendency in September 1972 to chess immortality – becoming the 11th World Champion.

As a chess aficionado, I was recently intrigued by a new and different chess book, Game Changer, by International Grandmaster Matthew Sadler and International Master Natasha Regan.

The book chronicles the evolution and rise of computer chess super-grandmaster AlphaZero – a completely new chess algorithm developed by British artificial intelligence (AI) company DeepMind.

Until the emergence of AlphaZero, the king of chess algorithms was Stockfish.  Stockfish was architected by providing the engine the entire library of recorded grandmaster games, along with the entire library of chess openings, middle game tactics and endgames.  It would rely on this incredible database of chess knowledge and it’s monstrous computational abilities.

And, the approach worked.  Stockfish was the king of chess machines and its official chess rating of around 3200 is higher than any human in history.  In short, a match between current World Champion Magnus Carlsen and Stockfish would see the machine win every time.

Enter AlphaZero.  What’s intriguing and instructive about AlphaZero is that the developers took a completely different approach to enabling its chess knowledge.  The approach would use machine learning.

Rather than try to provide the sum total of chess knowledge to the engine, all that was provided were the rules of the game.

AlphaZero would be architected by learning from examples, rather than drawing on pre-specified human expert knowledge.  The basic approach is that the machine learning algorithm analyzes a position and determines move probabilities for each possible move to assess the strongest move.

And where did it get examples from which to learn?  By playing itself, repeatedly. Over the course of 9 hours, AlphaZero played 44 million games against itself – during which it continuously learned and adjusted the parameters of its machine learning neural network.

In 2017 AlphaZero would play a 100 game match against Stockfish and the match would result in a comprehensive victory for AlphaZero.  Imagine, a chess algorithm, architected based on a probabilistic machine learning approach would teach itself how to play and then smash the then algorithmic world champion!

What was even more impressive to the plethora of interested grandmasters was the manner in which AlphaZero played.  It played like a human, like the great attacking players of all time – a more precise version of Tal, Kasparov, and Spassky, complete with pawn and piece sacrifices to gain the initiative.

The AlphaZero story is very instructive for us supply chain planners and retail Flowcasters in particular.

As loyal disciples know, retail Flowcasting requires the calculation of millions of item/store forecasts – a staggering number.  Not surprisingly, people cannot manage that number of forecasts and even attempting to manage by exception is proving to have its limits.

What’s emerging, and is consistent with the AlphaZero story and learning, is that algorithms (either machine learning or a unified model approach) can shoulder the burden of grinding through and developing item/store specific baseline forecasts of sales, with little to no human touch required.

If you think about it, it’s not as far-fetched as you might think.  It will facilitate a game changing paradigm shift in demand planning.

First, it will relieve the burden of demand planners from learning and understanding different algorithms and approaches for developing a reasonable baseline forecast. Keep in mind that I said a reasonable forecast.  When we work with retailers helping them design and implement Flowcasting most folks are shocked that we don’t worship at the feet of forecast accuracy – at least not in the traditional sense.

In retail, with so many slow selling items, chasing traditional forecast accuracy is a bit of a fool’s game.  What’s more important is to ensure the forecast is sensible and assess it on some sort of a sliding scale.  To wit, if you usually sell between 20-24 units a year for an item at a store with a store-specific selling pattern, then a reasonable forecast and selling pattern would be in that range.

Slow selling items (indeed, perhaps all items) should be forecasted almost like a probability…for example, you’re fairly confident that 2 units will sell this month, you’re just not sure when.  That’s why, counter-intuitively, daily re-planning is more important than forecast accuracy to sustain exceptionally high levels of in-stock…whew, there, I said it!

What an approach like this means is that planners will no longer be dilly-dallying around tuning models and learning intricacies of various forecasting approaches.  Let the machine do it and review/work with the output.

Of course, sometimes, demand planners will need to add judgment to the forecast in certain situations – where the future will be different and this information and resulting impacts would be unknowable to the algorithm.  Situations where planners have unique market insights – be it national or local.

Second, and more importantly, it will allow demand planners to shift their role/work from analytic to strategic – spending considerably more time on working to pick the “winners” and developing strategies and tactics to drive sales, customer loyalty and engagement.

In reality, spending more time shaping the demand, rather than forecasting it.

And that, in my opinion, will be a game changing shift in thinking, working and performance.