The rules by which the book argues with a record

The date of birth is later than today.

Usually the day and the month have been swapped, or the year is wrong.

The breed percentage is below zero or above one hundred per cent.

A share of blood is a part of a whole and cannot go beyond one hundred per cent.

Walking up the pedigree leads back to this same animal.

Traversal of the pedigree loops at that point. The inbreeding computation survives it — there is a guard in place — but at the cost that the true coefficient cannot be computed either for this animal or for any of its descendants.

At the birth of the offspring the parent was younger than 19 months or older than 20 years. Animals produced by embryo transfer are left out of this rule entirely.

Younger than nineteen months is physically impossible: gestation lasts about 279 days. Older than twenty years is almost always a link to the wrong animal — to a namesake, for instance. With an embryo transfer the age of the donor has to be counted at the day the embryo was collected, not at the birth of the calf: the herdbook has no such field yet, so the rule stays silent and the report says on how many records.

One dam has two offspring less than 270 days apart, and it is not a twin birth. Offspring produced by embryo transfer are not counted.

A cow cannot carry two calves faster than gestation lasts. Either one of the offspring is recorded under the wrong dam, or it is a twin birth entered as two separate calvings. An embryo donor is under no such constraint: she carried none of her offspring — different recipients did — and the gap between their births can be anything.

The number of litters recorded for one dam exceeds what she could have produced between her earliest possible calving and the birth of her last recorded offspring. Twins count as one litter; offspring produced by embryo transfer are not counted at all.

This is what a merge looks like: two or three different animals with similar numbers collapsed into one record, with all their offspring recorded under a single cow. Per-record rules cannot see such a merge — each link is plausible on its own, and it is their sum that is impossible. Nor is the error a single one: the pedigree of every one of those offspring is wrong, and with it their blood share, inbreeding and evaluation.

The sire or dam field holds this same animal.

Traversal of the pedigree loops on such a record, and the inbreeding coefficient cannot be computed.

A female is recorded as the sire, or a male as the dam.

Most often the sire and dam fields were swapped during entry.

The parent was born after the offspring, or on the same day.

A direct contradiction: usually the link points to the wrong animal — to a namesake, for instance.

The parent identifier copied from the certificate differs from the identifier of the animal the link points to.

The pedigree is stored twice: as a link and as text from the document. While the two agree it is a safeguard; once they diverge, one of the two is wrong, and only the farm knows which.

The inbreeding coefficient exceeds the limit that has been set.

A value like this comes from close mating and calls for parentage confirmed by documents.

The test did not confirm the parentage, yet the parents are still on the record.

A direct contradiction between the document and the record — and the only case in which the system knows for certain that the pedigree is wrong.

The age at first calving is under nineteen months.

Gestation lasts about 279 days, so such a calving would have required conception before puberty. This is an error in the date; there is no other explanation.

Number 1 stands on several calvings of one cow.

The calving number runs through the whole life of a cow; there is only one first.

A calving with a higher number carries an earlier date than the one before it.

Every value on its own is plausible; only their order is impossible. Usually the numbering was mixed up when the data was carried over from the previous recording system.

The interval between neighbouring calvings is shorter than two hundred and seventy days.

A cow cannot calve twice faster than gestation lasts. Either the date is wrong, or the second entry records an abortion.

One animal has two identical events on one date: two calvings, two inseminations or two test-day recordings.

Usually the same file was uploaded twice. Until the duplicate is removed, every average for the animal is computed on doubled data and the calving numbers run ahead.

Fewer than twenty days between the calving and the insemination.

Before the twentieth day the uterus has not recovered, and there is physically nothing to inseminate. This is not bold management but an error in the date.

The date of the pregnancy check falls before the date of the insemination itself.

The order of events is reversed: there is nothing to check before the insemination.

The service sire recorded on the insemination was born after its date.

The link points to the wrong animal — as a rule to a namesake, or to a match on a short identifier.

An animal has several calvings with the same number — two third calvings, for instance.

