James and James
 +44 (0) 1604 968 820

Six and Six Sigma – How to Design Fulfilment Systems to Achieve World-Leading Accuracy

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+

Six's systems were designed from scratch following best industrial practices. To demonstrate a few of the ideas and principles the company is based on, we thought we'd share a brief insight with you.


Count the number of F’s in the following passage:

Finished files are the result of years of scientific study combined with the experience of years...

Most count only 3 rather than 6, missing those in the three occurrences of ‘of’ which despite being obvious in retrospect we fail to see even when looking carefully. Unfortunately humans are not infallible and simply ‘trying hard and double checking’ will never remove all errors.


In another experiment:

‘participants are shown a… video during which they are asked to count the number of times two soccer balls are passed between a group of people… We also ask if anyone saw anything unusual in the video. About 20% of viewers will admit to having seen a gorilla... As we watch the video a second time, everyone sees a person dressed in a gorilla costume slowly walk into the scene, stop, face the camera, and beat his chest before calmly walking out of the scene. Those who did not see the gorilla during the first viewing sit gap-jawed upon realising that they missed it altogether.'

‘Humans are not good at inspection. We often miss large things (like a gorilla) when we are focused on anything else.’ (Conner)

From these two examples it is clear that even with 100 inspectors checking items coming off a production line, it is still unlikely that all the errors will be found all the time. In fact human error usually accounts for more than 60% of mistakes and the rate of errors does not decrease in proportion to the amount of experience of the individual in that task (Duffy & Suall, 2004).

Automation is key to reducing the human aspect of any process, and hence the probability of a mistake occurring, but we must be careful not to assume that simply using a computer will help reduce errors. In the medical profession, handheld computers were trialed as an alternative to keying written notes; “[Data input] error rates… were 67.5 errors per 1000 fields, compared to… 10 per 10,000 fields for paper-based double data entry.” (Shelby-James, 2007). The computer made no mistakes, but the human-computer interaction allowed for complacency and systematically allowed errors.

pexels-photo-132907 (1)

How good is good enough?

Many warehouses claim 99.9% accuracy, and those who report their accuracy often fall slightly lower, generally between 99.3 to 99.6%.

99.9% equates to one error in a thousand and might seem pretty good, but then consider that this is:

  • 400 Letters lost in the post every hour
  • 4,000 Incorrect drug prescriptions each year

Six Sigma techniques used by world leading production companies strive to obtain all operations with 6 standard deviations of the correct mean or better. In simple terms that’s less than 99.9997% accuracy or 3.4 defects per million.


A Six Sigma process would reduce this to:

  • 1 Letter lost in the post every hour
  • 12 Incorrect drug prescriptions each year

This might seem an unnecessary level of accuracy in fulfillment but the customers who didn’t get what they ordered probably wouldn’t agree. Consider how much time companies spend on handling returns and customer complaints due to late, poorly packed, or wrongly delivered goods. Then consider that for every complaint there are probably 10 people who were disappointed but who couldn't be bothered to complain, and each of those customers (those representing the easiest next sell) are now far less likely to buy again.

Forcing Correctness

A new method of working is required where the operator isn’t just assisted by technology, but is physically restricted from making mistakes. It may come as some surprise but the methods that follow are not the draconian dictatorship that one might image, but are infact designed to empower workers to spend their time thinking productively about their work rather then continually being forced to focus intently on every action.

Shigeo Shingo’s ‘Poka-Yoke’ (pronounced ‘poka-yoka’) techniques are an example. His philosophy relied on making an action physically impossible so that the operator had no option but the correct one. For example, given a plug with a ‘this way up’ sticker on it and with two identical pins, it is unlikely the operator will plug it in the wrong way up, but it is still possible. Add a third, redundant pin to create a 3-pin plug, and it instantly becomes impossible to get it wrong. While this example is a very physical one, the same concept can be used in any process to ensure the correct steps are followed in order.

At Six we use this technique in several places, for example, to ensure it is not possible for the wrong products to be packed, or for an order to be wrongly addressed. Many a time we’ve seen a others with a stack of paper orders for the day, going through each in turn packing. With 20-30 pieces of paper around at a time it is understandable that however vigilant, occasionally the wrong packing slip or address is likely to be used. Our solution, to only ever have one piece of paper on hand - the right piece of paper. In front of each packing station is a printer. Once the packer has confirmed the order is correct, the packing slip with integrated address label is automatically printed. (Just as a double check, the packer also scans a barcode on the packing slip, confirming it’s the right one).

