Transcript Also available as a Powerpoint document

How to profit from your
investments in data collection
systems
“Identify the root causes to delivery problems in
your manufacturing process with BayMiner QVM
(Quality Variance Management)
Ralf Ekholm
CEO
Bayes Information Technology Ltd.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
What is it all about?

The data analysis market is changing:
1. Data mining is not sufficient anymore.
2. Classic reporting is replaced by easy, fast multi-dimensional analysis methods.
 You can identify hidden reasons to variations in punctuality and completeness.


For managers:

A method to identify which of several variable combinations cause problems.

A method to get a realistic picture of the reasons to e.g.punctuality problems.
For users:


A new method to know better before deciding.
BayMiner QVM is NOT:

A calculation tool.

A reporting tool.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
What is new?
 Assumption: At present you get punctuality curves per main component/time
 It is impossible to recognize if behind delays are some unknown major trends.
 BayMiner QVM reveals them for you if the necessary information hides in your
data. (if not, you must further develop your data collection)
 When situation regarding one module gets better another gets worse.
 BayMiner QVM can help you predict punctuality problems.
 Future steps
 Subcontractors should provide data together with their product or service in such
format that you can utilize it for process development without high fixed costs.
 The form of the data should be such that you can use it to predict e.g.
punctuality problems.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
Advantages and Benefits BayMiner
QVM offers for quality development

You can identify root causes to various quality problems:
 Why some variation combinations are more difficult than other.
 Avoid up to 50 % of delayed deliveries.

You can utilize company knowledge effectively:
 Share knowledge over organizational borders.
 Avoid the use of scarce resources for unprofitable tasks.

You get a second opinion when you plan loading.
 Identify and correct sales budget variations.
 You can keep delivery commitments better.

Reveals hidden phenomenon in testing data.
 Speeds up corrective actions.
 Shorter time-to-market for new products.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
Familiar problems?
 Sales promise combinations that are too difficult to
realize (notwithstanding that you have a configurator
that should help sales to avoid such cases.)
 Networking has brought new risks.
 Your statistics are not trustworthy.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
These problems can be solved:
 With the BayMiner QVM method that:
 Elicits knowledge from sparse data.
 Presents information in an easily understood way.
 BayMiner PRO is a decision support development tool that:
 Learns from data about operations in the past.
 Visualizes problem clusters.
 Indicates the probable causes and their co-influences.
 BayMiner QVM includes a special version for on-line use.
 Easy to integrate - operates via the company's intranet.
 Highly visual - indicates results with simple traffic lights.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
Process example: How to compare two visually
defined groups with BayMiner QVM

The process
1. Check that the distribution of values for most important variables is sensible.
2. Identify important groups in BayMiner’s visualisation.
3. Select a sample from the middle of each group and name the selection.
4. Compare two groups by putting them against each other multi-dimensionally.
5. Look at the distribution difference indicators for possible hidden information.

Results and benefits

If you want to compare successful & less successful product applications
BayMiner’s visual representation immediately verifies that you have a
homogenous group.

If the visualisation shows that you do not have a homogenous group, but that e.g.
a statistically uniform group consists of two different subgroups your conclusions
for actions are very misplaced.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
Useful links
 http://www.bayminer.com/
 http://cosco.hiit.fi/
 the research group behind it.
 http://www.bayminer.com/files/papersetc/bnets.pdf
 theory, pretty heavy.
 http://www.kdnuggets.com/
 the most comprehensive Data Mining and Knowledge Discovery
site.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007
Thank you for your interest!
Bayes Information Technology Ltd.
Porttikuja 3 C
FIN-00940 Helsinki
tel. +358-9-72892680
www.Bayminer.com
CEO Ralf Ekholm
tel. +358-50-5497109
e-mail: [email protected]
We are a Finnish HiTech company.
Tekes (National Technology Agency) has supported development.
Academy of Finland has supported research in Bayesian Networks.
BayMiner QVM for Executives
© Bayes Information Technology Oy 2007