GIPR Technologies · A worked example

See what a run produces, before you commission one.

Synthetic data. Real engine. Every figure checkable.

Most vendors ask you to take a savings number on faith. This page is the alternative: a complete run on invented data, showing the shape of the findings you would get, how much of your estate a second source makes visible, and the arithmetic behind our headline — so you can check it rather than trust it.

01 — What a finding looks like

One duplicate, and why nobody caught it.

This is the artifact you would forward to internal audit: both halves of a duplicate payment, side by side, with the reason each department’s own review missed it. Invented data — but the format, and the reasoning, are what a real run returns.

The crossover, as a record · synthetic data Found line · stage 2
One finding, both halves, and the reason no single review could see it
FieldAccounts payableCard program
SUPPLIERBluestone Facilities ServicesBluestone Facilities Services
AMOUNT$4,850.00$4,850.00
DATE21 April 202623 April 2026
UNITFacilities · North siteIT · South site
CHANNELAP voucherPurchasing card · no purchase order
REFERENCEVoucher, FacilitiesPC-09001 · cardholder CARD-0007

One bill, paid twice. The invoice was vouchered by Facilities at the North site; the card charge landed in IT at the South site, two days later. Each unit’s own review saw one half and nothing anomalous in it — unit-level review structurally cannot catch this, which is why the finding attributes at the organization level rather than to either unit. The cardholder is carried as a pseudonymous reference, never a name: the engine has no field in which to put one.

02 — What a second source is worth

The same estate. One more feed.

This is the number worth taking away, because it is the one that applies to you whatever your data looks like: how much is already happening that no single system can show you. Nothing changed here except that the card feed arrived alongside the accounts-payable extract.

The two numbers worth taking away

Both come from adding a single card feed to an accounts-payable extract. Nothing else changed.

Spend that bypassed a purchase order

$86,880.00 $134,089.66

54% more of it visible. It was always happening. No single system could see it.

Found — duplicate payments

$42,650.00 $47,500.00

The difference is one $4,850 transaction — the same supplier, the same amount, invoiced once and carded once, two days apart, under two different units. Each unit’s review saw exactly half.

03 — The full run

Every line, at every stage.

Three runs over the same period, adding one source each time. Read across a row to see what each feed unlocked; read the bottom rows to see what the engine refused to answer because the data was not there.

Staged run — synthetic data, deterministic
  Stage 1
AP extract only
Stage 2
+ card feed, vendor master
Stage 3
+ contracts
Found dollars — the only figure we sum $42,650.00 $47,500.00 $47,500.00
Addressable band $63,539.63–$64,079.26 $63,539.63–$64,079.26 $75,059.63–$92,879.26
Findings — rules that fired 7710
Spend that bypassed a purchase order $86,880.00 $134,089.66 $134,089.66
Spend split across duplicate supplier records $136,345.57 $144,334.78 $144,334.78
Purchases split just under an approval threshold $58,980.00 $58,980.00 $58,980.00
Spend on contracts that had already expired no contract data no contract data $96,450.00
Spend outside contracts already negotiated no contract data no contract data $73,900.00
Purchase orders with no contract backing no contract data no contract data $221,500.00
Purchase orders with no approval on record could not check could not check could not check

04 — The arithmetic

Why our headline is smaller, and how to check it.

Add every dollar in the final column the way a savings headline is usually assembled, then compare it to the one line we are willing to sum. The gap is not a rounding difference — it is the difference between money you can act on and money that is merely somewhere worth looking.

Derivation · stage 3 Every figure from the table above
What gets added, and what does not
LineAmountKind
Found — duplicate payments, traced to the records on both sides$47,500.00FOUND
Purchase orders with no contract backing$221,500.00exposure
Spend split across duplicate supplier records$144,334.78exposure
Spend that bypassed a purchase order$134,089.66exposure
Spend on contracts that had already expired$96,450.00exposure
Spend outside contracts already negotiated$73,900.00exposure
Purchases split just under an approval threshold$58,980.00exposure
Addressable opportunity, high end of the band$92,879.26modeled
Everything added
Published
$869,633.70
$47,500.00

$869,633.70 ÷ $47,500.00 = 18.31. The six exposure lines total $729,254.44; the modeled band high adds $92,879.26. Not one of those dollars is wrong — each is a real characteristic of the dataset, and each is reported. What would be wrong is adding them together and calling the result savings.

What this does and does not tell you. It shows the engine running end to end on four differently-shaped sources, keeping found, exposure and modeled dollars apart, and declining to answer a question it had no data for. It tells you nothing about how much is in your estate — these dollars are invented, and the defects were planted by us. The interesting run is the one on your data, where we do not know the answer in advance.

Next — Your data

Now run it on an estate we have never seen.

An accounts-payable extract and a card feed is enough to make the crossover above findable in your own numbers. Purchase orders, contracts and an approval log unlock the rest of the pack — eleven of the nineteen checks read purchase orders.

Scope an assessment