Summary

Statistics has a crucial role to play in manufacturing industry, from the design cycle through to production, in the analysis of field data and, where there is a large customer base, in the management of those data. Statistical techniques originally developed within a wide variety of areas have immediate applications in manufacturing industry, yet they are hardly known about, let alone used. It is notable therefore - and not only in the UK - that the use of advanced statistical techniques in manufacturing is very rare.

In many countries - and in the UK in particular - manufacturing industry is very fragmented, with many suppliers contributing to a final product. It is very difficult in such a climate to impose disciplines from above even though this might seem to be the sensible approach.

The purpose of this web site is to provide a one-stop-shop for all industrial applications so that small companies in particular need look no further. The site will provide open access to lists of statisticians and also a search engine, mailing and discussion lists, a newsletter, and links to funding sources, business aids, standards and professional bodies. It will also offer, for a subscription, broadband access to cutting-edge software and searchable indexes of journals.

The overall aim is to place in the hands of the statistician in industry, or the statistically aware engineer, the leading technical and software approaches with full support and training where necessary from the statistical community, so that the best use can be made of this valuable resource. In this way, the industrial statistician or engineer - together perhaps with consulting statisticians in support - will have available the fullest possible library of methods to enable optimal solutions to be found to industrial problems.

While the initial site is aimed at a UK audience, international expansion will follow as soon as practicable. This offers UK export potential as well as international collaboration.

Introduction

The product may be formed from a variety of components, each of which has to work optimally and together with the other parts. The quality of the final product depends critically on the quality of each component and on how they interact.

Quality is about correct and consistent performance. Correctness is a function of the predictive part of the engineering, an area where engineers are particularly competent. Consistency is about whether a product

• can be mass-produced allowing for the variability in constituents and
• whether it will perform in the field under a wide range of conditions which may be hard to predict.

The science of uncertainty is called statistics. It is a discipline which draws problems from a wide applied field - much wider than manufacturing industry - and finds solutions sometimes requiring complex concepts. While this wide variety of applications areas can lead to some problems of terminology, it also means that solutions applicable to engineering problems can be found in work done in almost any area of intellectual pursuit where measurements have been made.

A product follows the following iterative cycle, after establishing the customer's requirements:

• Original design
• Manufacturing a prototype
• Optimising the design
• Verifying design against specification
• Planning and implementing production

• Monitoring the product in the market place

Characteristics of the UK manufacturing industry

In the UK, the contribution of manufacturing to the GDP has shrunk to about 23%, the smallest in the industrialised world, and this is becoming a threat to our technological base. In comparison, manufacturing in Switzerland - also with a large financial sector - contributes nearly 50% to the Swiss GDP.

A feature of the UK economy is its domination by small companies for reasons that may be cultural. This has long been noted and means that the economy is very flexible, but it makes manufacturing on a global scale difficult. According to which source you find, there are different numbers reported for the size of the manufacturing economy in the UK. The DTI claims that "98% of the 135,000 UK manufacturers are SMEs", while at more or less the same time claiming that there are "332,070 manufacturing businesses of which 99.2% are SMEs" (this link appears to be broken at the moment). The different definitions of SME cannot be responsible for a factor of over 2 but it is clear that SMEs form the vast majority of manufacturing businesses. Curiously the number of non-SMEs is therefore about the same for both these sources at about 2,700.

Perhaps a more reliable source is from the Office of National Statistics' web site which includes tables drawn from Inland Revenue and VAT returns. This shows a very interesting distribution indeed . For 2001, Table 9.2 of PA1003_2001.pdf reveals:

It is possible that different definitions of 'manufacturer' and 'SME' are responsible for this discrepancy - for example some companies which are really design consultancies that work for a number of clients may claim to be manufacturers. Whatever the number, it is likely that the true count of manufacturing or immediately involved companies is above 200,000 and that this is dominated by micro and small manufacturers. Some of the tiniest may be garage-manufacturers that are just satisfying a local and/or inexpensive market or artisans with a particular niche but on the other hand there will be also be a number of consultancies that work across industry particularly at the design phase and that are not included here.

