Solution · Data evaluation
Spot deviations in your data before they cause problems.
Our data analysis and quality control solution compares measured values, finds unusual results and helps tell a measurement error from a real deviation. Instead of going through spreadsheets by hand, experts get an overview of what to investigate. Each hypothesis has a validation criterion defined in advance.
Typical complexity level Level 3 · Complex · from 6,490 EUR excl. VAT
When it makes sense
- Quality control means manually comparing large numbers of measurements
- An unusual value could be a measurement error or a real problem
- Data sits in several systems and is not comparable
- You need to back decisions with numbers, not guesswork
How it works
- 01
We collect the measured data.
Values and their context from existing data sources.
- 02
We connect and align the data.
Python combines the data and creates comparable groups.
- 03
Models find unusual values.
Statistics, anomaly detection (Isolation Forest) and classification of deviations.
- 04
Experts get an overview of what to check.
A dashboard in Power BI or a web application for expert review.
Validation and security
The solution runs in the client's environment. We measure model accuracy on historical data, and it only goes into operation after expert validation.
In practice · Data evaluation
Data deviations under control
The model compares measured data and helps distinguish a measurement error from a real deviation. The team gets an overview of what to check.
Pilot and validation
Starting point
An unusual value could mean a measurement error or a real problem. Telling them apart required manual checks and data comparison.
Goal: Detect suspicious values earlier and verify when there is a real deviation.
Solution
A data pipeline prepares comparable measurements. Statistical methods and machine-learning models look for anomalies and help distinguish a measurement error from a real deviation.
Results
- A first application makes unusual values visible.
- Each hypothesis has a documented validation criterion.
- Model accuracy and practical use are being validated in the pilot.
The pilot aims to reduce manual checking. Deployment follows only after the results are validated.
Technology
- Python, pandas and scikit-learn
- Isolation Forest and Gradient Boosting
- PostgreSQL
- Power BI
Security and data
- Runs in the customer's environment
- Access for authorised expert roles
- Documented criteria and accuracy measurement
Operations
- Regular joint evaluation
- Expert verification before deployment
- Further operation based on validation results
This case study is anonymised. We do not name the company or its industry.
Frequently asked questions
Do we need a lot of data?
It depends on the task. During the assessment, we judge whether the data is sufficient and, if necessary, suggest how to start collecting it.
Will the model replace the expert?
No. The model shows what to investigate; the decision stays with the expert.
You might also be interested in
Solution · Stock planningDemand and stock forecasting
Demand forecasting from historical data and seasonality. Recommended quantities per item and location 4 to 52 weeks ahead. In one project, revenue grew by 5–6%.
Level 3 · from 6,490 EURMore PricingPricing
Fixed prices for the AI assessment (990 / 2,490 / 4,490 EUR) and for automation across four complexity levels (from 990 EUR). Assessment credited to your pilot.
MoreNot sure where to start? That's fine.
Get in touch. In a short call, we'll find out what takes up most of your time today and suggest a first step. The consultation is free and with no obligation.