Every manufacturing company has a list of processes to “automate” that is longer than its budget. Usually the first one on it is the one that failed most recently, or the one a system vendor talked about. Both criteria are wrong. A good first process is one where someone today retypes data between two places, where you can count how many times a day, and which can be completed in a few weeks. Below: how to find it and how to measure the result with numbers the company collects itself.
Where to look for candidates
Automation pays off where there is repetition and retyping. Walk through a working day of three people: the planner, the storekeeper and the shift manager, and write down every situation in which someone:
- retypes data from paper or from one screen to another (a shift report into the ERP, an order from an e-mail into the system, an issue from a notebook into a spreadsheet),
- waits for information someone else already has (whether the material has arrived, whether the order is closed),
- calculates something in a spreadsheet that the system could calculate itself (rejects, downtime, tool consumption),
- collects signatures or approvals by e-mail (leave, a purchase, a change to an order).
After one day, the list has a dozen or so items. These are candidates, not a “digital transformation”.
Three selection criteria
For each candidate, answer three questions with numbers you collect yourself:
- How many times a day does it happen? Count it for a week. A process that happens once a month is not the first candidate, however annoying it is.
- How many people and how many minutes each time? Measure it on a few cases, don't estimate. Retyping a production report is rarely “five minutes”; with correcting errors and looking for missing data, it is often much more.
- What happens when someone makes a mistake? An error in the stock level costs a line stoppage; an error in the training records costs you during an inspection. Write down what happened last time.
The first process is the one that wins on points 1 and 2 and has a clear answer to point 3. Not necessarily the biggest; the most frequent and the most measurable.
What not to choose to begin with
- A process that goes through five departments and needs the approval of each of them. The first project should have one owner.
- A process whose data sits in a system nobody has access to (an ERP without an interface, a vendor without support). Access first, then automation.
- A process that is itself disorganised: everyone does it differently. First agree on one way of doing it, then the tool. Automating a mess gives you a faster mess.
- A process nobody on the shop floor wants to change. Without a person who will use it and fight for it, the project dies after rollout.
How to calculate the effect before you start
The measure of success must be set before the start and collected using the same method after rollout. Measures that work:
- the number of manually retyped items per day (before: counted in the measurement week; after: the same),
- the time from the event to its entry in the system (before: the night shift report entered in the morning; after: at the moment of the event),
- the number of corrections and fixes in a month,
- the number of “is it in the system yet” phone calls.
The formula for time saved is simple: number of events per day × minutes per event × number of working days. Put in your own numbers. Don't put in percentages from a vendor's presentation, because nobody measured them at your company.
On top of that there is a cost that cannot be calculated in advance but can be described: what the company gains when data is in the system at the moment of the event. The planner sees the machine status without a phone call. The manager sees rejects as they happen. This becomes apparent after rollout, and it is worth recording as an observation, not as a number.
How to run the first project
- An audit of one process: who, what, when, in which tool, what goes wrong. One day with the people who do it.
- A decision on the form: integration of existing systems, a small application, a retyping robot, a panel at the machine. The simplest thing that closes the process.
- A pilot on one line or in one warehouse with the same measure as in the audit. With the option of rolling back without losses.
- A decision on extending it based on measurement, not impressions.
A pilot that fails is also a result: the company knows what not to do, for a fraction of the cost of a full rollout.
Pre-start checklist
- One process selected, with one owner on the company side.
- Counted: frequency, time, cost of a mistake.
- An agreed measure of success and a “before” measurement week.
- Access to data in the systems the process touches.
- A person on the shop floor who wants the change and will test it.
- A plan for rolling back the pilot.
This is how we start at ErgoFlow: with an audit of one process and a pilot that can be measured and stopped. We describe ERP, MES and WMS integrations, workflows and custom applications on the Automation page, and examples of applications built for a specific process on the Applications page.
This article is for information only. The frequency of events, time and cost of mistakes vary between companies; we only give specific numbers after a process audit.