The UK tendency to favour small businesses - Napolean was a little misguided to confuse these with shopkeepers - has been exacerbated by such recent practices as downsizing and outsourcing - and means that manufactured goods require a long supply chain. It therefore means that manufacturing in the UK becomes a bigger organisational problem than in other countries. The US has a similar economy, though much larger but they have in recent years managed to increase their proportion of manufacturing GDP. However the US is much less exposed to currency fluctuations since the majority of US trade is internal.

This is not to be despondent from a UK perspective - there are many excellent small high-tech companies in the UK that are working at the frontiers of technology in electronics, telecommunications, computing, aerospace, plastics, petrochemicals, life sciences, etc. that will pull through.

In some sectors, such as financial services, pharmaceuticals and life sciences, the UK is noted as a global player. In engineering, the main global presence is in aerospace, which dominated by a large defence (ie government) demand. Elsewhere, there is no longer any global industry: vehicle manufacture, ship-building and general manufacturing of many commodity items have long left these shores.

These features, while particularly evident in the UK, are visible in other countries, and as de-regulation etc. becomes more prevalent globally, what is happening today in the UK is likely to become the norm elsewhere eventually. Industrial statisticians world-wide therefore find it very difficult to get the message across to the millions of small manufacturing companies that are sub-contractors or lower.

The fragmented nature of the UK manufacturing economy poses a number of problems such as how best to organise the long supply chain so as to take advantage of its flexibility without imposing unnecessary overheads. The UK Government is very concerned about the apparent low productivity in this country and various documents have been released, and programs have been started, with the aim of improving the situation. The documents can be found on the Office for National Statistics and DTI web sites.

A problem in manufacturing

A number of questions arise:

• Is the apparent low use of statistics really a fact?
• Are there other approaches that may used?
• Why do engineers and manufacturers avoid statistics - or call it something else altogether.
• Does manufacturing need statistics at all?
• Are we right to be worried?

Given that concern is shown at these elevated levels, we may therefore reasonably deduce that there is a problem. At the working level, the reasons for concern are clear to anyone involved on a day-to-day basis in both statistics and industry. There is a communications gap.

Statistics is not, by and large, a subject that can be taught in an abstract fashion, and it is notable that some engineers (as well as other applied scientists) fail completely to appreciate its subtlety. Equally, many statisticians were originally qualified in other disciplines, but found that a fascination with uncertainty drew them into statistics. Some fundamental statistical principles have been encapsulated within the quality control mantras that can be heard in Board Rooms - SPC, Deming, Taguchi, 6-s etc., etc. These have an engineering origin and raise the issue of uncertainty, and are certainly proven tools, but there are three problems in their implementation:

• They are all top-down management-driven philosophies that depend on being pushed down into the supply chain through being championed either by the client company or by dynamic individuals,
• They are taught as self-contained topics in fairly short courses as relatively stationary concepts whereas the underlying statistical approaches are dynamic - always being developed, refined and improved, and adapted to the problem in hand. In principle of course all users should attend frequent refresher courses but, like training courses in the first place, these need to be tailored to the company and there are many other demands on engineers,
• Not all statistical issues and approaches are covered by these and, as in all cases of lack of knowledge, the practitioner in industry may not recognise that there is a problem or alternatively may struggle for a solution when one is readily available.

For individual projects the procedures can be insisted upon and, in the short term, other client companies will benefit; but retention of culture is an important problem in the dog-eat-dog company culture and it is all too easy to belittle the apparently unproductive research or quality control departments. Eventually of course such companies should disappear since customers will turn away, but observation shows that accountancy bottom lines - which ultimately control the longevity of a company - have little to do with long term planning or with product longevity, and more to do with this year's profits and the current share price.

Therefore it is very difficult to promote or make available statistical approaches in such a fragmented and widely distributed manufacturing economy; top-down approaches are difficult to maintain as a supplier moves from client to client. The supplier itself is often aware of the problem but does not know how to tackle it, nor is there the financial facility. Many suppliers do not know how to find the right resource - apart from the globally successful industries which employ large numbers of statisticians, most other statisticians work in academia or for the government. And the small - even tiny - size of many companies means that they cannot individually allocate resources to solve statistical problems. No company of 10 people will directly employ a specialist that they believe they don't need and will be very reluctant to pay consultancy rates that may be a large proportion of the monthly salary bill.

