Category Archives: UK Trainwrecks

I shot the Sheriff (but I didn’t update the warrants database)

Leave a reply

The headline doesn’t scan as well as the original Bob Marley lyric, but that jarring dischord is nothing compared to the problems with the PNC (Police National Computer) in the UK.

A review is underway of the processes used to update the PNC database across the UK covering the “national process and practice for withdrawing warrants, involving courts, the police, and the crown prosecution service”. The review is also extending into Magistrate Courts (lower level ‘district’ courts in England and Wales) due to “differing practices” which may require procedures to be clarified.

At the heart of the issue is the withdrawal of warrants for thousands of defendants who never turned up for their court dates and have escaped justice because no warrants were issued for their arrest, no police pursued them and their ‘failure to appear’ wasn’t logged on the PNC. The power to withdraw a warrant rests (or should rest) with judges only…

Regular visitors to this site will recall the IQ Trainwreck that emerged about the DNA database in the UK a while ago… oh dear.

ahem… Information Quality problem with the trains…

2 Replies

http://www.theregister.co.uk/2007/11/13/eurostar_glitch/

Again culled from el Reg, here’s an example of an information quality problem that prevents a process (a very trivial process) from being completed.

It seems that the on-line booking system for the EuroStar train linking the UK with Europe doesn’t recognise the existence of the 29th of February 2008. As the leap year day is the day when women can traditionally propose to their partners (as opposed to dropping less and less subtle hints about the need for a big diamond on their finger which happens the other 3 years out of 4) this might pose a problem for some.

Why is this an IQ Trainwreck?…

  • Well, it involves trains so it is just too easy an association to make.
  • A process has been (apologies for this one) derailed by poor quality information (or master data or a poorly designed/implemented data quality check)
  • The comments from the readers of the post sum up the likely responses of people encountering the problem..
    • This is a trivial bug, if that’s how careless they are do I really want to travel by train?
    • They obviously don’t want my business, I’ll fly instead
    • Good grief…. how embarassing for them.

Apparently the Eurostar site isn’t the only one to have this issue… but it is the one that makes the best IQ Trainwreck.

oops Amazon did it again…

4 Replies

Also from “The Register”, it appears that Amazon have again created ‘interesting’ relationships in their data that have had uncomfortable results for their customers.

We previously reported on the IQ trainwreck that occured when Amazon sent email recommendations for sex toys to people who’d never bought such items from Amazon. Today they seem to have gone one better with a search for “Spiderman Watch” on Amazon.co.uk returning a quite prominent sex toy.

The Register posted their story at 15:03 GMT+1 today, and at 16:14 this correspondent found the same item, this time TOP of the search results.

I’ve uploaded a screen grab of my search results and clicking here will bring you to my search string… rather than risk offending readers the screen grabs are linked to rather than displayed as thumbnails.

Spiderman Watch Search Result (Image not presented for fear of offending)

As Amazon uses linkages within its data to present recommendations, purely in the interests of research I followed a few links on the product page for the unexpected result to see what might have lead to the association being created.

Under the product detail for the sex toy Amazon proudly lists that customers who bought items like this also bought “Spiderman – The Animated Series” and a number of innocent children’s toys.

The sex toy and a number of others of similar kind (which revealed themselves when I clicked on the producer name) appear to be categorised “Toys and Games” in the Amazon database… which means that they may (indeed WILL) appear in other searches. For example, if you search for “Rabbit” under the “Toys and Games” category, look what appears in the “New Arrivals” section on the left hand side…

Link to Screenshot of Search results for ‘Rabbit’ in ‘Toys and Games on Amazon’.

Further investigation indicates that the root cause here is the nature of the classification and tagging of these particular ‘toys’ in the Amazon database… searching for “Rabbit” under “Toys & Games/Dolls & Accessories” produces an interesting result on the first page… Example of classification/metadata issue

Why is this an IQ Trainwreck?

  1. Reputational Damage – The fact that the item is returned in a search for a children’s product is damaging to Amazon’s reputation as a retailer. As the story has appeared on The Register, it is possible that a ‘slow news day’ will result in it appearing in local nor national press in the UK (and it has been mentioned here).
  2. Derived from information – Amazon search results are returned from Amazon’s database… somewhere in the database a relationship has been created between the term ‘spiderman watch’ and this particular sex toy. This may have happened by accident or through malicious intent on the part of an individual. However the fact that it can happen suggests a lack of control over the information (should it be ‘consistent’ for a search for a children’s watch to return a sex toy? What controls might Amazon consider to improve the quality of their searches and prevent possibly inappropriate content from being shown to children?)
  3. Information is of poor quality – it fails to meet or exceed expectations.

Personally, I wouldn’t want to have to explain to any kids what those particular toys were for.

I am reminded of a story I heard about a particular court case in Ireland a few years ago where a children’s party hire shop sued a classified directory enquiries provider for listing them in the Adult party hire section of the directory….

It’s not only Amazon who have pricing trainwrecks…

Leave a reply

Courtesy of the correspondents over at TheRegister.co.uk, we have the story of Woolworths.co.uk who advertised a childrens book for a somewhat astronomical price.. stg£99 million (and 99p) and then charged £2.74 delivery as well…

http://www.theregister.co.uk/2007/01/16/rather_expensive_book/

This is the counter-weight to Amazon’s underpriced TV issues mentioned here previously.

The error is now fixed on the Woolworths site…

Store offers TVs at £0.49

1 Reply

The BBC reported in September 2005 that a well known UK retailer was caught out when a technical error led to customers attempting to buy televisions from its website for £0.49 (less than a US$1.00 then).

