Sheetless: Our Philosophy

Tue Oct 29 2019

by John Hill

Here at Sheetless, we’ve made it our mission to help organisations design better decisions. We are a team of modellers and computer scientists who’ve been inspired by the way that engineers design and improve physical systems and, with Sheetless, we are bringing the same ideas to human systems.

A Short History of Physical System Design

In 1901, two bicycle mechanics from Ohio built something that changed the world. It didn’t look like much - just a coffin-sized wooden box with a fan at one end. It was the world’s first wind tunnel. In this wind tunnel, they tested scale models of aeroplane wings as they raced to be the first people to achieve powered flight. In just 2 short years, they were able to rapidly design and test hundreds of designs, and finally scaled up the best one. On the beach at Kitty Hawk, North Carolina, in 1903, the Wright Brothers’ Plane took off. And the rest is history.

The Design-Test-Modify Loop

The wind tunnel succeeded in reducing the time taken to complete the design-test-modify loop to such an extent that the Wright Brothers gained the upper hand over a much better funded rival in the race to the skies. Today, wind tunnels are rarely used. The advent of computers meant that computational models could be designed and tested faster and cheaper than physical models. Computational Simulation is now the method of choice for optimising physical systems.

Screenshot 2019-10-29 at 22.35.30

Organisational (Human System) Design

Back on the ground, organisations continue to iterate through the design-test loop in the equivalent of the pre-wind tunnel way. Decisions are designed by people, but then tested in the real world where mistakes cost time, and money. It can take months, or years before the data reveal whether a decision was a good one or not. Bad decisions can be passed off as bad luck, or bad timing. In an increasingly volatile and competitive world, this is no longer good enough.

What we set out to build was a "wind tunnel" for decision-making in organisations. A tool for understanding the outcome of decisions and strategies, before implementing them in the real world. There are of course, some differences between designing physical systems and human systems - or organisations - as most people call them. While wind tunnels were used to study the flow of air across physical surfaces, in organisations, the key flows are of information and data across people and entities. But there are many parallels between the two domains.

Models Matter

What the Wright Brothers figured out was that you could use a model of a wing - an approximate representation of a wing - to learn about the performance of the real wing. The beauty of models is:

  1. They are easy to subject to testing against simulated scenarios

  2. They provide a focal point for collaboration in the design process

  3. They provide an objective way of communicating the relationship between design elements

You’re (Probably) Already Using Simulation!

In fact tens, maybe hundreds, of thousands of organisations are already using computational simulation to help them make decisions. They are using spreadsheets to model how their supply chain evolves under a range of macroeconomic scenarios, or to understand how customers are likely to react to changes in pricing, or marketing strategies. Millions more organisations do simulation outside of spreadsheets - they do it in the computers in their heads! We are all capable of subjecting our reality to scenarios and thinking through the consequences. It’s one of the cornerstones of human intelligence.

The bottom line is that these models already exist, whether in complex, error-prone spreadsheets, or inside decision-makers’ heads. The models are the right ones, but the tools we’re using aren’t helping us to reach better decisions. The value they can offer our organisations has been locked away. Why is that?

How do you Design Good Decisions?

At Sheetless we think that good decisions are made when:

  1. They are well-tested against a range of future scenarios

  2. People and departments work together to share their knowledge and data

  3. Decision-makers understand how the parts of a system relate. They understand cause-and-effect.

Because simulation techniques fell at the advanced end of the analytics spectrm, typically they were brought inside organisations by external consultants. Even if the models are correct, and they provide useful directional insights for decision-making, this outside-in thinking doesn't become a part of the organisation's DNA. And since good models tend to be "grown" rather than "off-the-shelf", unless they become part of the fabric of decision-making, they don't get used and improved.

Models as Shared Understanding

Decision-making and analysis should be done by people closest to the problems. Motivated individuals with in-depth knowledge and know-how about their function should be able to embed that inside models, and proliferate it across the organisation. These insiders are the real agents for change inside organisations. They are there for the long-haul and are thus incentivised to maintain and support models.

Spreadsheets Are Wonderful (No, Really!)

Today, the most widely-used tool for building models in any organisation is the spreadsheet. Spreadsheets have gained such ubiquity for three key reasons:

  1. Their flexibility

  2. Their usability

  3. Their ability to integrate data from the real-world

Anyone can pick up a spreadsheet and turn their knowledge and know-how about the real world into a quantitative model that can be tested, shared, and scrutinised by others. This has been transformational to the way we run our organisations. They are also a great way to introduce real-world data into models, to ensure that we our models reflect reality - a key concern for any modeller.

But Limited

There are things that spreadsheets can’t do, and we think this is holding organisational decision-making back. While great for building models, scenario testing is a slow process in spreadsheets. Finding the key model inputs to change can be laborious - they may span multiple worksheets and lurk inside complex formulae. Sometimes, when a previously unknown scenario needs testing, we find that the structure of the spreadsheet needs to change. A spreadsheet built to answer one set of questions is rarely flexible enough to answer a new set of questions.

Spreadsheets can also be a nightmare to collaborate on. The great strength of spreadsheets - their almost unlimited flexibility - is also their achilles heel. The same problem can be solved in a bewildering number of ways. Ask 10 analysts to build a simple model in a spreadsheet, and you’ll get 10 different spreadsheets! Furthermore, the cells which make spreadsheets so easy to develop incrementally also hide away the structure of the model. Understanding how data mutates through a model requires lots of digging into cell references, and chasing formulae across worksheets. Being handed somebody else’s spreadsheet can be a nightmare. It’s often quicker to rebuild spreadsheets from scratch, rather than trying to disentangle their model structure. This is incredibly wasteful. How many millions of analyst hours each year are being spent rewriting functional models, simply because the structure is so opaque?

This opacity also means that spreadsheets aren’t a good means of making decisions. It can be hard to understand why something changes. Following the data through such an opaque model structure is difficult. Good decisions are reached when we have a good understanding of cause-and-effect. Spreadsheets - and particularly the large and complex spreadsheets found inside most organisations - fail this test.

There is a Better Way

We think that the structure of a model is just as important as the data and mathematics that goes on inside it. When models are well-structured, two great things happen:

  1. Models become much easier to collaborate on

  2. Cause-and-effect is easier to understand

We also think that spreadsheet are missing a crucial dimension of reality. Time doesn’t always fit easily into a rows-and-columns view of the world. When we set out to build Sheetless, we knew that time would be front-and-center of all models. Drawing on insights from Computational Simulation, Complexity Science and System Dynamics, we developed the Sheetless platform. Models are composed of simulation elements - building blocks that govern how things change through time - and connected together to form complex model structures which can visually describe any model structures - even quite complex ones like processing queues and feedback loops. Ask 10 analysts to model the same thing in Sheetless, and you’ll probably get 10 identical models!

What this Enables

By design, the Sheetless platform transforms the way organisations can leverage the Design-Test-Modify Loop to make better decisions. By placing time and model structure at the heart of Sheetless models, analysts can iterate rapidly through scenarios and modifications (or decisions) to identify better decisions. We even see a future where we can use computational techniques to speed-up (or automate) the process of finding optimal decisions.

Built for Organisations, In the Cloud.

We’ve combined the usability of Excel, with all the benefits of well-structured models, which place time as a first-class dimension. For the first time, organisations can build models built for decision-making. And you can get started, for free, today.