The irony is that the UK is without doubt a global player in statistics - many statistical techniques were pioneered here. It is arguable that the pragmatic culture in the UK that derives from experimentation and practical applications also gave birth to the first Industrial Revolution and many other events. Statistics may be seen as quantifying this pragmatism and is therefore an inheritor of a fundamental part of British culture.

We are therefore left with essentially a marketing problem - how to encourage industry to make proper use of statistics, which is one of our flagship services. This is not only a UK problem but one that permeates the world. But the world-class UK statistics industry needs to be put together with the UK's highly distributed manufacturing. The solution to this problem therefore offers opportunities on a global scale.

Implications

It is a natural defensive reaction, when faced with lack of understanding, for people to collect together in critical masses - it is a response to an alienation born of suspicion. Since statistics is the science of uncertainty, its solutions are expressions of the uncertain. People unable to comprehend uncertainty - particularly those able enough to comprehend certainty very well - find this very threatening and therefore want nothing to do with uncertainty.

In manufacturing, there are for the most part so few statisticians that they cannot form the critical mass necessary. Thus statisticians gather together in those industries where their skills are recognised; manufacturing is not included.

It is impracticable for statisticians to go out in person to do all this - they are busy people with full agenda. It has been seen that holding conferences - interesting and stimulating though it is to the delegates - does not by and large succeed in promoting statistics directly into industry. Conferences tend to become gatherings of statisticians interested in industry rather than the other way around and very few manufacturers attend.

Solution

Engineers working in manufacturing are all highly numerate and technically competent. They may realise that statistical help is required but are faced with tight financial constraints or a management that expects something for nothing, or does not see the point. Optimum solution may require specialist statistical software that individual companies may not be able to or want to afford for one-off applications.

The small-company structure described above means that such companies are unlikely to invest in statistical software or even have sufficient IT maintenance skills. This means that solutions, where they are found, are going to be calculated using general purpose software or will require engineers to spend a lot of time writing their own programs from scratch. Both of these approaches have associated concerns of quality control.

We cannot force industries to change old habits but it would be much easier if there were a simple resource that they could use to check out at least some of their problems. With many small, widely distributed manufacturers, potential clients are widely scattered geographically. Equally these clients have applications that, while they may be statistically special, can be professionally handled by any statistician who will recommend or engage with a specialist if necessary.

We forsee the solution to this dilemma to be based on an Internet application. We are therefore in the process of assembling a web site which will be a single entry portal for industrial statistics. The site is

The point of industrial statistics rather than engineering, manufacturing or business statistics for example is that it is a more general application and we do not want to exclude parts of industry that may not regard themselves as manufacturers even though the manufacturing process is the main target. The power of statistics is also it's Achilles' heel - it is a science that can be applied to almost anything so that all disciplines regard it in some ways as an interloper.

Features

The Industrial Statistics web site will include at Level 1, available free:

1. A list of statisticians, departments, institutions and consultancies that are interested in working with industry, broadly categorised by special interest, with links, and geographical locations;
2. A search engine that will locate potential collaborators;
3. Various closed mailing and discussion lists that can act as introductory help forums;
4. A monthly newsletter delivered by email that may include interesting solutions, notification of events, etc.;
5. A monthly digest of journal contents with industrial statistics applications;
6. Industry- and geography-specific pages;
7. Links to (geographically dependent) funding sources, including co-operatives, and information as to availability;
8. Links to institutions and companies capable of satisfying a wide range of educational needs - long and short courses, academic courses, training, etc. delivered onsite, on-campus or electronically;
9. Links to statistical definitions, notes and other useful information;
10. Case-studies either by link or by video-streaming;
11. Provision of on-line video-streamed courses either from major institutions such as the Open University, training institutes or on a bespoke basis for a few clients - all on a paid-basis;
12. Links to business aids such as insurance, business link, contract forms etc;
13. Links to standards and to bodies such as the NPL, ISO etc.;
14. Links to professional and learned bodies, such as the RSS;
15. Access via a browser.

