SP04 - Matching

[Former Member]

I have the following question related to matching.

I have a repository with 6 records. I have set no search criteria in the record mode of the repository and defined a simple matching strategy. I seem to get different results depending on how I perform matching...

When - in matching mode - I select one record and..

(1) Selected vs Results : I get 4 match-scores

(2) Selected vs All : I get 4 match-scores

(3) Results vs Results : I get 5 match Records.

(4) Results vs All : I get 5 match records.

I would expect that also for Selected vs Results and Selected versus All, I would get 5 match records. One for each record in the repository.

Any explanation?

[AndreasSeifried]

Hi Dirk,

depending on your "How" (1)-(4), MDM does compare different sets of records.

(1) only the selected record(s) against all in the result list (1:6)

(2) the selected against all in the repository: (1:6)

(3) all in the result list against all in the result list: (6:6)

(4) all in the result list against all in the repository: (6:6)

This explains the difference. (3)&(4) should typically give you more "hits" than (1)&(2).

It would be weird though, if your test would give more hits for (1) or (2) than for (3) or (4).

Hope this made it clear

Regards

Andreas

[Former Member]

Andreas,

I fully agree with your statement and this is how I tought this would work as well. However, for (1) and (2) I would still expect that I would get 5 match scores for my selected record and not 4. ( As this would mean my selected record is not matched to all others; as the results list contains all records.. )

Regards,

Dirk

[AndreasSeifried]

Hi Dirk,

your selected record is matched against all in the repository in (1).

Remember that you do not see any "hit" if the resulting score is below the "low threshold" defined in the strategy.

(3)&(4) give you additional hits because e.g. record 2 is matched against records 1,3,4,5,6. So alltogether there are n*(n-1) comparisons, with n=6 in your case.

And this might give you additional hits exceeding the low threshold.

Regards

Andreas

[Former Member]

Andreas,

I set-up the following test case.

Started with empty product repository.

Added one material, populating the Division Field and the Material Type field.

I copied the newly created record 4 times with the "Duplicate Record"

I have now 5 records in my database, all the same, except for my material number which I renamed into 1,2,3,4 and 5.

I have one matching rule, with Token Equals on the field "Division". Success score is 20, failure 5 and undefined 10 I have one strategy, min score5, max score 20, low treshold 0, high treshold 20.

As the records are all the same ( except for the material number, which I do not use in matching), I expect the matchhing behavious to be identical. But it isn't.

Simply using the option "Selected versus All"

-> I select record "1" and "Selected versus All" returns 3 matches. Record "1" is compated with records 3,4,5. It had no result with the comparison with record 2 ???

-> I select record "2" and "Selected versus All" returned 4 matches, with records 1,3,4,5. This is in line with my expectations.

-> I select two records "1" and "2" and using "Selected versus All" I get for each record 4 matches. Record 1 is matched to records 2,3,4,5. Record 2 is matched to records 1,3,4,5. Again in line with my expectations.

If all my records are identical, all my matching rules and strategies are identical.. why is there a difference in matching results? I could replicate the same behaviour with other matching fields, other number of products...

Regards

Dirk

[AndreasSeifried]

Hi Dirk,

I tried to reproduce following your description, but I always get consistent results, giving me 4 hits per selected record. All looks fine.

Are you using an old build ? I did this with 5.5.33.13.

Regards

Andreas

[Former Member]

Andreas,

I use 5.5.33.13. I just did the following and again the same results.

- Created a brand new repository

- Added one record ( name is populated with New Item)

- Duplicated the record 3 times ( so I have 4 records)

- Set-up matching rule on "Name" with Equals. Succes =3, Failure=4, Undefined=5

- Created Strategie ( Low=1, High=2) using the single rule

- When selecting one record and using the "Selected versus All", I get consistenly shows 2 matches for each of the "duped" records, while for my first record I get 3 matches.

- When selecting all records and using the "Selected versus All", I get 3 matches for each record.

..

Anyone else facing the same ?

Regards,

Dirk