While the methods used may vary, the technique is always the same - to make it very hard or impossible for an action to be done incorrectly in the first place.

Catching the Mistakes

Unfortunately when dealing with other peoples' products there is not the opportunity to design the product packaging to assist with Poke-Yoke techniques and so it is inevitable that at some point, it will be possible (although much more unlikely than by normal methods) for the wrong product to be presented, especially when dealing with awkward unmarked and unpackaged items, for example plumbing fittings.

At Six we employ yet another check, one which takes literally no time at all. Rather than weigh all our parcels at the end of the day, each is weighed just before it is sealed. At this point the actual weight is checked against a calculated weight, based on the individual weight of the items, the weight of the chosen box or padded bag and an allowance for packaging materials. If the weight falls outside of preset boundaries, then the entire process is stopped and the order is investigated to find the source of the problem.

Continuous Improvement

The act of stopping the entire process, rather than sidelining the order is one that was used by Toyota and proved very successful. When a problem arose on the production line, a Toyota employee would pull an Andon cord, stopping his section of the line and calling an engineer. If the problem could not be solved within around 45 seconds then the rest of the line would also stop until everyone had stopped work. Amazingly, this helped Toyota have the best car production quality.

Initially a rather counter intuitive improvement process, Jidoka is the idea that if an unexpected problem occurs, you shouldn’t just make a note of it, continue with your work and correct the issue when it is convenient. Instead you should stop your work and correct the problem, even if this means disrupting others’ work and pulling management away from their tasks. Once everyone is involved the problem will be solved much faster. In these situations any errors, which are almost inevitably due to flaws in the process, directly result in the process being corrected before any other problems can arise. It also means that every problem gets fixed now, and not just added to a wish list.

Another Japanese method ‘Kaizen’ (which translates to continuous improvement) takes the process of automation a step further. It says that once you have automated and reduced a process such that human errors have been removed as far as possible, you should teach employees how to continuously evaluate the process to improve it further.

The Japanese talk of 'reduction of waste’, to clarify improvement. Here waste is meant as a waste of effort rather than physical rubbish. So having to correct a label is waste, as it should have been done correctly first time. Keying information in to a computer when the computer could have calculated or populated the fields automatically is also waste, as the latter is more efficient and more accurate. By removing the need for workers to be continuously and repetitively checking for errors, you reduce their stress levels and allow them the opportunity to improve work methods.


Forcing ‘First In First Out’ (FIFO)

FIFO is extremely important in all distribution systems and more so when time-dependant products are involved (e.g. Food). The ‘Kanban’ system (visual card), while originally intended for the re-order of product, can be applied in ensuring that FIFO is observed by creating a physical sign from either a mental or electronic record. The logic is this: If the minimum stock of address labels required is 500, then you have two trays of 500 labels. Labels are always taken from the top tray and when it is empty, you reorder 500 labels. While there may be a inventory record showing the number of labels left, having a physical empty tray is much simpler, more visual and likely to evoke a response. Likewise if there are two or more batches of a product, then they should each have their own shelf location. A card saying ‘picking’ is place on the shelf to pick from. When the shelf is empty the card is returned to the office, where the next shelf location to pick from is selected. This ensures not only that the correct batch is always picked, but also confirms when each batch is finished – adding another inventory check point. At Six this card is not physical as in the example above, but it a token passed within the software, though the effect is the same.


The techniques outlined above could reduce the occurrence of customer fronting errors by a factor of a hundred or more for little or no extra effort or cost, so you would expect their usage to be widespread. Unfortunately these basic techniques need to be well implemented, and while there are many books on the subject, many companies fail to do so. Many review their processes once but fail to continuously improve, stagnating their development, or use the techniques inappropriately where they can become restrictive and unhelpful. So while picking an order and packing it in to a box is very simple, the process for managing the accurate picking and packing of thousands of orders is very challenging.

Many companies have come to expect a certain number of customer complaints due to erroneous dispatches, though clearly this needn't be the case.


Conner, G. Poka-Yoke: Human-Proofing Your Process

Duffy & Suall (2004) Proceedings of ICONE12: 12th International Conference on Nuclear Engineering

Shelby-James, Abernethy, McAlindon, Currow (2007) Handheld computers for data entry: high tech has its problems too

October 11th, 2010 by Hannah Newman

Join the revolution

Thinking about outsourcing your eCommerce fulfilment? Our expert team can help find a package that’s right for you.

Do NOT follow this link or you will be banned from the site!