1. Broadband access to a wide variety of specialist software either running directly on IS servers or via the Internet. This software will be available openly but with associated quotas of time and space;
2. The charging rate for the software will depend on installation and maintenance costs. It is intended to include all mainline statistical software, whether Commercial or Open Source, subject to licence agreement;
3. There will be the facility to buy more time and/or space as required so that companies can avoid mounting the software themselves;
4. Access to searchable indices of statistics such as Current Index of Statistics;
5. Access to journals online;
6. Access to library ordering systems such as the British Library;
7. Access and programs may run either via thin client software (Citrix or VNC) or a native X-server (eg X on Unix or Exceed).

• to carry out the correct calculations and
• to interact fully and professionally with any consulting statisticians or with their own statistical advisors.

The site will also try to help with collaborative working. Individuals and companies with separate level 2 access may allocate a group structure to their data so that others may share it using standard Unix facilities. This may if required be extended to sharing any additional processing resources purchased so that, for example, one lead company will purchase all the processing required and let the other companies share this resource - perhaps subordinates in the supply chain or statisticians working on the problem.

As each user will be a relatively standard Unix user, email and other facilities will also be available subject to sensible safeguards. This will enable each user to employ the @industrial-statistics.com domain name as well as providing web facilities.

Marketing to statisticians

List members will not become the object of mail shots since the mailing lists will all be closed - only mail from subscribed members will be allowed, and spam filters will be installed.

For those statisticians who are not able to work via University departments, there will be a either a click-through Professional Indemnity Insurance if we can obtain it, or a group PI and ISO9000 membership subject to clear standards of working and a percentage of the fee. These will be charged on a project basis and other business aids will be placed on the web site. This will make the freelance consultant able to handle occasional contracts where they may not be able or want to work full time.

For statisticians who do not want to work freelance, we also hope to provide a shell so that the project contract will actually be with Industrial-Statistics.com and the statistician will be an episodic employee. Thus we will take care of all tax and National Insurance as well as providing the PI and ISO assurance to the client. It is possible that this mode of working will be attractive to people from outside the UK for tax reasons subject to legal constraints.

Marketing will be direct to individuals, to Universities and to all institutes that are associated with statistics such as 6-s training institutes so that the client will have a wide choice of how to proceed.

Presentations will be made at statistical conferences, meetings etc.

It is crucial that statisticians see this approach:

• as a way to promote themselves or their departments into industry;
• as a way of retaining some ownership of statistics, which is such a broad-based discipline frequently taken over by others - for example neural networks and data mining which both originate in statistical concepts;
• for academic statisticians, as a way of improving links between Industry and Higher Education - much lauded these days;
• as a way for academic statisticians to gain access to industrial problems and data that will be very useful for longer term research or for research students.

Marketing to industry

General marketing will be carried out:

• by extensive Internet research, web promotion and contact via email;
• by newsletters, discussion lists and digests will keep the topic at the top of the agenda;
• by appearance at manufacturing and trade fairs;
• by conference presentations;
• by articles placed in trade journals where possible;
• by promotion via government agencies if possible;
• through peer sponsorship by leading manufacturers - the fear of being left behind being the motivation.

Thus the Level 1 activity with newsletters and discussion lists is actually part of the Level 2 marketing campaign.

Working practices

Most statistical work for industry to date is for large companies and institutions. The fees involved are not particularly onerous for such bodies but for a small manufacturer, unless budgeted by their client, the fees can be a problem. Much of the purpose of industrial-statistics.com is to overcome this hurdle.

The conventional paradigm therefore needs to be addressed and work will not necessarily be done out-of-line by the statistician. The collaborative facilities of the Level 2 design will enable the statistician to co-worl with client staff in an advisory role or even for a technically competent engineer to do much of the work him or herself. There is another way of working - as a percentage of the savings made.

Four approachs are envisaged:

1. by conventional consultant working essentially by the day or project and billed as appropriate,
2. by collaborative working where the statistician engages with the client's staff as an advisor, guiding calculations done by the client,
3. completely by staff within the company, availing themselves of the contemporary software at Level 2 and suitable textbooks, doing the calculations themselves, and
4. where a statistician - or group of statisticians - take a percentage of the savings from the improvements over an agreed period of time.

