OPOP
OPOP is Austrian soft sensor software for the process industry. It trains machine learning models on historical plant data to predict critical parameters in real time, helping operations teams cut production costs without additional hardware or CAPEX.
What is OPOP?
OPOP - short for Operational Optimization for Process Industry - is soft sensor software for continuous process plants. A soft sensor is a machine learning model trained on a plant's historical process data: from the measurements that are available continuously, it infers a quantity that is not, such as a product property normally checked in a laboratory or a value read by a physical analyser. The vendor frames this as finally putting existing process databases to work, with no additional hardware and no CAPEX, because the value comes from data the plant already stores.
The software is deployed on-premise and designed as self-service, aimed explicitly at plant engineers and operations teams rather than at data scientists or external consultants. The product screenshot published on the home page shows a Windows desktop application, version v0.1.0, organised around three tabs: Models, Training and Databases. A Model Details panel exposes the model name, upload date, accuracy (84.69% in the published example), model type (random_forest), status and the field being predicted, in that example cold filter plugging point in degrees Celsius. A Live Model Prediction Graph overlays the real value and the prediction, with a configurable number of points and a refresh interval set to 10 seconds in the example. Two further charts, Performance on Training Data and Performance on Test Data, compare actual and predicted values across the history, so the engineer can judge a model before relying on it.
The White Paper claims three benefits: lower costs through better use of resources and energy, lighter quality monitoring, and more flexible, more robust processes thanks to a digital backup measurement and smarter monitoring. Three use cases are published with figures attached: desulphurisation, furnace outlet temperature and flue gas analysis.
Behind the product are three founders based in Vienna: one from process engineering and operational optimisation, one software engineer working on complex connected systems, and one physicist specialised in machine learning. The site footer displays support from Wirtschaftsagentur Wien, INiTS and the Sustainability Challenge. No trading company is named anywhere on the site.
What it does
- Predict a critical process parameter in real time from data the plant already records
- Replace a failed analyser with a digital backup measurement so the unit stays controllable
- Build a soft sensor from historical process data, in about 30 minutes in one published case
- Train and evaluate machine learning models directly from the application
- Connect the software to an existing process database
- Follow prediction against actual measurement on a live graph refreshed every 10 seconds
- Check model accuracy on training and test data before relying on a prediction
When to use OPOP / When not to
A quick filter to help you decide if OPOP is the right fit.
When to use OPOP
- Refinery and continuous-process teams that want to steer a unit on a live quality figure instead of a weekly laboratory sample
- Plant and operations engineers who want to exploit their own process data themselves, without hiring a data scientist
- Sites that need a backup measurement when a physical analyser fails, without buying a replacement instrument
- Organisations sitting on years of stored process data that currently produces no operational value
- Furnace and combustion operators looking to hold optimal firing when an outlet temperature or oxygen reading is unavailable
When not to use OPOP
- Buyers who want to sign up, download and evaluate a tool on their own: entry runs through a contact form or an introductory call
- Teams looking for a ready-to-use SaaS platform, since the software is described as an on-premise deployment
- Procurement processes that need a published price or a free tier before opening a file: no rate card, free plan or trial exists
- Engineering teams that require a documented API, named connectors or a mobile app, none of which is offered
- Buyers with strict vendor-maturity or compliance requirements: the product screenshot is stamped v0.1.0 and no terms, DPA or security page is published
How to use OPOP
A typical end-to-end flow, from setup to results.
