Files in, deliverables out

The Pipeline

This is not a description of how the review would work. It is the review, running. Load the files the Embassy and Athora will actually send, watch them pass or fail the intake gate, set the actuarial assumptions, and generate the Survey Report from your own data. If you have no files to hand, the sample set below is the same shape as the real thing, including a version carrying the faults an insurer file usually arrives with.

1 · Load the files→2 · Map the columns →3 · Pass the gate→4 · Set assumptions →5 · Generate
Nothing here is uploaded. Every file you load is read inside your own browser and never leaves this machine. There is no server to send it to. Participant data is personal data under Regulation (EU) 2016/679, and for a tool meant to be handed to a client the safest architecture is the one with no transmission in it at all. You can confirm this by disconnecting from the network and running the pipeline anyway.
1

Load the files

No files loaded

Participant census

Required

Actives, deferred members and pensioners. Dates of birth, hire and plan entry, pensionable salary, part-time fraction, benefits in payment. From the Embassy's HR record.

Insurer reserve statement

Optional, needed for conformity

Year by year: opening reserve, contributions, interest credited, benefits, charges, closing reserve, and whether generations are split. From Athora Belgium.

Employer contribution record

Optional, needed for reconciliation

What the Embassy actually remitted, by year. This is what the insurer's receipts are reconciled against, member by member and year by year.

No files to hand? Use the sample set

Two complete sets, in the formats these files genuinely arrive in: a spreadsheet with a title block above the table, Dutch column headings from the insurer, amounts written 1.234,56 and dates written dd/mm/yyyy. The as received set carries the faults our certifying actuary warned us to expect, so you can watch the gate catch them.

Clean set, passes the gate Census .xlsx Insurer statement .xlsx Contribution record .xlsx Census .csv
As received, carries real faults Census .xlsx Insurer statement .xlsx Contribution record .xlsx
The same census in every other format we read, so you can test the claim rather than take it .xls (legacy Excel) .ods (OpenDocument) .csv (semicolon, European) .csv (comma) .tsv (tab) .psv (pipe) .json

2

Map the columns

Waiting for a file

Nobody is going to rename their columns for us. The pipeline finds the header row, reads English, Dutch and French headings, and reports how confident it is in every match so you can correct it. Anything it could not resolve on the evidence is stated rather than assumed.

3

The intake gate

Waiting for data
3b

Correct the data

Nothing to correct

Some of what the Embassy holds arrives as a scanned document, and a figure keyed from a scan has to be checked by something other than the person who keyed it. Edit any record here and the gate re-runs immediately, so a mistyped reserve breaks the reconciliation rather than passing quietly into the valuation. Nothing you type changes the file you loaded.

IDStatusSex BornHiredEntry SalaryPT ReserveBenefit Form
4

Set the assumptions

On the approved basis

The parameters that move the answer most are here. Everything opens on the basis this engagement recommends, and any departure is marked. The complete surface, all of it adjustable, is in the Assumption Lab.

5

Generate the deliverables

Waiting for a valuation

Survey Report

The deliverable named in the solicitation. All twelve Guideline elements, produced from your data on the assumptions set above, with every departure from the approved basis marked. Opens ready to save as PDF.

Data exception register

Every exception the gate raised, classified and itemised, naming the record and the fault. This is the file the Embassy hands to the insurer as a work list.

Member-level extract

Every member with the obligation and service cost attributable to them, so any total in the report can be traced to the records that produced it.

Projected benefit cash flows

Forty years of expected benefit payments, undiscounted and discounted, which is what a funding conversation actually needs.

Assumption record

Every parameter used against the approved basis, flagged where they differ. The audit trail for how this particular set of numbers was produced.

What this does and does not claim. The pipeline reads any tabular file we have met: the spreadsheet family including legacy Excel and OpenDocument, delimited text with the separator detected rather than assumed, and JSON. It will not read a scanned PDF or a photograph of a statement, because nothing can do that reliably enough to value a pension on, and we would rather say so than let you find out at award. Where a file defeats the automatic mapping, the mapping is editable by hand and the work is minutes rather than a renegotiation.