WBONO000701
WBON0000701
From: robert worden + GRO
To: Andrew Parsons
i “GRO
Ce: 'Chris Emery’
Subject: 1700 Peaks
Date: Wed, 27 Mar 2019 11:32:36 +0000
Importance: High
Jonathan Gribben
Andy, Jonny -
I'd like to give you an early heads up on a type of analysis I have been doing recently, which I think is very
promising.
It can give me another independent way to estimate the 0.4% upper limit on the impact of bugs on claimants,
that can be done quite quickly as follows:
¢ If there is a bug affecting a branch's accounts, the branch FAD code is quite likely to appear in a Peak
(as is confirmed by Callendar Sq, Receipts Payments mismatch, suspense account). Say the probability
is as low as 50%; it is probably higher, for any significant financial impact.
© Out of 218,000 Peaks, there are about 1,700 which mention the FAD code of one or more claimants, at
a date when the claimant was in post. (I have written a program to find them)
¢ To calculate the total impact of bugs on all claimants accounts, we just have to sum the likely financial
impact for each Peak that indicates a bug, over those 1,700 Peaks.
¢ To get a total impact of £18.7M, the average impact on a branch per Peak has to be £10,000 (or £5000,
if you allow for the 50% factor above). This is very high indeed, for an impact on one branch. PO
would have been going mad about it.
¢ For most of the 1700 Peaks, it is pretty obvious that they had no impact at all, and were not bugs -
they are about ISDN or something irrelevant.
¢ It would be quite simple to examine a random sample of 100 Peaks out of the 1700, and scale up by 17
¢ [have not done this, but I will bet that the resulting number is tiny - probably less than 1% of £18.7M
This would give three very independent upper limits for the proportion of claimants’ losses arising from bugs
- which will probably be 8%, 0.4% and say 1%.
These limits come respectively from the claims data, KELs, and Peaks. Three very different sources and
assumptions. It will be very hard for the Cs to get away from these three independent analyses - i.e. to prove
that they are all wrong.
Note that for future trials of lead claimants, it is very easy to pick out the small number of Peaks which
mention each claimant's FAD code or codes, and examine them one by one. Those Peaks will tell you all
about remote access as well as bugs - to put each claimant under the spotlight - even without ARQ data.
WBD_000571.000001
WBONO000701
WBON0000701
Peaks have been churning out at a rate of 50 per working day since 2000 - there is not much they miss.
Perhaps we can discuss tomorrow at chambers?
Robert
WBD_000571.000002