## Trussell Trust

Paddy and Carol Henderson were activists employed by the UN and working with street children in Sofia. Carol’s mother, Betty Trussell, had left them some money and in 1997 they established the Trussell Trust, a registered charity (no. 1110522). In 2000 they decided to switch operations to the UK on hearing about a mother having difficulty getting enough money to feed her children. In 2005 they set up the first food bank in Salisbury.

After a couple of years they handed over the running to Chris Mould, the former CEO of Salisbury District Hospital. He proved successful at fundraising and organising:

They now have more than a dozen employees and around 30,000 volunteers (according to the Independent 12 Dec 13). In a recent report (apparently a kind of extended web press release, organized as short headlines and charts rather than a more conventional text-based report, at least I cannot find a text report after 10 minutes of looking) Trussell claims to run over 400 food banks and to have given 3 days emergency food to nearly a million people in the UK during the year to 31 Mar 14, roughly half because of “benefit delays” or “benefit changes” and another 20% because of “low income”.

The obvious question is how real the need is. Are the clientele people who have no (legal) way of feeding themselves and their families without these foodbanks, or are they people who are incompetent at managing their (limited) resources, or are they people who simply prefer to conserve their scarce cash if someone is offering free food?

Unsurprisingly, there is plenty of anecdotal evidence for all three explanations and more besides, so the resulting debate tends to generate more heat than light. It is the kind of issue that somehow brings out the worst in the media, because it so easy to get emotionally-charged interviews that support your own pre-conceptions – or those of your target audience.

Of course, this issue has occurred to Trussell, so we have on the “How a foodbank works” page of the website:

The Department of Work and Pensions is not convinced, so we get the chairman, Chris Mould

… responding to accusations that his charity was “aggressively marketing”, Mould said: “You can’t get free food from the Trussell Trust by walking through the door and asking for it; you must have a voucher. More than 24,000 professionals – half of whom work in the public sector and health service, the police, and in social services – ask us to give this food to clients because they’ve made the decision that this individual or family is in dire straits and needs help. We’re not drumming up demand.”

The DWP last week claimed that food poverty has gone down under this government, pointing to a recent report by the Organisation for Economic Co-operation and Development. It found that the proportion of people in the UK who said they were finding it difficult to afford food had fallen from 9.8% in 2007 to 8.1% in 2012.

Clearly many GPs and other professionals do not relish a role as gatekeeper to benefits. There is ample anecdotal evidence that the standards they apply, for example, in providing sick-notes vary widely. Equally, is a volunteer food bank manager with Trussell, who sees someone without a voucher but in clear distress, really going to send them away without food? Does anyone really believe that Trussell’s clearly PR-savvy management would not back them up in that approach? The potential damage to the organisation from reports that they were turning away the genuinely hungry because some stupid box had not been ticked would be huge.

The careful reader will also have noticed that I gave three possible types of client, not just the genuinely needy and the scrounger, but also those who find it difficult to prioritize their limited funds. That has been a contentious issue all my lifetime. The old, paternalistic approach was to accept that this “incompetent with money” category is substantial and that it is better to give benefits in kind – the state pays the rent direct, gives food vouchers etc. The other approach was a mixture: partly that such an attitude is dehumanising and insulting and people must be trusted to make their own decisions about their own lives, sometimes coupled with a flat denial that the category exists, partly that it is important to help people to get better at such decision-making. There is no easy answer.

Unfortunately, even debating the issue now seems to be politically incorrect. Food banks came up at the last “Any Questions” (over the Easter weekend). None of the panel dared to address it. They all preferred to accept that all food bank clients are genuinely needy.

To be clear, I am not criticising the Trussell’s work. Indeed I applaud it. There is obviously a genuine problem and the organization is doing a good job in dealing with it.

Last night Greater Manchester Police confirmed that Mr Crompton had been arrested for administering a noxious substance.
A spokesman said: ‘On Tuesday, 2 Apr 14, police in North Manchester received a report of concern over medicine administered to a resident at a Blackley care home. Officers from the public protection investigation unit started an investigation and a multi-agency strategy meeting was held with partner agencies to discuss how to best establish the facts. Subsequently an 83-year-old man was arrested on suspicion of administering a noxious substance.’ Mr Crompton has been bailed until Apr 28.

[The Mail, Fri 11 Apr 14]

Well, the boys in blue got one thing right. Prescription medicines frequently are noxious substances which should not be administered. But a “multi-agency strategy meeting” to figure out if a man should be allowed to hand his wife her pills? Has all common sense disappeared in the welter of tick-boxes and arrest targets?

## Codejam

Codejam is here again. It is a test in devising algorithms against the clock, open to all-comers world wide. It was supposedly started by Google as a way of trying to select good programmers to work for them, or at least of providing a more objective measure of candidates’ abilities than the usual interview process. For the first few years they subcontracted the process, but now they handle it in-house. The main contest is once a year. There are four rounds:

Qualification round
Round 1
Round 2
Round 3
Finals

The finals are held in a Google office, typically in a different city each time. Google pays travel expenses for the top 25 from Round 3 to attend. The other rounds are online.

The qualification round is easy. It typically lasts about 24 hours to do problems for which one would normally be allowed less than 45 mins each. This time full marks was 90 and the pass mark (for moving to Round 1) was 25. 20,595 advanced to Round 1 and another 4867 scored something (plenty more entered but either did not compete or failed to get anything right).

