Back to all posts
Projects & Retrofits7 min read·November 15, 2026

Which Plant Process Should You Automate First? A Scorecard

TB
ThynkBlox Automation Team
Automation Engineering

Automate first the process that repeats often, costs you real money or downtime when it goes wrong, already runs in a stable and well-understood way, and has data you can trust. To find it, score each candidate on those points plus safety impact, then compare the totals. Do not start with a process the team cannot yet describe consistently. Automation repeats whatever you build into it, and that includes the problems.

Why use a scorecard instead of picking the loudest problem?

Most plants have more automation candidates than budget. The process that gets chosen is often the one discussed at the last production meeting, or the one a supplier happened to demonstrate. Neither is a good reason.

A scorecard makes the team say why one process should go first. It also keeps the record of that decision. When someone asks six months later why the packing line was automated before the batching area, you can show the reasoning.

The scorecard does not replace engineering judgement. It gives that judgement a structure, so the team argues about evidence rather than opinions.

What criteria should the scorecard include?

Use five criteria. Score each from 1 (low) to 5 (high), in the direction that favours automating first.

CriterionWhat it asksScores high (4–5) whenScores low (1–2) when
Repeat frequencyHow often does the task or sequence run?It runs many times a shift, in much the same way each timeIt runs occasionally or differs each time
Cost of errors or downtimeWhat does a mistake or stoppage cost in scrap, rework, lost output or customer impact?Errors are frequent or expensive, or stoppages hold up downstream areasErrors are rare and cheap to put right
Process stabilityIs the process defined, repeatable and agreed by the people who run it?Steps, setpoints and sequence are documented and followed consistentlyOperators run it differently, or it changes often
Data availabilityCan you measure the inputs, outputs and states the automation needs?Signals already exist, or can be added easily, and are trustedKey values are guessed, recorded by hand or missing
Safety impactDoes automating reduce people's exposure to hazards, and how complex is the safety design?It removes manual handling or exposure, with a manageable safety designIt needs significant safety-related design with no clear exposure benefit

Safety impact needs care. A process that reduces exposure to a hazard is a strong candidate. Any change that touches safety functions still needs a risk assessment with a qualified assessor. A high score here means it is worth doing, not that the safety work can be skipped. The guide on planning a safety assessment for a machine retrofit covers that step.

Should the criteria carry equal weight?

Not always. Agree the weights before you score anything, so they cannot be adjusted afterwards to favour a candidate someone already prefers.

A reasonable default is to count process stability double. A stable process with average scores elsewhere is usually a safer first project than an unstable one with high scores everywhere else. Some plants give the cost criterion extra weight because their customers or production plan are especially sensitive to late or faulty output.

Stability should also act as a gate. If a candidate scores 1 or 2 on stability, take it off this round's list, whatever its total.

What does a worked example look like?

Illustrative example: a plant shortlists three candidates. The scores below are invented to show the method. They are not measurements from any real site.

  • Candidate A: case packing and palletising. Operators load cases by hand at the end of a filling line.
  • Candidate B: ingredient weighing and dosing. Operators weigh minor ingredients by hand against a paper batch sheet.
  • Candidate C: changeover of a forming machine. Settings are adjusted by experienced staff, and each person does it a little differently.

Weights: stability ×2, all others ×1.

Criterion (weight)A: PackingB: DosingC: Changeover
Repeat frequency (×1)542
Cost of errors or downtime (×1)354
Process stability (×2)4 → 84 → 82 → 4
Data availability (×1)432
Safety impact (×1)432
**Weighted total****24****23****14**

A and B are close. In this case the team would look past the totals. Dosing errors could put a whole batch out of specification, and the paper batch sheet means the plant is already short of data on that process. Packing has the higher total and offers a clear benefit in manual handling. The team might start with packing as the lower-risk first project and add measured weighing to the dosing area alongside it, so that data is ready for a second phase.

Candidate C falls at the stability gate. Its cost score is high, which shows the changeover matters. What it needs first is a standard method, not automation.

Which processes should you leave alone for now?

Leave out processes that are unstable or poorly understood. Warning signs include:

  • Different shifts run the process differently and each believes their way is correct.
  • Nobody can write the sequence down without saying "it depends" at most steps.
  • Quality problems have no agreed cause yet.
  • The process, product or equipment is due to change soon.
  • The process relies on one person's knowledge that no one else has written down.

Stabilise these processes first. Document the steps, agree the setpoints, find the causes of variation and collect some data. Mapping the process and writing it down often brings much of the improvement people hoped automation would deliver. Once the process is stable, score it again.

Also hold back candidates where the control equipment underneath is near the end of its supported life. That decision belongs to a separate review. The guide on prioritising obsolete plant control equipment explains how to rank that risk, and the two lists should be read side by side.

How do you collect honest scores?

Score with the people who run and maintain the process, not only with managers. Operators know how often a task really happens and where it goes wrong. Maintenance staff know which signals can be trusted.

For each score, write one line of evidence, for example "runs roughly every cycle, per shift log" or "weights recorded by hand, no electronic record". A score with no evidence behind it is a guess and should be marked as one.

Where the cost of errors is uncertain, use a range or a rough order of magnitude rather than a precise figure no one can defend. The scorecard compares candidates, so consistency matters more than precision.

What happens after you pick the first process?

Choosing the process is the start of the project, not the end. Next steps usually include:

  1. Write down the current process and what the automated version must do.
  2. Check that the area is ready for installation: power, space, network, access and drawings. See checking site readiness before a controls installation.
  3. Decide who will design and deliver the work. The guide on choosing a system integrator for a plant project explains what to look for.
  4. Agree how success will be judged before work begins, using measures you can check once the system is running.

What mistakes should you avoid?

Automating to fix an unclear process. Automation makes a process consistent. It does not decide what the process should be.

Choosing the most ambitious project first. A first project that runs late damages confidence in everything that follows. Start with something that can be delivered cleanly.

Scoring alone. One person's view misses frequency, failure modes and data gaps that the floor team would spot straight away.

Changing weights after scoring. If the weights change, rescore every candidate and record why.

Ignoring data availability. A process with no reliable signals often turns into a sensor and instrumentation project first. That can be the right call, but plan it as such.

Treating the total as the decision. Close totals need judgement. Use the scorecard to support the discussion, not to end it.

Frequently asked questions

How many candidates should we score?

Score enough candidates to make a real comparison, often a handful. Including too many spreads the floor team's time thinly and weakens the evidence behind each score. Shortlist using a quick discussion first, then score the strongest options properly, with one line of evidence recorded against every criterion.

Can an unstable process ever be automated first?

Rarely, and only when there is a strong reason, such as removing people from a hazard. Even then, stabilise the method as part of the project and document it before design begins. Otherwise the team designs against a moving target, and commissioning becomes an argument about what the process should do.

Should the scorecard include cost and payback?

Keep project cost as a separate step after scoring. The scorecard ranks which processes are worth automating; costing then tests whether the leading candidates are affordable. Mixing the two too early tends to favour cheap, low-value projects and hides the difference between a good candidate and an inexpensive one.

How often should we revisit the scorecard?

Revisit it after each project finishes, and whenever a product, line or major piece of equipment changes. Processes that failed the stability gate may now qualify, and data added by earlier projects can raise other candidates' scores. Keep earlier versions so the reasoning behind each decision stays visible.

Next step

Want to apply this to your plant?

Every project starts with a fixed-fee assessment, and you can stop after the report.