The calving number runs through the life of a cow and cannot repeat. A repeat means either a doubled upload or a slipped link: the calving was recorded against the wrong animal. The second is the more dangerous — the event disappears from one cow and appears on another, and the number of calvings, from which the lactation number and the age group are derived, lies for both.

Fewer than 265 or more than 295 days lie between the fruitful service and the calving. With an embryo transfer the count starts at the day of transfer, and both bounds are a week lower.

Gestation in Holsteins lasts about 279 days and varies within a week. Shorter than 265 days is usually a slip in the calving date, or an abortion entered as a calving. Longer than 295 usually means the wrong service is marked as the fruitful one, or a later service never reached the herdbook.

The offspring of a calving is held by a different farm or herd than the dam was on the day of that calving.

A calf is born where its dam stands. A divergence means one of two things: either the calving is recorded under the wrong cow, or the calf was moved and the movement never reached the herdbook. Location is an independent witness: a date in a record is changed with one keystroke, while a move leaves a trace with the other party.

The mating is recorded with a bull who on that day was held by a different farm — or on a different site of the same farm — than the cow. Artificial inseminations are outside this rule.

Natural service requires the bull and the cow to stand together. A bull from another farm means either a link to the wrong animal — usually a namesake — or an unrecorded lease of the bull.

Lactation yield below five hundred or above twenty-five thousand kilograms.

Twenty-five thousand occurs in world record holders, forty thousand in no one. Limits drawn this wide catch an error in the unit of measurement without touching a good animal.

Fat content outside the range from two to six and a half per cent.

Beyond these limits it is usually not fat content but swapped columns.

Protein content outside the range from two to five per cent.

The same again: beyond the limits lie swapped columns, not a record.

The record of a bull carries milk yield, fat, protein or lactations, as a cow record would.

A bull has no production of his own: there is nothing to milk. The figure got there from someone else’s spreadsheet or from a wrongly entered sex, and it spoils any herd average the bull enters alongside the cows.

The breeding value for fat in kilograms differs by more than one standard deviation from what follows from the evaluations for yield and fat percentage.

Kilograms in an evaluation are not an independent figure: they follow from the yield and the percentage. A divergence means the figures were assembled from different sources — and what goes into the index is something the animal does not have. Each figure on its own is plausible, which is why the plausibility limits do not see this error.

The breeding value for protein in kilograms differs by more than one standard deviation from what follows from the evaluations for yield and protein percentage.

The same as with fat, and more dangerous: protein carries fourteen per cent of the weight in the Association profile, and its spread is three times narrower than that of fat. Forty extra kilograms of protein pull an animal into the top one per cent of the book.

A male animal carries a cow age group — first-lactation cow, cow in second lactation or cow in third and later.

One of the two fields is wrong, and from outside it is not visible which. If the sex is wrong, the animal counts as a cow in every report by group and can be given a calving. If the group is wrong, a bull falls into selections of cows and spoils their averages. The system does not choose on behalf of the person: all it knows by itself is that these two values cannot exist together.

An animal in the calf or heifer group has milk yield, fat, protein or lactations filled in.

There is no lactation before the first calving — there is nothing to milk. Either the age group is out of date and the animal calved long ago, or the production row arrived from another animal. The first spoils reports by group, the second the evaluation of a bull on his daughters.

The animal is older than the age limit and is still listed as alive.

Almost always it means that a disposal was never recorded, not that the cow lived to a record age.

A disposal reason is filled in while the status has stayed as in the herd.

The animal keeps appearing in herd lists and counts, having left the herd everywhere except on paper.

The animal has a test-day recording, an insemination, a calving or a weighing dated later than the day of disposal.

A cow that has left is neither milked nor calving in this herd. Usually it means one of two things: the disposal date is entered wrongly, or the event was recorded against the wrong animal. The second is the more dangerous — the event has disappeared from one cow and appeared on another, and both of them lie.

The herd holds records where the yield differs from the rest by two orders of magnitude: some in kilograms, some apparently in tonnes or centners.

One by one such records are caught by the plausibility check, and the farm sees fifty identical warnings instead of one cause. While the units differ, any herd average and any comparison between animals is meaningless.