# Admit a SmartFACT product's export

Register the source, admit the numbered parts in order and know what the result can show.

Source: https://help.smartfact.ai/manuals/lottrail/admit-a-smartfact-export

Review edition: 16 November 2026

Roles: supervisor

CheckStation, WorkSteps, KitProof and StoresFlow can each write their own lot and serial records as files LotTrail admits as they stand. The product sends nothing and LotTrail fetches nothing: a person downloads the files in the product and a Supervisor admits them here, one review each. They are declared files like any other, so LotTrail does not check them against the product.

## Register the source as a dataset

In the product, an Owner or supervisor opens the panel Evidence for LotTrail. Under To register this source in LotTrail it shows four values. Ask for them exactly as shown, or sit beside the person who has the panel open.

In LotTrail, open Declared datasets and use Create a declared dataset. Type each value into the field with the same name.

| On the product's panel               | In LotTrail                                                          |
| ------------------------------------ | -------------------------------------------------------------------- |
| Source instance ID                   | Source instance ID                                                   |
| Source generation                    | Source generation                                                    |
| Declared scope code                  | Declared scope code                                                  |
| Relationships this source can record | Relationships this source can record: tick exactly the ones it lists |

Tick no other relationship, and do not tick operation. None of these products records a material going into an operation, and a dataset that claims a relationship its source does not record makes later answers look more complete than they are.

Complete the data owner, authority and retention fields from your own organisation's agreement, and name the readers. A dataset's source instance, generation, scope code and relationships cannot be changed after it is created. If one was typed wrongly, the product's files will be refused: create the dataset again with the right values.

## Admit the parts in order

One export is usually several files, called parts. The product's panel lists them in order, each with its part number, the record numbers it holds and its checksum. The file names sort in the same order.

Admit them in that order, one at a time: the oldest export first, and part 1 of each export first. For each part, open Review evidence files, choose the dataset and the file, and choose Check and preview file. Compare File checksum with the checksum on the product's panel, read the warnings, tick the declaration and choose Accept this reviewed file. [Review and accept a declared evidence file](/manuals/lottrail/review-and-accept-files) describes the review itself.

To find where you are, open Declared datasets. Each dataset says which records it holds under Records held, for example "Source records 1 to 7,500, as they stood at 16:00 on 10 November 2026. Declared complete." The next part to admit is the one whose records start at the next number.

Only the last part of an export says when the product read its records. Until that part is admitted, Records held says that the newest file states no time. That is expected in the middle of an export.

A part marked Corrections on the product's panel holds only correction notices. It is admitted in its turn like any other part. Until the part after it is admitted, an investigation can say that a correction has withdrawn a record and its replacement is not yet available.

## When a part was skipped

LotTrail does not stop a part being admitted out of turn. It tells you first. If a file starts beyond the records the dataset holds, the review says so in words like these:

"This file starts at record 5,001. This dataset holds records 1 to 2,500. Records 2,501 to 5,000 are in neither. If this is one part of a numbered export, admit the earlier part first."

The safe answer is to leave this review, admit the earlier part, and come back. The review of the later part then shows no such warning.

If you go on, a second tick appears under the declaration: Admit this file with those records missing. Accept this reviewed file works only with both ticks. The file is then admitted, and every investigation of the dataset reports a gap in the recorded source sequence until the missing records arrive. Use this only when the earlier records are truly lost or will follow later.

## When an earlier part is refused

Once the dataset holds a later part, the earlier part is refused as it was first written, because it says the source's records end before the point the dataset has already reached. The review warns that the file moves the declared coverage backwards or narrows it.

The gap is not permanent. Ask the person with the product's panel to open If a part was skipped or lost and use Write this part for a dataset that already holds a later part, naming the latest part LotTrail has accepted. The product writes the same records in a file LotTrail can admit. Its name ends with the word after and the later part's numbers. Admit that file in the ordinary way. The dataset's record range does not move and the gap closes.

Do not edit a file, renumber it or change its coverage by hand to get it admitted. A changed file has another checksum and its records no longer match what the product wrote.

## A new source generation is a new dataset

The product starts a new history when its installation has been restored from a backup, when its mapping changes, or when an Owner or supervisor decides to. Its panel then shows the next Source generation, and with a new mapping a new Declared scope code.

Register that as a new dataset with the new values and admit its files from part 1. Files of the new generation are refused by the old dataset, and files of the old generation by the new one. The old dataset stays as it is, with its saved investigations.

## What an investigation can show

These sources record associations, not physical inputs and outputs. On the investigation page choose the dataset under Declared dataset, and under Relationship meaning choose Related records and context. Leave Direction at From input towards output, except where the table says backwards: there choose From output towards input.

| Start from                            | Dataset from | Shows                                                           |
| ------------------------------------- | ------------ | --------------------------------------------------------------- |
| A job or order number                 | CheckStation | Its units, then each unit's inspections and their outcomes      |
| A lot the source named for its units  | CheckStation | The units of that lot, then their inspections                   |
| An inspection, searching backwards    | CheckStation | The unit, then its lot and its job                              |
| A job number                          | WorkSteps    | Its units, then each unit's completed procedures                |
| A job number                          | KitProof     | Its kits                                                        |
| A component serial or a component lot | KitProof     | The kits it was checked for                                     |
| A kit, searching backwards            | KitProof     | The serials and lots checked for it, and its job                |
| A supplier's lot                      | StoresFlow   | The jobs it was issued to, with the quantity less returns       |
| A supplier's lot, searching backwards | StoresFlow   | The purchase documents it was received against, with quantities |

Each of these says what the product recorded and nothing more. A component checked for a kit was checked, not proved fitted. A lot issued to a job left stores for that job; the record does not say which unit it went into. A completed procedure is about a unit and is not a material input.

Recorded physical relationships gives no answer from these datasets. It says that the selected source does not record all the relationship types needed for this search. That is correct, and it is not a fault in the file.

No single search answers which finished units hold a given supplier's lot. StoresFlow says which jobs the lot was issued to. Take each job number to the CheckStation, WorkSteps or KitProof dataset and search from it there. The answer is every unit made for those jobs, which is a wider set than the units that hold the lot.

Where two products hold the same unit or the same lot, a Supervisor can join that one pair with a reviewed identity link. See [review an exact same-identity link](/manuals/lottrail/review-identity-links). A job number cannot be joined: carry it by hand from one search to the next.

## Limits to know

An investigation reads every event of the datasets you choose and takes up to 50,000 events across them. Beyond that it is refused, however the files were admitted. As a guide, 50,000 events is about 10,000 CheckStation units, 12,500 WorkSteps units, 5,000 KitProof kits or 10,000 StoresFlow lots, and fewer when two datasets are searched together. The product's panel shows how many events a history has exported.

If the product could not write a record, for example because a code is too long for the file format, every file of that history says the history is incomplete. The review warns that the source declares the file does not cover its complete history, Records held ends with Declared incomplete, and investigations say so. It stays so for that source generation. The product's panel lists the records that were left out.

A file is at most 10 MiB, a review lasts thirty minutes, and sixteen ordinary files and four correction files can wait for review at once. For a first export of many parts, stage a few, accept them in order, and stage the next few.
