Spreadsheet Screw-ups (or bad choice of Excel for serious purposes)

https://www.theregister.com/2021/06/02/uk_special_forces_data_breach_whatsapp/

got both of these thanks to the Modeler’s Miscellany:

Oldest spreadsheet error

(that we know about)

Plimpton 322

https://personal.math.ubc.ca/~cass/courses/m446-03/pl322/pl322.html#errors

Accounting for the errors

As confirmation of both the interpretation of the table and this conjecture regarding p and q , the four apparent errors can be reasonably explained:

  • The number [9, 1] in row 9 should be [8, 1] - a simple copying error.
  • [7,12,1] in row 13 is the square of [2,41] , which would be the correct value - a mistake particularly easy to make since the squares also appear in the conjectured calculation.
  • The correct value to replace [53] in row 15 is [1,46] , which is twice the erroneous value.
  • As for the fourth error in row 2, where [3,12,1] occurs instead of [1,20,25] , there have been a couple of solutions proposed. None are entirely convincing. The possibility proposed by Gillings suggests strongly that those who made up the table had values of p and q at hand.

More on Plimpton 322:

Also, the spreadsheet was set to manual calculation and they forgot to F9.

1 Like

I am not quite sure what’s going on here.

#EndSARS panel blames computer for errors in report

The Lagos judicial panel on police brutality says the state government is trying to evade responsibility by citing computer errors in the #EndSARS report.

According to the panel’s report, protesters were killed at the Lekki tollgate on October 20, 2020 in what could be described in context as a “massacre”.

In a white paper on the report, the Lagos government accepted 11 out of the 32 recommendations of the panel, rejected one and accepted six with modifications.

The Lagos government also said the finding of the panel that nine persons died is “irreconcilable” with the testimony of John Obafunwa, a pathologist.

PANEL REACTS

Responding to the Lagos government’s white paper in a statement released by Ebun-Olu Adegboruwa, the panel said the state could have reached out to members for clarifications where necessary.

“The chairperson, all panel members and indeed the secretariat of the Panel were all within the reach of the LASG for clarifications if there was sincerity, other than picking holes in order to evade responsibility on account of computer errors and tabular alignments of cut and paste ,” the statement reads.

The panel also said the duplication of names on its list of casualties was as a result of a computer error, adding that they considered several dictionary definitions of the word ‘massacre’ before it was used.

This was a misnomer from the spreadsheet that ought to have terminated at Page 297 but mistakenly overlapped to Page 298 with the same names and same numbers ,” the panel said.

“It was the computer error of the secretariat of the panel which could have been corrected as the secretariat of the panel was domiciled in the ministry of justice at all times.

“In any event, the mere fact of repetition of same names on a table cannot without more, nullify the uncontroverted evidence of death.

“The panel considered several definitions of the word MASSACRE and adopted one of the dictionary meanings of MASSACRE as being ‘the act or instance of killing a number of usually helpless or unresisting human beings under circumstances of atrocity or cruelty’.

“The Panel considered that firing live bullets at unarmed, peaceful and unresisting protesters which led to the death of some of them, was cruel and atrocious on the part of the military and the police. The White Paper ignored these explanations and findings by the Panel.”

£46.55 million fine – and spreadsheet error was a part of it.

The notice is here:

Error 4

2.14 On 28 November 2018, the Treasury Markets team identified that a cell in the spreadsheet used for preparation of the FSA047/048 return included a positive value where a zero or negative value was expected. This was investigated by SCB, which concluded that this was caused by an error in SCB’s ILAS reporting system, which had erroneously included in the USD Gap 2 Metric client collateral held at the London Clearing House (“LCH”) but not the return of that collateral to clients (“Error 4”). Error 4 resulted in an overreporting of the USD Gap 2 Metric by USD 7.9 billion and a breach of the PRA’s Expectations (meaning this error caused SCB to fall below the USD Gap 2 Metric survival period of 91 days). SCB notified the PRA of this error on 1 April 2019, over four months after it was identified. The error was escalated to the FORC and to senior management shortly after it was notified to the PRA and was logged as a “Significant” ORE.

4.16 On 29 November 2018, Treasury Markets identified that a cell in the LMM spreadsheet (“Line 49”) was showing a positive number (around USD 10 billion), even though it related to liabilities and should therefore have been either zero or negative. Treasury Markets called GLRR the same day to discuss the finding. Following the call, GLRR emailed GFS to request a breakdown of the figure that Treasury Markets had identified as unexpectedly positive.

An £8 billion error, a £47 million fine. A spreadsheet without proper controls.

To be sure, that wasn’t the only error or lack of controls (after all, this is only error 4). But jeez.

News coverage of the fine & errors

more:

“Staff in my department have confirmed that a few values for revenues for the current fiscal year, 2021-22, on page 15, were misallocated and presented on the wrong lines,” Compton told the legislature.

