Why Do Industrial Vision Companies Make Less Money the More Orders They Take?

Many industrial vision companies make less money the more orders they take.
Doesn't that sound strange?

A project worth several hundred thousand: the customer is willing to pay, and the company successfully signs the contract. But after the project is completed and the accounts are calculated, there is little profit left – sometimes even a loss.

Why?

Because the industrial vision industry has a very big contradiction:

Customers want customisation, while companies need standardisation.

01Customised Projects: The Harder They Are, the More They Pay – and the Harder They Are to Profit From

For example.
Reflective steel plate inspection, transparent part inspection, flexible material defect detection, complex assembly inspection…
These scenarios are often the hardest to do, but customers are most willing to pay.
Because others cannot solve them.

The problem is, these projects have almost no standard answers.
Algorithms must be retrained.
Light sources must be redesigned.
Mechanisms must be readjusted.
On-site debugging must be continuous.
In the end, a project consumes a large number of engineers; the revenue looks good, but the profit is not high.

On the other side, all industrial vision companies are desperately pursuing standardisation.
Standard cameras.
Standard software.
Standard algorithms.
Standard solutions.
Because only standardisation enables scalable replication.

But reality is cruel.

The demands that truly create value are often not the 80% that have already been solved.But the remaining hardest 20%.

Especially in fields such as new energy, semiconductors, aerospace, and high-end equipment manufacturing.
What customers are truly willing to pay for is precisely the scenarios that others cannot handle.

02The Past Competed on Algorithms; the Future Competes on Industry Understanding

So industrial vision is entering a new stage.

The past competed on whose algorithm was stronger.The future competes on who better understands the industry.

Many people are still researching how to improve YOLO accuracy by 1%.
But truly competitive companies have already begun researching:
how to understand battery processes.
how to understand PCB manufacturing.
how to understand semiconductor defects.

Because algorithms are rapidly becoming platformised.
Industry knowledge, however, is becoming increasingly scarce.

With the development of vision large models and multimodal models, previously one project corresponded to one algorithm engineer.
In the future, one industry model may serve hundreds of projects.
Algorithm development costs will become lower and lower.
And the value of industry know-how will become higher and higher.

03The Future Is Not 100% Standardisation, but 80% Platformisation + 20% Industry Customisation

So the future industrial vision industry will not move toward 100% standardisation.
Instead, a new model will emerge:

80% platformisation.20% industry customisation.

Platformisation is responsible for reducing costs.
Customisation is responsible for creating profits.
Algorithms are the foundation.
Data is the fuel.
Industry knowledge is the true moat.

04The Higher the Level of Manufacturing Automation, the Harder the Remaining Problems

Many people think industrial vision has entered a red ocean.
But what is seen is quite the opposite.
The higher the level of manufacturing automation, the harder the remaining problems.
And these hard problems are precisely the greatest opportunity for industrial vision in the next decade.

So the most valuable companies in the future may not be the ones with the strongest algorithms.
But rather companies that simultaneously possess algorithm capability, engineering capability, and industry know-how.

Because only in this way can problems that others cannot solve be truly turned into solutions that can be delivered, implemented, and profitable.