Why is this an IQ Trainwreck?

Why is this a trainwreck?—The store was able to repudiate the transactions because of the huge discrepancy between the offer price and the actual price. It was too good to be true. However, the reputational loss was significant. Argos was the butt of jokes for weeks and customers’ trust in their pricing was badly affected. It is easy to conceive that a simple piece of validation could have prevented the problem. “If the offer_price is less than 50% of actual_price then query value”

Cumbrian Train Crash—Poor IQ Management implicated

1 Reply

According to a report in the Guardian of 2007-08-27  the Cumbrian rail crash in Febuary 2007 that killed one person and injured many others had an Information Quality component which was the final link in the causal chain:

Alongside concerns over the points, the study’s focus will be a breakdown in communications among Network Rail’s Cumbria workforce which contributed directly to the crash. It is expected to state that track inspections were not carried out as planned, that records of inspections were flawed and that safety certification used by some engineers had expired.

Industry sources also confirmed reports yesterday that two different inspection teams thought the other had inspected the points prior to the crash and therefore failed to inspect a crucial stretch of track at Grayrigg. As a result, a Virgin Pendolino train travelling from London to Glasgow on the night of February 23 was derailed by a broken set of points that should have been noticed earlier by track inspection teams.

There are at least three information management issues here:

  1. The maintenance of inspection logs
  2. the maintenance of staff certification information
  3. training of staff in the importance of information and its use

The cost of the rescue effort was huge as was the cost to the NHS, the managers and maitnenance workers stand lose their bonuses and one person has been arrested. so once again we can see that Poor Information Quality can:

  1. kill
  2. cost huge sums of public money
  3. send people to prison
  4. hit individuals in the pocket

Hardware horrors…

Leave a reply

The Irish Examiner newspaper carries this story in the Business section of its website today. A UK hardware chain is having to rebate customers following an ‘administrative error’ and what the Examiner describes as a ‘system failure’ over the weekend.

Customers who paid by credit or debit cards found themselves being charged multiple times for the transactions resulting in some of them being put overdrawn on their bank accounts. The chain in question is refunding the customers their money, but given the impact this could have had on customer with direct debit payments for utility bills or mortgages the impact to customers is potentially significant. Not quite a full-on trainwreck but one might say that some of the carriages were left behind at the last station.

This would appear to be case of duplicated data (or duplicated submission of data) indicating a weakness in the processes and controls for managing information, which has resulted in an information experience for customers that failed to meet their expectations.

 While customers are being rebated, spare a thought for the additional administrative costs incurred by the hardware store in tracking down the affected customers, calculating the amount of overcharge, contacting the customer, processing the refund, arguing with the customer about the impact on their credit rating because the direct debits for their phone bill, credit card and mortgage all failed because there was no money in the account (note – there is no information to suggest that this has actually happened, but it is a risk and would be a long phone call). These are the costs of scrap and rework in the information process. Payment by electronic card (credit card or debit card) is a ‘information age’ payment process and is fraught with the risks of non-quality of information fromf poorly managed processes.

Buying tools by debit card… so good they billed you twice, no.. three times… no wait that could be four times….

CSI – won’t get fooled again?

1 Reply

According to The Register the UK’s police forces have a bit of an IQ trainwreck with their police DNA database (like the one that they use on cop shows like CSI to absolutely identify the criminal with flashing lights and snazzy computer graphics -only slightly slower and less ‘rock and roll’ in the user interface department).

Apparently the interface between the DNA system and the Police National Computer was rejecting a large number (100,000 identified so far) of records due to a mismatch of key information.

The Register describes the root of the problem as follows:

Police DNA samples are identified and linked to both the DNA database and the PNC with two unique numbers. One of these is automatically printed on a label, along with a barcode, in a police officer’s DNA sampling kit. The other is the Arrest Summons Number, which a police officer writes by hand onto the sample label.

The samples are sent to the forensic service firms for processing, where both numbers are inputted into a Laboratory Information Management System (LIMS). This information is then emailed to the DNAD Unit, where another computer system automatically extracts the information and sticks it in a skeleton DNA record, which would have been created when a police officer took the original sample.

If the lab technician misread the police officer’s hand writing, there would be mismatch because both numbers need to match a corresponding record in the PNC in order for the DNA record to be loaded successfully.

This resulted in between 10% and 12% of records failing to load due to a mismatch in the data due to errors in transcription of handwritten information. To compound the problem it would seem that politics reared its ugly head as the Reg tells us that “poor co-operation” between the police forces, the DNA database unit and the forensic labs meant that the problem wasn’t tackled until a Data Quality and Integrity Team arrived on the scene to sort it out.

Of course, they sorted out for the police and forensic labs without them having to get their hands dirty. We hope that a forensic review of the root causes was put in place or else this problem will recur.

The key issues here from an IQ point of view are:

  1. Proofing processes against error – the hand-written serial number was a source of error in a process that has to be water tight. What could be done to address this?
  2. Failure of governance (or of a recognition that goverance was needed) to ensure good quality information for law enforcement purposes.
  3. Politics and what seems like ‘not our problem, its your data’ type cultures.
  4. Reliance on scrap and rework by an external agency (who appear now to be seen as the ‘saviours’ of the DNA database)

Poor quality information arises in a number of different ways. Often it is as inconvenient for you as the wrong price on a packet of peas in your supermarket. Other times it is the difference between a crime being solved or a criminal getting away.