How to create new optimization world records

How good are you at optimization?

It’s a valid question – and one we’re often asked by prospective customers. Given that the difference between great optimization and good optimization can run to millions in additional savings, prospects naturally want the very best.

So how good are we?

We could, of course, rely on customer case studies to prove our point. We have plenty of those, but they tend to be persuasive rather than conclusive.

Or we could ask ourselves that question and answer it in the only way that counts in the optimization world: We could attempt a new world record.

Optimization experts enjoy pitting their skills against each other, and there are famous academic optimization challenges on the web with recognized benchmarks. As our optimization platform has been implemented at many leading logistics providers, we set our sights on a well-known logistics optimization challenge known as the vehicle routing problem with time windows (Gehring & Homberger’s 1000 customer instance: c1_10_4).

A couple of academics had just set a new world record and published a paper in the Parallel Computing Journal (Mirosław BŁOCHO, Zbigniew J. CZECH, A parallel memetic algorithm for the vehicle routing problem with time windows).

Could we beat it?

The option of locking some of our PhDs in a room, and only releasing them when they’d broken the record and produced another academic paper, wasn’t going to cut it. Contrary to popular belief, optimization experts just wanna have fun. And so we created the optimal environment for us to have truckloads of fun, and launched our first Global Optimization Challenge.

We opened the competition to all Quintiq employees worldwide, and encouraged collaboration by allowing prizes to be shared. As we operate on four continents, our experts met to share insights in a specially created forum. A team was on hand to support requests for changes to our optimization tools, and provide development versions as quickly as possible.

Did we have fun? You bet! At the end of the six short weeks, one of the competitors posted a wistful “I wish it was not closed yet :).”

And the world record? Smashed by a convincing 170 points, thanks to the combined efforts of all the participants. The evidence is right here on SINTEF’s Transportation Optimization Portal.

Shortly afterwards, one of the winners sent this email. It’s reproduced here as a tribute to ‘the Quintiq way’ of commemorating achievements (literally the icing on the cake) and sharing success.

From:            Victor Allis
To:               Colleagues MY
Date:            11/2/2012 9:08 AM
Subject:        Cake Treat

Dear Colleagues,

As CEO I would like to encourage you all to break many more World Records and therefore I have invested the prize monies that I won with the Optimization Challenge in treating you all to cake. There will be cake distributed to the pantries from 1.00 pm today. Please feel free to help yourselves.  Enjoy!

For technical insights, visit

Detecting hidden optimization potential in a field services company

The director of field services was skeptical.

How could there be any optimization potential – hidden or otherwise – when the company already had an impressive utilization rate?

I was in a meeting with the CEO and director of services of a company with a large mobile workforce and an international client roster. Their workforce was both highly skilled and diverse: each of their field service engineers had a specific type of technical expertise for a particular kind of assignment. To add to the complexity, some clients had strong preferences for certain engineers, and many of the tasks could only be assigned to personnel with the right certification.

I didn’t know if there was hidden optimization potential either, so I did what I always do in such circumstances: I listened and asked some questions.

“Please tell me how you plan your operations now.”

“We have a system that supports our planners by highlighting the engineer with the right skills who is nearest the client’s site.”

But what if you have an assignment that can be completed by a junior engineer when the engineer who’s nearest is relatively senior? And what if there’s another task – a little further away – where the skills of a senior engineer are required?”

I pressed on.

“What if you have two senior engineers available and one of them is 200 miles east of the task while the other is 100 miles west. Which one are you going to assign? The one who’s nearest? But suppose there’s also another assignment that’s 200 miles west of the first task. Now the engineer who’s still available has to travel 400 miles to get to the assignment, making the total distance traveled by both engineers 500 miles when it could have been just 300.”

“Well, perhaps our planners catch these things.”

“But in your system, your planners only see the nearest qualified engineer. They’re not just assigning a couple of engineers to a few tasks; they’re dealing with hundreds of engineers and tasks every day. How often do you think your planners miss opportunities like these?

“Can’t say. No idea really.”

Those of you who’ve read my previous posts have probably spotted the ‘coping mechanism’ behind this company’s hidden optimization potential. The question now was, What would happen if we replaced that sub-optimal ‘rule of thumb’ (nearest qualified engineer) and applied intelligent decision support to a representative day?

The results surprised even me:

  • Time lost by field service engineers (for example by waiting and travelling) was reduced by 18%
  • The number of occasions when an over-qualified engineer was assigned to a task dropped by 20%

Continue reading

Uncovering hidden optimization potential (Pt 3)


At Disney, employees called ‘Imagineers’ are encouraged to let their imaginations run wild when developing new attractions.

‘Imagineering’ has earned Disney millions. It can do the same for you.

So how would you organize your operations if you could ignore all practical planning limitations and optimize the real KPIs you need to optimize?

Many leading companies around the world have already accepted the challenge of re-thinking their processes and closing the gap between the KPIs they steer by and their real KPIs. Their starting point was a simple question: What does my business need to achieve?

In every case, their hidden optimization potential lay in the gap between:

  • The actions taken to optimize proxy KPIs, and the actions that should have been taken to optimize real KPIs
  • The buffers they relied on to cope with complexity, and the optimal buffer size
  • Their old product portfolio, and the new products that were possible once they were no longer limited by complexity

Continue reading

Uncovering hidden optimization potential (Pt 2)


Matching demand and capacity precisely can be a complex, time-consuming challenge.

