The Information Factory: Why Manufacturing Improvement Shouldn't Stop at Production

Manufacturers understand the importance of improving production flow. But what about the information, coordination and decisions that make production possible?

Walk into a manufacturing business and you’ll probably find that considerable attention is paid to what happens on the factory floor. Production schedules are monitored, quality problems are investigated, machines are maintained, and people look for ways to reduce waste, shorten lead times and improve productivity. Some businesses have formal Lean manufacturing programmes; others have developed their own practical approaches to continuous improvement. Either way, the work of making things is usually visible enough for people to observe, measure and discuss.

But do we apply the same discipline to the work that happens around production? Customer enquiries, quotations, order processing, production planning, purchasing and engineering approvals are all essential to manufacturing, yet the effort involved in coordinating them can be surprisingly difficult to see. A delayed quotation might appear to be an estimating problem, for example, when the real difficulty lies in obtaining technical information or agreeing who can make a decision. For some manufacturers, understanding this less visible work may reveal opportunities that improvements on the factory floor cannot address.

Every manufacturing business has an information factory

We tend to describe manufacturing in terms of physical transformation. Raw materials enter a factory, people and machines turn them into finished products, and the business manages quality, capacity, cost and delivery along the way. At the same time, another kind of work is taking place. Information is being created, interpreted, checked and passed between people so that the physical work can happen at all. A customer enquiry becomes a quotation; a quotation becomes an order; and an order creates requirements for materials, production capacity, technical specifications and delivery commitments.

Some of that information is straightforward. A part number, quantity or delivery address can be recorded and transferred between systems, provided that the original details are correct. Other information requires interpretation. A customer’s drawing may be ambiguous, a requested tolerance may be difficult to achieve, or a delivery promise may depend on capacity that has not yet been confirmed. These are not merely administrative questions. They influence what the business can promise, how it plans its resources and whether it can deliver what the customer expects.

A manufacturing business can be understood as operating through two interdependent systems: one transforms materials into products; the other transforms information into coordinated action. I find it useful to call the second the information factory. I’m not proposing a new management theory: Lean value stream mapping has long recognised information flow alongside material flow. Nor am I suggesting that information behaves like a physical component. It can be copied without being consumed, its meaning depends on context, and its quality may not be obvious until somebody tries to use it. These are not literally two separate factories: production and information work are intertwined. The metaphor is useful because it invites us to investigate information work with the same seriousness that we apply to physical production. We can look for delays, duplication, rework and unnecessary handovers, while also asking a harder question: how does information become understanding, and how does that understanding become a decision?

Why does a quotation take five days?

Consider a hypothetical manufacturer producing engineered components to customer specifications. A customer sends an enquiry, and an estimator reviews the requirements, calculates the likely manufacturing costs and prepares a price. The estimating calculation itself might take twenty minutes, yet the customer receives the quotation five working days later. From the outside, it looks as though the business takes five days to produce a quotation. That description is accurate as a measure of elapsed time, but it tells us very little about where those five days went.

Perhaps the drawing omitted a dimension that engineering needed to clarify. Purchasing may have had to establish the availability of a specialist material, while production planning needed to confirm whether the requested delivery date was realistic. The quotation might also have required a director’s approval because the customer wanted unusual commercial terms. None of these activities is necessarily unreasonable. Engineering should not approve something that cannot be manufactured, and a commercial commitment should not be made without understanding its risks. The delay arises from the way information and decisions have to be assembled across the organisation, rather than from the speed at which the estimator can perform a calculation.

Buying faster estimating software might reduce the calculation from twenty minutes to five, but the quotation could still take five days. Manufacturing managers will recognise the distinction between processing time and elapsed time; it is familiar in production improvement. The challenge is to apply that distinction to commercial and managerial work as well. We can see the estimator preparing a price, but we may not see the dependencies that determine when the estimator is able to finish. Until we understand those dependencies, we risk improving the most visible activity rather than the activity that actually governs the outcome.

Information moves, but understanding doesn’t always follow

Speed is only part of the problem. Suppose a customer changes a drawing after an order has been accepted. Sales forwards the revised document to engineering, engineering confirms the change and production planning receives the update. On the face of it, information has moved efficiently. Yet the original drawing may still be attached to the production order, purchasing may already have ordered material against the previous specification, and the production supervisor may be working from a printed instruction. Someone also needs to decide whether the change affects the agreed price or delivery date.

In that situation, the organisation has distributed a document without necessarily establishing a shared understanding of what has changed. This is why information improvement cannot be reduced to sending fewer emails, transferring data faster or connecting more software systems. We need to know what the information means, which version is authoritative, who must interpret it and who is responsible for acting upon it. A spreadsheet can contain accurate figures and still cause disagreement if different departments understand them differently. A shared document repository can make files easier to find without resolving which file should be trusted.

An automated notification may help, but it cannot by itself establish that somebody has assessed the consequences of a change. The important question is not merely whether information has arrived. It is whether the right people can use it to make appropriate decisions and coordinate what happens next.

Improving one activity might achieve very little

Manufacturing managers also know that making one machine faster does not necessarily improve the performance of the production system. If a downstream operation limits throughput, increasing the output of an upstream machine may simply create more work in progress. Information work is subject to similar dependencies, although the consequences are not always as easy to observe.

Imagine a manufacturer introducing automation to extract customer-order details from emails and create draft records in its business system. That might remove repetitive data entry and reduce transcription errors, both of which could be worthwhile. But suppose no order can be released until engineering has approved the specification. If engineering approval is the limiting factor, faster order entry will not necessarily produce faster release to manufacturing. It could simply create a larger queue of orders waiting for approval.