- Read the White Paper, which downloads freely from the home page as a PDF without any form, to test the approach against your own unit
- Send the contact form on the home page (first name, surname, e-mail, phone, company, role) or request a first conversation, as there is no online sign-up and no public download
- Agree the scope with the vendor, then have the on-premise software installed on a workstation
- Log in to the desktop application
- Open the Databases tab and connect the application to your historical process data source
- Go to the Training tab and train a model on that history; the published example shows a model uploaded on 30/03/2025, type random_forest, status Finished
- Open the Models tab to review the model details: accuracy, model type and the field being predicted
- Read the Performance on Training Data and Performance on Test Data charts before putting the model into service
- Run the Live Model Prediction Graph, choosing the number of values displayed and the refresh interval, to follow prediction against actual measurement
- Repeat the cycle for each further parameter; one published case reports a soft sensor built in about 30 minutes
Pros & Cons
Pros
- Three concrete industrial use cases with figures attached: around 2% of hydrogen saved, up to 10% of fuel avoided, a soft sensor built in about 30 minutes
- The no-CAPEX argument is coherent: the value comes from process data the plant already stores, with no instrument to buy
- On-premise deployment keeps process data inside the industrial site
- Self-service design explicitly aimed at plant engineers, without a data scientist or an external integrator
- The interface exposes model accuracy and actual-versus-predicted curves, so the user can check the result rather than take it on trust
- The White Paper downloads without a form, and the founders publish individual e-mail addresses for direct contact
- Institutional backing is displayed: Wirtschaftsagentur Wien, INiTS and the Sustainability Challenge
Cons
- No published price and no pricing page: the cost of the software cannot be known before contacting the vendor
- No free trial and no free plan is announced
- No terms of service, no data processing agreement, no security page and no retention commitment
- The privacy policy is four short paragraphs and covers the marketing site only, never the customer's process data; no hosting country, certification or sub-processor list is given
- No API, no technical documentation, no mobile app, and no third-party system named anywhere (historian, DCS, OPC and the like)
- The German and English Impressum publish two different legal addresses, and no legal entity is identifiable: no legal form, no company register number, no VAT number, the Impressum naming a natural person
- Early-stage signals throughout: the product screenshot is stamped v0.1.0, three pages of the site are unedited Wix templates, one product page is empty, and there is no social presence
Pricing & Plans
No price is published for the product. There is no pricing page, no amount on the home page and no figure in the White Paper. There is no free plan and no free trial, so no entry price point can be quoted. The commercial model observable on the site is contact-sales: a form on the home page or a first conversation precedes any quotation. Readers should note that the amounts of EUR 80, EUR 120 and EUR 150 visible on three program pages of the site are unedited Wix template prices, carried alongside placeholder descriptions; they do not price the software and should not be read as an entry cost.
Data, GDPR & hosting
A consolidated view of how OPOP handles your data.
GDPR overview
OPOP never uses the words GDPR or DSGVO anywhere on its site, so it makes no compliance claim at all. The privacy policy runs to four short sections. It says personal data is collected and used only within legal requirements, lists the browsing data captured (IP address, browser type, operating system, access time), and recognises three rights: access, rectification and erasure, to be exercised through the contact details given in the Impressum, in practice office@opop.at. There is no named data protection officer, no legal basis cited, no retention period, no sub-processor list and no Article 27 representative. As the publisher is established in Vienna, Austria, the GDPR applies by law whatever the site says. The website itself runs on Wix infrastructure, a technical processor the policy does not disclose. No data processing agreement is offered anywhere.
Who owns the data?
The published privacy policy covers only the marketing website. OPOP states that it automatically collects an IP address, browser type, operating system and access time, and uses them solely for technical optimisation and security. Nothing on the site addresses ownership of the customer's process data: no terms of service, no licence agreement and no data processing agreement are published, so no written clause assigns rights over plant data to either party. In practice the software is described as on-premise, which means historical and live process data stay inside the customer's own systems. That is a consequence of the deployment model rather than a contractual commitment.
Reuse rights
For website data, OPOP declares a narrow use: technical optimisation and security, plus anonymised statistical analysis through analytics cookies that visitors can switch off in their browser. The policy states that data is not passed on to third parties unless the law requires it. Nothing is said about reusing visitor data for any other purpose, and the site takes no position, in either direction, on training models with customer data. Inside the product the logic is different: the software trains models on the customer's own historical process data, on the customer's premises, with model types such as random_forest and an accuracy figure displayed in the interface. That is the product doing its job, not the vendor exploiting data. Because no terms of service are published, end users have no written permission and no written restriction governing what they may do with the models or the predictions they produce.
Data retention & training
Hosting summary
The White Paper states that the OPOP software can be deployed on-premise, meaning it runs on the customer's own infrastructure and process data need not leave the industrial site. Beyond that, the site names no hosting country and no hosting region, for the product or for the company. No sub-processor list is published and no data processing agreement is offered. The marketing website itself is hosted on Wix (IP 185.230.63.107, AS58182 Wix.com Ltd., an anycast address), which indicates where the brochure lives and not where any customer data lives; the privacy policy does not disclose Wix as a technical processor. The publisher is established in Vienna, so Austrian and EU law is the applicable jurisdiction for the company, but the site makes no explicit statement about data residency or about the jurisdiction governing customer data. A buyer who needs a documented hosting arrangement will have to obtain it directly from the vendor.
