How to Choose the First Process to Automate
By DBR77 Team · Published · Updated
The best first process to automate is repetitive and stable, runs at steady volume on more than one shift, causes a pain the plant already feels (a bottleneck, heavy lifting, quality complaints), can be described with real numbers, and is small enough for one budget and one owner. It is rarely the most visible project in the plant. To find it, make a long list from the shop floor, score each candidate on six criteria, drop the ones that fail basic checks, and confirm the winner with data before you ask suppliers for offers.
What the first project has to achieve
A first automation project has two jobs. It has to pay back on its own, and it has to teach the plant how to buy and run automation. The second job is easy to underestimate. In McKinsey's survey of industrial leaders, 61 percent named a lack of automation experience as a barrier to adoption. The first project is where that experience is built: writing a brief, comparing offers, running FAT and SAT, training maintenance.
That is why the NIST Manufacturing Extension Partnership, in its Automation 101 guide, advises starting with "something you can build upon" and with the "dull, dirty, difficult, dangerous" tasks nobody wants to do. A clear first win makes the second project easier to approve, and a troubled one makes every later project harder.
Make a long list from the shop floor
Start with five to ten candidates. Good places to look:
- bottlenecks, the stations the line waits for on peak days
- overtime and extra shifts, where the plant pays to catch up
- heavy or awkward work: palletizing, sack handling, machine loading, lifting above shoulder height
- the station new people leave first
- the sources of scrap, wrong labels or customer complaints
- internal transport routes where operators push carts or drive forklifts between fixed points
Ask shift leads and maintenance as well as managers, because they know which station stops the line and why. Where machines already have controllers, a few weeks of machine data shows real cycle times and stop reasons better than opinions. The First 30 Days of IoT in a Brownfield Factory describes how to collect it.
Score the candidates
Score each candidate from 1 to 5 on six criteria. Higher is better on every line.
| Criterion | Scores 1 | Scores 5 |
|---|---|---|
| Repetition and stability | Method varies by shift, many exceptions | Same task every cycle, written standard work |
| Volume and shifts | Low volume, one shift | Steady volume, two or three shifts |
| Pain | "Would be nice" | Bottleneck, injuries or complaints the plant already tracks |
| Data available | No cycle times, no product data | Cycle times, volumes, product sizes and weights on file |
| Interfaces | Many machines and IT systems to connect | One or two simple signals to the line |
| Size | Needs board approval and a new building | Fits one budget, one owner, one area |
Before you add up the points, run three knockout checks. Is the process about to change (new product, new packaging, line move)? Is it unstable today? Is there no one who will own it? A "yes" to any of them removes the candidate for now; When Not to Automate (and When Waiting Is Right) lists what to fix first.
Illustrative example: a metal parts plant with 400 employees scores three candidates. End-of-line palletizing on two packing lines scores 26 of 30: very repetitive, three shifts, heavy lifting, data on file, one signal from the line. CNC machine tending with 14 part numbers in small batches scores 19, mainly because of variants and changeovers. Final visual inspection of painted parts scores 21: the pain is high, but nobody has written down what counts as a defect. Palletizing goes first. Inspection becomes a candidate for the second project once the defect catalog exists.
Found your first candidate?
Describe it as a Technology Challenge on DBR77 Marketplace and compare concepts from integrators in one format. Free for manufacturers.
Check the winner with data before you commit
Scoring shows where to look. On a connected line the obvious station is not always the constraint, so confirm where the loss is before you commit.
A plastic injection molding plant in central Poland, making components for household appliances, looked at automating part of its operations. The plant has over 200 employees, works three shifts and covers 20,000 m², including an 8,000 m² warehouse. It built a digital twin of a selected part of the plant in DBR77 DT, with material flows and task times, and compared alternative automation variants, each with an economic analysis.
The simulation showed the bottleneck at a manual process done by the first operator: 90 seconds of activity and 98 seconds of waiting. In the simulation, replacing that operator with an autonomous cart increased efficiency by over 100%. The variant analysis for the area took a few days. With the target transport automation concept in hand, the plant could efficiently select a technology provider on DBR77 Marketplace. The full story is in the case study.
The first operator was waiting more than half the time, and that only showed up once the whole flow was modeled instead of a single station.
Traps when choosing the first process
The prestige project. The most visible line attracts sponsors, and also scope creep and spectators. Save it for when the team has done one project.
The hardest problem. The process with the most variants, the tightest tolerances or the worst quality is tempting because it hurts most. It is also where a first team is most likely to fail.
Starting from the technology. "We want a cobot" or "we want an AGV" picks the answer before the question. Pick the process, then let suppliers propose.
The whole hall at once. A plan that covers ten stations needs ten sets of data, many approvers and a budget that is hard to defend.
From choice to project
Once the candidate is chosen, turn it into something suppliers can price. Write one paragraph that states the problem and the goal. Collect the baseline: volumes per shift, current and target cycle time, products with sizes and weights, variants, floor space, existing machines and controls. Name the owner and the approvers.
Then draw the scope line with How to Scope an Automation Project Without Overcomplicating It and write the brief with How to Write an Automation Brief Integrators Can Quote.
How DBR77 Marketplace helps
The baseline you collected for scoring goes straight into a Technology Challenge on DBR77 Marketplace, whose guided form ends with operational assumptions such as shifts, operators and cycle time. DBR77 experts review the challenge before it goes live, and integrators from a network of more than 800 technology suppliers and integrators answer with concepts first and offers later. To check where the constraint is before you publish, model the flow in DBR77 DT, as the plant in the case above did.
FAQ
Should the first project be the one with the fastest payback?
Not necessarily. A project with a fast payback on paper and many unknowns can fail, so prefer a slightly slower payback on a stable, well-described process.
How many candidates should we score?
Five to ten is enough. Fewer and you may miss a good option. More and the exercise turns into a study.
What if two candidates score the same?
Pick the one with better data and fewer interfaces. It will be easier to describe, quote and accept.
Conclusion
Build the long list with shift leads and maintenance, score the candidates and drop any that fail the knockout checks. Confirm the top scorer with data, then write the brief for that one station.
Start with one station
Post your first Technology Challenge on DBR77 Marketplace, or talk to our team about which candidate to describe.
Post your Technology Challenge · Contact the Marketplace team
Sources
- McKinsey & Company, Unlocking the industrial potential of robotics and automation, 2023 (survey conducted in August 2022)
- NIST Manufacturing Extension Partnership, Automation 101: How to Plan for a Successful Implementation, 2024
- DBR77, Case study: Development of an automation concept, 2024