Company audit

If such services are carried out under approach 4 above, the company is not committing money forward. The statistical consultancy is taking a risk that no savings can be made or that the company will not survive. Hopefully both these are very unlikely but it does mean that the percentage of savings should be quite generous - typically 50% over a two year period but this depends on the company turnover and present profit, the size of the customer base and many other factors. This is the way in which some management consultancies work.

By entering such a partnership arrangement, a company can be assured of continual interest by the statistical consultancy as there is every incentive for the statisticians involved to look in detail at the total production process. It is likely that statistical techniques will continue to be employed after the period has expired and a cultural change will have been achieved. There is no reason in principle why this facility should not be arranged by Industrial-statistics.com, which could act as banker for the interim to pay the individual statisticians.

Income

• Sponsorship from collaborating statistician/training institutes, academic departments or companies either by annual fee or affiliate schemes;
• Sponsorship from flagship industries that want to be associated with cutting edge techniques;
• Sponsorship from software manufacturers either by annual fee or percentage click-through;
• Sponsorship from business services such as click-through project-based PI Insurance;
• Percentage kick-back or arrangement fees for courses etc;
• Sponsorship from book or journal sales;
• Initial support from regional and central government in the UK and EU;
• Standard subscriptions at Level 2;
• Supplementary purchase of processing at Level 2;
• Income from PI and ISO9000 cover;
• Income from providing a tax and NI shell,
• Income from providing a statistical audit service.

Legal

There may be a need to retain legal expertise so that, for example, a standard Contract form can be suggested. This is frequently a problem when working with small companies who do not have legal departments. The Contract should aim to be equitable between statistician and client and prescribe a mode of working.

There will be a professional requirement for all work done via IS to be covered by Professional Indemnity Insurance and the group or click-through PI facility will mean that no work - however small or protracted - need not be covered. Various insurers will be invited to make these facilities available and provision of ISO9002 will require the statistician to work in a prescribed manner to a QA manual, which will be available for downloading. Work carried out by direct contact - even though effectively introduced by the site - will be to the standards mutually decided.

Disputes will be handled at cost by professional colleagues who may be appointed either directly by the parties concerned or by IS. The mandatory use of PI should ensure that this will never happen.

Licences will be required for some Level 2 software and this may be expensive. In some cases, this will only provide a limited number of seats and therefore a processing limit will be placed once a queue develops.

Requirements

• This web site will need to be mounted in a secure location which will be provided by the Multimedia Centre at Manchester Metropolitan University, which has a 155Mb/s link to the Internet.
• The main systems will all be Linux-based.
• The Level 2 system will require a farm of computers linked together. There are a number of ways of handling this - the present selection is to use Mosix.
• At Level 2 it may be necessary to have some NT servers for software that cannot be Linux-mounted. It may alternatively be possible to use Win4Lin, which will be substantially cheaper and provide immediate links to the underlying data directories.
• The hardware requirements are proportional to the Level 2 membership so a part of those fees will support the necessary farm of machines.
• Hardware and software costs are therefore expected to be minimal for Level 1 applications.
• The web site will be clean with minimal overheads and no fussy graphics. It is not intended to be mass-marketed and clients will appreciate simplicity with a purpose.
• Workaround pages and other features will be included to ensure search-engine promotion of all the facilities.
• Where links are installed to other web sites, it will be suggested that reciprocal links are made to improve search engine rankings.
• It is intended to offer some sophisticated graphical exploration software that will require the clients to have broadband access.
• It is also intended to offer some case-studies and courses by video-stream, which will also require broadband access.
• Online data transfer will also be chargeable to cover the graphic-intensive applications.
• Facilities will be required for customer orders, although credit card will be preferred
• Customer tracking will be essential - in particular where there is click-through counting.
• The client software will be either browser based (Level 1) or Citrix/VNC/X based (Level 2). The Level 2 subscriptions will pay for the Citrix Server while in both cases the client software is free.
• One of the largest costs will be employing suitable staff - technical, marketing and journalistic.

International expansion

It will be necessary to approach EU funding mechanisms to ensure that sufficient language facilities are available. This is not an issue for expansion into the US or other English speaking areas.

John Logsdon
Manchester UK
4 July 2001