Things to keep in mind
Risks and trade-offs to weigh before adopting OPOP.
- The legal identity is not established: the Impressum names a natural person, with no legal form, company register number or VAT number, and no OPOP entity was found in the Austrian company register or in GLEIF. The Czech boiler manufacturer OPOP spol. s r.o. is an unrelated namesake.
- The German and English Impressum publish two different legal addresses (Media Quarter Marx 3.4, Maria-Jacobi-Gasse 1/5. Stock, 1030 Wien, and Helene Thimig W. 8, 1230 Vienna). Both are stable in web archives, so this reflects two versions maintained in parallel rather than an update in progress.
- Three pages of the site are unedited Wix templates, including a team page presenting four fictional people and placeholder text. Nothing on those pages describes the real company, and a reader in a hurry could easily mistake them for staff information.
- The amounts of EUR 80, EUR 120 and EUR 150 found on three program pages are Wix template prices attached to placeholder descriptions, not the price of the software; one product page is empty, and three page titles still carry a leftover project name.
- The site's llms.txt file announces an MCP endpoint and API documentation. This is Wix platform boilerplate, not a product API, and quoting it as an integration capability would be a mistake.
- Maturity risk: the product screenshot is stamped v0.1.0 and the example model dates from March 2025. There is no public documentation, no support address and no social presence to fall back on, and the Supported by logos (Wirtschaftsagentur Wien, INiTS, Sustainability Challenge) are public bodies and an incubator, not customers or integrations.
- A soft sensor is an inferred value, not a measurement. The 84.69% accuracy shown in the vendor's own example is a useful reminder that predictions drift as a plant changes, and that operators can drop their guard when a screen keeps producing a plausible number. Treating a prediction as a safety-relevant reading, or letting physical instrumentation and human verification lapse because a model is running, would be a serious misuse.
Setup & Integrations
Technical difficulty
Moderate. OPOP describes its software as self-service, intuitive and integrable into existing workflows without an external service provider, and one published case reports a soft sensor built on historical data in about 30 minutes. In practice, reaching that point requires an on-premise installation and a connection to a process database through the Databases tab, so IT and OT involvement is needed, along with a usable history of process data. No public installation documentation, help page or support subdomain exists. Because access runs through a sales conversation, the vendor is present during setup.
Deployment
Behind OPOP
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What exactly is a soft sensor in OPOP?
Do we need to buy additional hardware?
Where does the software run?
How long does it take to build a soft sensor?
What savings does OPOP claim?
How much does OPOP cost?
Is there an API or a mobile app?
How does OPOP handle the GDPR?
Who publishes OPOP?
What kind of model does the tool use?
Should you pick OPOP?
OPOP makes a clear, testable technical proposition: train a machine learning model on process data a plant already stores, and use it to predict continuously a value that is otherwise sampled in a laboratory or read by a physical analyser. Three published use cases give the argument substance, covering sulphur content in a desulphurisation unit, furnace outlet temperature and flue gas oxygen, each with a figure attached. The on-premise, self-service design fits the stated audience: plant engineers who want measurement redundancy or fewer laboratory delays without a data science project or an external integrator.
Commercial maturity is another matter. There is no published price, no terms of service, no data processing agreement and no technical documentation, and the product screenshot is stamped v0.1.0. The showcase site itself is untidy: three pages are unedited Wix templates, one of them listing four fictional staff and others carrying placeholder program prices, one product page is empty, and the German and English Impressum publish two different Vienna addresses that have both been stable for months. No legal entity could be found either: searches of the Austrian company register and of GLEIF return nothing, and the Impressum names a natural person rather than a company.
The natural buyer is an operations team that already holds a usable history of process data, is looking for a digital backup measurement or a way to lighten laboratory testing, and is comfortable engaging a very young Vienna vendor through a direct conversation. Anyone who needs a rate card, contractual data commitments, named connectors or documentation before committing will find nothing on the site to work with. On its engineering claims OPOP deserves a call; on its paperwork it is still an early-stage supplier.
- Choosing a selection results in a full page refresh.
- Opens in a new window.