This is where systems thinking, including the perspective offered by the Theory of Constraints, becomes useful. Improving a part of a system does not necessarily improve the performance of the whole. That does not mean every administrative delay is a constraint, or that faster quotation turnaround matters only when it increases factory throughput. A delay might affect customer confidence or the chance of winning an order even when production has spare capacity. We therefore need to be clear about the outcome we are trying to improve, understand how the activities depend upon one another, and establish which delays actually matter.

The work that nobody deliberately designed

One of the most revealing parts of organisational improvement is discovering the work that people have developed to compensate for difficulties elsewhere. An employee maintains a spreadsheet because the main system doesn’t contain the information they need. Someone checks every order manually because mistakes have occurred before. A production planner telephones sales each morning because delivery priorities are not reliably communicated. An experienced manager is repeatedly asked to resolve discrepancies because they know the customers and remember how earlier decisions were made.

These arrangements may never have been formally designed. They have emerged because people needed to keep the business functioning, and it would be a mistake to dismiss them all as waste. The order check may prevent an expensive manufacturing error; the experienced manager may be exercising judgement that the formal process does not accommodate. What concerns me is not simply the presence of extra work, but the possibility that important parts of the organisation depend on informal practices that are poorly understood and difficult to sustain when particular employees are absent.

Before removing an activity, we need to understand the purpose it serves. Sometimes the answer is to eliminate duplication or an unnecessary handover. Sometimes it is to make a valuable check more reliable and less dependent on one person. In other cases, the wider process needs redesigning so that people no longer have to compensate for missing information or unclear responsibilities. Those are different interventions, and choosing between them requires understanding the work rather than judging it by appearances.

Where does AI fit?

Artificial intelligence offers useful possibilities here. It can help retrieve information, examine documents, identify inconsistencies, summarise correspondence and prepare material for decisions. Alongside conventional automation, it may also reduce the effort involved in routine information transfer and monitoring. But I would be cautious about beginning with the question, Where can we use AI? That risks selecting a technology before establishing what needs to change.

Return to the quotation example. AI might help extract requirements from an enquiry or flag missing details in an attached specification. That could be valuable, particularly if it helps people recognise a problem earlier. But if quotations are delayed because nobody has clear responsibility for resolving technical questions, introducing AI will not necessarily address the cause. Likewise, automatically processing drawing revisions could make matters worse if the business has not established reliable controls over which revision is approved and who must assess its consequences.

There is also a difference between preparing information for a decision and making the decision itself. AI might help an engineer notice an inconsistency in a drawing, but that does not mean the system should determine whether the component is suitable or safe to manufacture. The appropriate role of AI depends on the evidence available, the consequences of error and the judgement required. My preference is to use AI, where helpful, to investigate the work before deciding whether AI should become part of that work. The best improvement may instead be clearer responsibilities, a simpler process or better use of software the business already owns.

Start with one customer order

There is a practical way to begin investigating your own information factory without starting a major improvement programme. Choose a recently completed customer order that involved several departments, preferably one that required clarification or coordination. Follow it from the original enquiry to the point at which production could begin, using the actual emails, records, drawings and approvals rather than relying only on the documented procedure. Speak to the people who handled it and ask them to explain what they really did.

At each stage, ask five questions:

1. What information was required before the work could proceed?
2. Where did it come from, and how did people establish that it was accurate and current?
3. Where did work wait for clarification, approval or a decision?
4. What informal checks or workarounds did people rely upon, and what risks did those practices manage?
5. Which delays or difficulties affected the outcome — for example, the customer promise, manufacturing readiness, cost or quality?

Use the actual sequence of events to distinguish active work from waiting time. Resist the temptation to label every check as waste: establish why it exists before proposing to remove it.

You might discover that an apparently unnecessary check prevents a recurring problem, that the same information is entered into several systems because nobody trusts the original record, or that one experienced employee resolves questions nobody else knows how to handle. None of those findings automatically points to a software purchase. The value of the exercise is that it makes the relationship between information, coordination and decisions visible enough to investigate. Only then can you make a sensible judgement about what to improve.

Faster information is not necessarily better judgement

It is tempting to measure improvement in information work through fewer emails, faster approvals, shorter searches and reduced processing times. These can be useful measures, but they are not sufficient. A drawing revision that reaches production in seconds rather than hours is an improvement only if the correct revision reaches the right people, obsolete instructions are controlled and someone has considered the consequences. Similarly, a quotation approved in five minutes is not necessarily better than one approved in an hour if the faster decision overlooks a significant technical or commercial risk.

I would therefore distinguish between improving the efficiency with which information is processed and improving the quality of the decisions that information supports. Good organisational improvement needs to consider both, including occasions when they pull in different directions. The aim is not to introduce delays in the name of caution, but to ensure that speed does not come at the expense of the understanding and judgement on which reliable performance depends.

Manufacturing improvement has traditionally paid close attention to making physical production more efficient. Yet every product also depends on the information, coordination and decisions that determine what should be made, how it should be made and when it can be delivered.

Improving that work does not necessarily require new software, automation or artificial intelligence. It begins by understanding how the work actually happens — and by recognising the difference between moving information and making a sound decision.

The next significant improvement in your manufacturing business might not be on the factory floor. It might be in the information factory that makes production possible.

Further reading

• Lean Enterprise Institute, Value Stream Mapping. On mapping material and information flows together.
• Galbraith, J. R. (1974), Organization Design: An Information Processing View, Interfaces, 4(3), 28–36. On organisational design and information-processing requirements.