New MDM object and Matching in MDM

[Former Member]

Hi, we are just considering whether to use MDM in our project. We have no experience with MDM so far. Following questions have emerged during this consideration:

1. If I want to create my own new object, what built-in services can be (re)used for that new object? Can I use any of the built-in existing matching algorithms? Can I use search mechanisms for that new object? Can I reuse any UI component for maintaining that new object?

2. Is it possible to inherit from a built in object like business partner and extend it with some new attributes? If so, again what services defined for business partner can I reuse? How can I define new matching algorithms, search mechanisms and presentation for the new part?

3. What does exactly the notion “reference object” in the matching process mean, and how is it selected?

4. In case of an automatic match of the objects A and B, how will the attributes of A and B merged together? Which one will be overwritten by the other one? In case of a cleansing case, does the MDM specialist have the decision to merge the attributes arbitrary? Is it possible to link a new object to a master but not to merge it (i.e. let the new object as it is, maybe slightly different from the master it is linked to)?

5. A new object A is inserted in the MDM system, which causes a cleansing case. Let’s assume that the MDM specialist solves this case only on the next day. Is this record available already for the clients or it cannot be accessed until the cleansing case has been released? In the latter case: is it possible to change CI in the way that in case of a cleansing case a new temporary master will be created which can be seen immediately by the clients and later, when the manual match has been made, this master will be deleted?

6. What services are available to search among the master records? How can I extend them?

7. Where can I find detailed documentation about MDM and about all the processes and components of it?

Thank you for your answers in advance

[ch_loos · Accepted solution]

Hi,

I'd just like to add a couple of things:

4) In case of an automatic match between two MDS objects, the selection of the new reference object is random. However, you can customize this with your own implementation.

5) The process will remain in step manual cleansing until all cleansing cases related to this process have been closed. In the post step the object will be created on the MDS, and then distributed to clients.

Regards,

Christian

[Former Member]

Hi Markus,

there are some problems with ILTC:

a. I do not only need such an insert in the initial load but rather in the normal operation of the system

b. I do not want to use the Central Data Management scenrario, and according to the docu it is part of it

can you comment on that?

A question belonging to it:

Would it be possible not to match at all at an insert but rather make a database cleansing periodically--lets say every day once? Do you have some experience on the performance of such an overall DB cleansing?

Thanks, Andras

[Mark63]

Hi Andras,

Good question.

I assume leaving out the matching may be achieved by setting process chains accordingly, but since this is just an assumption, I'll forward this to the experts.

Stay tuned.

Regards, Markus

[KlausDavid]

Hi Andras,

with MDM 3.0 there was a new process type (Cleansing on the Database introduced). Please look in the online documentation under process types for "Data Cleansing"

You have to create a process Chain, which includes the process step "Data Selection". This process step allows you to define a selection variant to either select specific objects on the MDS DB, or also to include all changed objects since the last run.

Concerning performance I have no experience.

Regards Klaus

[Former Member]

Hi all,

question 5 again:

I have just read about Straight-Through Processing (STP). Could that be a solution to the problem in question 5? According to the online docu, during STP the inserted new data will be stored in MDS and distributed to the clients immediately, no manual decision is required. "Cases where a user’s decision is normally required (for example, 80% matches) are ignored. This means that the cases are treated as if there is no match at all."

Corresponding questions:

5.1 Can the matching be triggered later after STP?

5.2 I do not really need distribution, just to see consistent data in the central store immediately. Can STP still be used (because distribution is an essential part of STP)?

Thanks, Andras

[Mark63]

Hi Andras,

There is a process called "Initial Load from Trusted Client" (see http://help.sap.com/saphelp_mdm300/helpdata/EN/b5/4cda406478dd50e10000000a1550b0/frameset.htm) which involves data cleansing (see http://help.sap.com/saphelp_mdm300/helpdata/EN/80/aada4023a28631e10000000a1550b0/frameset.htm) at a later point in time.

Could this be an option?

Regards, Markus

[Former Member]

Thanks for the answers.

to question 6: is TREX also a database search engine, where eg I can select all business partners where name=SAP* or alike? If yes, is it fast enough? Is it transformed to an SQL query at the end? I know only the google-like text-query side of TREX from the online docuementation..If NOT, what supports the database query in the master data DB?

[Mark63]

Hi Andras,

With regard to question 4, please be informed as follows:

Strategies for automatic matching/merging can be defined in the Content Integrator. SAP delivers preconfigured, customizable examples. See also

http://help.sap.com/saphelp_mdm300/helpdata/EN/9c/860a3369454f409cc61215d8b76bd2/frameset.htm.

Additional information about CI matching strategies can be found in the Service Marketplace under the quick link <i>/instguidesnw04/</i>. There, choose <i>Operations</i>.

When it comes to data cleansing cases, the Master Data Manager matches data and merges attributes according to his or her expertise.

Regards, Markus

[Former Member]

Thank you for the answers. It would be great if the other experts mentioned in your answer could also react on the missing questions.

Thanks, Andras

[Former Member]

Hi,

I might give some input to a few points, but I can't cover everything.

1)

there is the Master Data Framework, Trans. MDF on MDS,which allows you to create own objects. As far as I know, the matching and cleansing strategies won't work for these own objects - but distribution works. So if you follow the scenario central mdm you're fine. Searching for the objects is supported.

2)

the easy enhancement workbench covers that - check the mdm enhancement guide. it is possible to extend objects.

look at http://service.sap.com/instguides go to netweaver04 installation --> documents: mdm 3.00 - extending the product and business partner

see also: help.sap.com

-->SAP NetWeaver. Enter the search term Easy Enhancement Workbench.

3. Reference Object is created in matching process, on the CI I reckon... its being used, when matching candidates appear - in my understanding you get a cleansing list, if the matching determines possible dublettes and then you determine the reference object - i am not 100% sure. the reference object is being used for later matches and is being distributed to depending MDCs

4. good question - I hope Mr Markus Ganser can comment on that or Klaus David - so much: you can determine thresholds for the matching eg: matching probability < 75% = no match; between 75 and 95 = manual decision; above 95% automatic release, if the latter happens I am not sure how the reference object is being selected. You can adjust the settings and the matching algorithms - but everybody keeps telling, the standard rocks.

Concerning linking: in the mdm scenario consolidation you leave objects like they are in the MDCs - only the knowledge about redundancies is being created - on the MDS the key-mapping info is available (eg object A in System 1 matches object Aa in System 102)

5) I have no clue, only a guess: i assume you have to get the cleansing case clear - otherwise no data available for depending systems...

what i know is: you have to get all the entries from a cleansing case list (which might appear when loading great amounts in one go) before any of them get released. so if you have like 1000 cases to decide manually - enjoy!

6) the TREX is the search engine in mdm - pretty fast thing as it is not written in abap...

you have database search and a search engine service (google-like engine search) - for both searches you get a quick and advanced search - searches can be reused (selection criteria). search is also provided as a webservice

7)

help.sap.com

service.sap.com/instguides and howtoguides

esp. config guide: harmonization nd central mdm

and master guide

hope you get along,

cheers matthias kasig