“This was due to an administrative issue with a spreadsheet that was not linked to the proper columns when they were re-sorted before printing.”

She said the overall totals were correct, but the numbers in some of the columns had been misplaced. Those number corresponded to provincial tax revenues for things like the province’s gas tax, liquor tax and carbon levy.

A government staffer later told CBC News the problem was based on the province’s use of Excel to compile its budget.

Dozens of people included in a massive privacy breach at the Teaching Council/Matatū Aotearoa have now learned their names were made public – four months after the council claimed all those involved had been contacted.

The breach, first revealed by TVNZ in December 2021, saw information about complaints against teachers, principals and schools uploaded by the council to the internet. The data had been indexed by Google, which allowed it to be searched by the public.

Sixty-eight people were emailed in late March and told their names had been included in the breach.

Enquiries by Stuff confirmed their names were found in “hidden rows” of a spreadsheet. The Teaching Council said no-one on its incident response team had found the hidden rows in the four months since the investigation into the breach began.

If you’re going to do a fraud, don’t use Excel. It’s not secure.

No, not even if you use a password.

oh jeez

holy shit

This whole thing

the whole thread

… I don’t know what to do with this

1 Like

A top elections official said Monday that “human error” in tracking the results of the May 17 primary election led Pennsylvania to inadvertently certify a county’s vote counts that the state deems to be inaccurate.

The embarrassing revelation came in a filing before Pennsylvania’s Commonwealth Court, where the state is seeking to break a standoff with three counties — Berks, Fayette, and Lancaster — that refuse to include undated mail ballots in their official totals in defiance of guidance from the Department of State. The department, which oversees elections, sued the counties last month, asking a judge to order them to do so.

But in a court filing Monday, Jonathan Marks, the deputy elections secretary, acknowledged that a fourth county, Butler, had also refused to count those ballots — and that the county had notified the department three weeks before the lawsuit was filed.

Marks apologized to the court for what he described as an oversight resulting from “a manual process” — a spreadsheet — the department had used to track which counties were counting undated ballots. Butler County was misclassified in the spreadsheet, he said, and from that point forward was left out of the state’s campaign to push counties that hadn’t included them.

-sigh-

when will they ever learn…

[hmmm, time to write a “Where have all the flowers gone” parody…]

1 Like

I’ll put this here. I can’t remember if I told this story in the AO, but I know I used to tell my UConn students about it.

1 Like

"A revised version of the Liberal financial framework shows the party
underestimated the net provincial debt, which would come in at $256.8
billion at the end of their mandate instead of the $240.6 billion
initially projected on Sept. 4.

Leitao, who is not seeking re-election but is presiding over the Liberal
campaign, told reporters in Quebec City that he wasn’t embarrassed and
said the calculation of debt is complex.

Late Tuesday, the Parti Québécois had to correct its platform hours
after presenting it, blaming a spreadsheet error for overstating the
> cost of its commitments and of the projected deficit in the final year
> of a PQ mandate.

The financial framework has been verified by an economics professor at
the University of Ottawa, and PQ Leader Paul St-Pierre Plamondon boasted
that his was the only party to present a framework checked by an
independent expert."

https://www.recorder.com/Charlemont-Selectboard-takes-School-Committee-to-task-following-budget-change-48033640

TMS, the district’s former business management service that it
shared with the Mohawk Trail Regional School District, counted the
transportation line twice when calculating the overall FY23 budget of
roughly $2.64 million. This error led to a budget with $118,398 more
than it should have included.

[Girard] said he was unhappy to read about “a $118,000 oops in the
newspaper.”

Bertsch said “there were a series of spreadsheet errors.”

Sort of a screw up, but not for the sloppy spreadsheet creator. Who knew sloppy spreadsheets could raise $2B?

1 Like

So, on one end, a person can raise $2B with a crappy spreadsheet.

On the other end, a person gets overpaid 10K pounds by accident, and it sounds like this council in South London is just sloppy in general

In December 2020, the council overpaid Miss X £10,000 due to a human error in compiling the spreadsheet for calculating payments. The council told her she would need to repay the cash and agreed to carry the debt over to the next financial year.

The report reads: “When I spoke to Miss X, she explained she had borrowed from family and friends when payments were made late, and she incurred daily interest charges on her business overdraft which was eventually revoked due to late payments. Miss X also says she incurred recruitment costs when staff resigned because of late payments.”

The report also says: “This is fault and caused Miss X an injustice as she had to chase the council on each occasion to establish exactly what was owed before it was paid. Being without large sums of money would have caused further uncertainty for Miss X.”


The council has now apologised to Miss X and paid her £150. A spokesperson for Wandsworth Council said: "This case involved an overpayment around nursery early years funding which was followed up with some late and inaccurate payments while we were trying to recover the initial overpayment.