Take this wooden puzzle for example.

Getting those last few pieces of wood to fit isn’t going to be easy. But what if you had lots of pieces in various shapes and sizes so there was always a shape that fitted a space precisely?

Solving a planning puzzle is easy – as long as you’re willing to throw resources at it. If you have 200 service engineers when there’s only work for 100, it’s easy enough to incorporate all your business rules and labor regulations to arrive at a feasible schedule. The real challenge lies in improving your KPIs by doing more with less: Any reduction in the number of resources dramatically reduces the number of feasible schedules.

Arriving at an ‘optimal’ schedule (one that’s feasible and enables you to achieve key business goals) unaided, is about as likely as winning the lottery.

There’s a world of difference between buffers that enable a business to deal intelligently with volatility, and buffers that exist because the business has very little control over its operations.

Which kind do you have – and how can you tell?

What should a supply chain manager in search of hidden optimization potential do next? Find out why a visit to Disneyland may not be such a bad idea when you catch the final installment in the series.

Uncovering hidden optimization potential (Pt 1)


Large, highly successful companies often find it difficult to take their game to the next level. The increasing complexity that accompanies success can seem like a formidable barrier to continued growth and success.

But isn’t.

The real barrier is simply the limitations of the human brain. While the average person finds it difficult to remember more than seven digits, the average planner regularly makes decisions that involve millions of options and a bewildering array of business rules and constraints. It goes without saying that if planning is a struggle, re-planning to recover from day-of-operations disruptions is a nightmare.

How, then, do businesses cope with the obvious mismatch between limited human brain power and apparently limitless complexity?

The coping mechanisms of complex businesses are so common that it’s worth reminding ourselves that they are far from ideal: They only exist because human beings just aren’t very good at processing large quantities of data.

Here are a couple of coping mechanisms that you may recognize.

  • Reducing complexity by oversimplification

This is planning by ‘rule of thumb’. Instead of considering all the factors that determine whether one course of action is better than another, planners focus on optimizing a particular KPI. For example, planners may focus on minimizing the number of empty miles, even though this KPI is only one of many that determine the cost of a trip.

  • Ignoring complexity by controlling what you can

A policy of ‘divide and rule’ provides an illusion of control by dividing a planning challenge into manageable parts. For example, instead of optimizing the entire sequence of a multi-stage production process, planners may focus on optimizing individual stages while ignoring how each decision affects the process as a whole. Similarly, a service organization consisting of several departments may optimize the utilization rate of employees in each department while ignoring opportunities to save costs by deploying employees across the organization.

The culprits in both these cases are proxy KPIs that stand in place of the real KPIs the business needs to optimize. In the logistics example, the real KPI was the total cost of trips – and not the number of empty miles. In the case of the multi-stage production process, the real goal was to improve the productivity of the entire process rather than the output of individual stages.

Controlling what you can control, and planning by rules of thumb put complexity in the driving seat. Your business is organized around complexity, and the KPIs you steer by are chosen because there is simply no way to steer by the real indicators of operational excellence.

Are there other signs that a breakthrough is possible? Check out Part 2 in Arjen’s series on uncovering hidden optimization potential

What Usain Bolt can teach you about optimization

I often come across C-level executives who are well aware of the complexity of their planning puzzles. These are people who are familiar with the concept of ‘NP-hard problems’, and phrases such as ‘an exponential increase in complexity’ and ‘more possibilities than there are atoms in the universe’. While they suspect they could be solving their planning puzzles better, they’re also deeply skeptical of the claims of ‘solution providers’. After all, aren’t their puzzles virtually unsolvable?

Yes – and no.

It’s true that any business of a significant size will have planning puzzles that are impossible to solve optimally. According to academic research on NP-hard problems, it’s impossible to guarantee that optimal solutions to difficult problems will be found in a reasonable time frame. However, with almost all real-life planning puzzles, you can get excellent results very quickly. How quickly? Seconds and minutes when needed; hours when you can spare the time.

In the real world, we’re more interested in knowing what an excellent result looks like (the 9.63 seconds that won Usain Bolt an Olympic gold) than in theorizing about the fastest time humanly possible. Most of our customers aren’t interested in theoretical discussions either. They want practical results.

The way to those results was summed up by a former CIO of United Airlines. When he discovered that we’d both spent many (too many!) years in academia, he said, “Victor, when we were academics we took simple problems and tried to find very complex solutions to them. Now that we’re in business, we take very complex problems and try to find simple solutions.”

Finding the right solution to your planning puzzle isn’t about aiming for an unknown theoretical optimum. The trick lies in locating practical results that are really, really good – but not necessarily optimal.

After all, what makes more business sense? Arriving at a solution that’s 99% optimal in less than two minutes, or achieving 100% optimality in days, months or even years? Compared with what planners with old-fashioned tools are achieving, a solution that’s 99% optimal could well yield millions of additional dollars per year.

In  many cases, you don’t need to know how good we can ultimately get. We’re perfectly happy, just like Usain Bolt, to significantly outrun the competition.