Data validation

The rules by which the book argues with a record

An open list: what is checked, why it is an error, and what to do about it. Rules are stated in advance — like an exam syllabus, not a remark after submission.

77

rules in the registry

36

find a contradiction in the data

10

report that data is absent

31

draw attention without holding up confirmation

39

carry a numeric limit stated aloud

Three states, and “fix it” is not among them

A check never asks you to fix a record, and that is not politeness. By construction it cannot tell a correction from a fudge: a rule saying “yield of 45,000 kg is out of bounds” would, at the word “fix it”, invite someone to put 12,000 — the block lifts, the mark of the Association goes on, and the book records a number nobody ever measured. So a check names the state of the record, and the decision stays with people.

There are three states. “Data missing” — the fields are empty: no date of birth, no breed, not a single parent. There is nothing to fake here, and the only way on is to upload what is absent. “Data does not add up” — the data is there and contradicts itself: a calving before the birth, DNA that rules out the recorded sire, a blood share above one hundred per cent. Here editing the record is precisely the thing that must not be done without grounds. “Worth a look” — the rule has noticed something odd and may be wrong.

The difference between the second and the third is not one of strictness but of whether the rule itself can be wrong. “Parent younger than offspring” cannot be wrong — it is a contradiction. “Yield above twenty-five thousand” can be: such a cow is rare but possible, and forbidding her would mean vouching for what we do not know.

What happens to a record that has questions against it

Nothing is deleted and nothing is rolled back: the data stays in the book, and the farm sees it and edits it. What is held is something else — the confirmation of the Association, the very mark the record is kept for. Until a finding has been dealt with, the animal is not confirmed and does not count as pedigree by the book.

How it is lifted differs. “Data missing” is lifted by uploading what is absent — and by nothing else. “Data does not add up” is lifted by an explanation the expert accepts: a document, a laboratory certificate, a letter from the farm saying where the discrepancy came from. The expert may judge a finding immaterial and waive it — but the waiver is by name, for this record and this rule, and it stays in the history of the application.

The third outcome is lawful too: to accept that the animal does not count as pedigree. The book is under no obligation to confirm everything uploaded into it, and a record without confirmation is not a defect but an honest statement of where things stand.

What the book asks for besides consistency

The rules below check the data against itself. Apart from them the book looks at completeness: confirmation means not only “no discrepancies” but also “everything required has been submitted”. The list is short and the same for every application.

Without these a record is not confirmed:

The list

Every rule has a number, and the number is what to quote in a letter or in a conversation with support: rule 45. It reflects the place of the rule in the registry and shifts if a new rule is inserted in the middle; the permanent key is the code of the rule, which goes into exports and never changes.

data missing: 10rules in the registry

Passport2 rules

Fields of the record itself: identifier, date of birth, breed, breed percentage.

1.1No date of birth

data missing

The record has no date of birth.

Without it there is no age, no age at first calving and no herdbook certificate.

1.3Breed not stated

data missing

No breed is selected on the record.

The breed determines the comparison base the index is computed against.

Parentage2 rules

Parents, the pedigree and everything computed from it. An error here costs more than the rest: it does not spoil one record, it distorts the evaluation of every descendant.

2.9Neither parent stated

data missing

There are neither links to parents nor their identifiers from the documents.

A record without parentage takes no part in pedigree evaluation and will not carry a herdbook certificate.

2.15Imported animal has a pedigree shorter than three rows

data missing

The animal came from another country, and the herdbook holds fewer than two parents, four grandparents or eight great-grandparents — counting only those entered as records.

Imported semen spreads across herds years before a certificate is ever issued for the bull, and by then it is his daughters whose pedigree is asked for — and it stops at him. Three rows of ancestors on import are required by both EHRC and WHFF rules: without them there is nothing to confirm the parentage with.

Limit: 2 parents, 4 grandparents, 8 great-grandparents

Reproduction1 rule

Calvings and inseminations — not one by one, but as a sequence in time.

3.20Natural-service bull without confirmed parentage

data missing

The mating is recorded with a bull who has no DNA test with the verdict “parentage confirmed”.

An AI bull is vetted by the stud; a natural-service bull is vetted by nobody, while his offspring enter the herdbook alongside the rest. Until the parentage is confirmed by a test, the pedigree of all his descendants rests on the farm’s word — and collapses all at once if that word turns out to be wrong.

Production3 rules

Milk yield, fat, protein and how well they agree with one another.

4.6It is not stated who computed the imported evaluation

data missing

The record carries an imported evaluation, and at least one of the two beside it is not filled in: whose evaluation centre it is, or which comparison base it was computed against.

An unnamed evaluation cannot be read: a TPI figure from an American catalogue and an index from a regional centre are different quantities, and without the name of the source there is nothing to compare them against. The same field is what makes a divergence from our own computation meaningful: without it, any divergence is explained away as “different bases” and therefore explained by nothing.

4.7It is not stated how the imported evaluation was computed

data missing

The record carries an imported evaluation, but the method of its computation is not filled in: from markers, from daughters, or from the pedigree.

A genomic evaluation is read off the animal itself; a pedigree evaluation is the average of its parents — and for a semen buyer these are claims of different value. In the book they sit as identical numbers, and this field is the only thing that tells them apart. An evaluation that arrives unlabelled stays unidentified forever: there is nowhere to recover the method from afterwards, and with it goes the ability to choose which of an animal’s several evaluations is the current one.

4.8Progeny evaluation without daughter and herd counts

data missing

The evaluation is marked as based on daughters, but the number of daughters or the number of herds they stand in is missing.

A daughter count without a herd count means nothing: two hundred daughters in a single herd is an evaluation of the herd, not of the bull, because feeding, technician and climate cannot be separated from him in such a sample. What separates them is the bull being used under different conditions, and conditions are herds. A progeny evaluation missing either number cannot be verified: it says nothing about how reliable it is, nor whether it can be compared with an evaluation from elsewhere.

Status and disposal2 rules

Whether the status of an animal agrees with what is recorded about it.

5.3Disposal without a reason

data missing

The status is not in the herd, and no disposal reason is stated.

The reason for disposal is part of breeding records: it is what shows why a herd loses animals.

5.4Disposal without a disposal date

data missing

The status is not in the herd, and no disposal date is entered.

Disposal reports count by the date, not by the status: without it the animal appears in none of them, and the herd disposal rate comes out lower than the real one.