The top 1737 got 90
the next 2 got 84
the next 85 got 79
the next 34 got 74
the next 6 got 71
the next 572 got 66
the next 27 got 65
the next 2 got 63
the next 457 got 60
the next 6682 got 55 (including me)
the next 6 got 54
the next 33 got 50
the next 80 got 49
the next 5 got 47
the next 878 got 44
the next 2 got 43
the next 16 got 41
the next 605 got 39
the next 7 got 38
the next 681 got 36
the next 21 got 35
the next 3 got 33
the next 2 got 31
the next 73 got 30
the next 195 got 28
the next 8384 got 25
the next 3 got 22
the next 18 got 20
the next 64 got 19
the next 2 got 17
the next 1315 got 14
the next 13 got 11
the next 42 got 8
the next 3410 got 6

There are usually 3 or 4 problems. Each has a page or so of description, always precise, but often not particularly easy to follow. In most cases you are then given two datasets, a small and a large. The idea is that you write your program (in more or less any computer language), then download the small dataset, run your program on it and submit the output to be marked (the downloading, running and submission must be completed in 4 mins). You then get either nil or full marks. If you get nil, you can keep trying (each time with a different dataset).

You then do the same with the large dataset, except that this time (A) you only get one shot, (B) you get longer to submit (9 mins? I cannot remember), and (C) it is not marked until after the contest is over. For each submission you have to submit your code as well as your output. In principle, they can check your code as well as your output, although in practice that only seems to happen in the event of suspected cheating or maybe in the final.

Any correct algorithm usually works for the small dataset, but for the large dataset, the algorithm usually has to be reasonably efficient. Sometimes code which works fine for the small dataset is too slow for the large one.

Google expected the last question to be the hardest, but actually qu 3 was the hardest. I wasted 3 hours failing to solve it. Maybe I would have solved it if I had stuck at it for another hour or so (and I had another 6 hours before the deadline), but I had to go off somewhere else. The bunching on 55 suggests I was not alone.

From Round 1 onwards, there is time pressure. But Round 1 is held at three different times of day to suit those in different time zones. Moreover, Round 1A, 1B and 1C are held on different days, so anyone prepared to put up with unsocial hours effectively has three shots at getting into Round 2. They happen at weekly intervals, starting next Saturday. Last year I got through Round 1, but with depressingly low marks and did not even attempt Round 2. So this year my goal is to reach Round 3. We will see. I certainly got off to a bad start.

## WebVan Mark 2

Webvan is a strong contender for the biggest dot-com bust ever. In the increasing frenzy of the late 1990s Louis Borders persuaded the boys at Benchmark Capital that he had an even bigger hit (at that point Borders bookstores, which he had founded in 1971, were doing well). It was so exciting that due diligence would have been perverse. They went for it big time.

It had one of the world’s silliest ever dot-com ideas: delivering fresh food to customers within a 30 minute window following an online order. Don’t be too distressed if you cannot figure out why that is so dumb. The top venture capitalists couldn’t either. I have now mislaid the book (about Benchmark) where it is all spelt out in gory detail, but from memory they threw about $200M at it. They quickly got Goldmans behind the float and raised a whole lot more. To what extent Benchmark et al cashed out at that point I cannot now remember. They did not need to hire anyone with a supermarket or groceries background, after all a cabbage is pretty much like a book, isn’t it? Of course, the business needed warehouses. A lot of warehouses. It placed a$1 billion order for fancy warehouses with a top engineering group and simply ran out of money. The remnants, an online grocery delivery service in Seattle and LA was acquired by Amazon to become “AmazonFresh”.

A dumb idea was, of course, made much worse by the money-no-object execution. Ocado has made an “operating profit” since 2011. But however you look at it, this brilliant business idea never deserved to raise $10M let alone the amounts it did. The valuations were just crazy. So now we have Just Eat plc. To be fair, its business model is not exactly the same. But the valuations are plumb crazy. It raised £94M (for the company) whilst the VC backers got £250M. But the company is valued at £1.4 billion. Its net assets prior to the float were about £50M and it made just under £7M on revenues of £97M. The shares were issued at 260p (see ticker JE on the London Stock Exchange) and now stand at 250p, so investors are crazy too. It describes its business as taking online orders for takeaway food. On page 42 of the prospectus it compares itself to Domino’s Pizza. That has a market cap of £0.9M with revenues of £270M and profits of well … That is a little hard to tell. Its accounts (for calendar 2013) show profits of a little under £18M or just under 11p/share. Its price now is 532.5p, so it is on a (historic) multiple of almost 50x. It claims that its earnings would have more than doubled to 24p/share, a multiple of 22x but for “exceptional items”. Indeed 2012 earnings were 19p. But Domino’s growth has hardly been stellar. In 2009 its revenues were £155M and its earnings 21p. With that kind of growth 22x looks way over the top too me. although many investors disagree since its share price climbed 3.5x from ~ 200p in 2009 to ~ 700p a year ago before falling back. So how on earth is Just Eat worth half as much again? The company would no doubt claim that the reason is that it is an aggregator, not a provider. In principle, a monopoly forum is the ultimate dot-com business. It is rather like being Bernie Ecclestone. You do not have to do any of the grubby, expensive, messy things. The people who pay to populate your forum do those. But you take a piece of everyone’s business. The classic example is eBay. Once you get established as the auction site, everyone prefers to buy or sell on your site rather than any other, because that is where most people go. Of course, eBay also demonstrates that it is not quite that simple. eBay got too greedy. It cut customer service to the bone and made it impossible for small customers to protest effectively about abusive behaviour by large suppliers. But it has survived that and still has a market cap of$70 billion and substantial earnings (PE 25x).

My difficulty with Just Eat is that I cannot see any way they can establish a lucrative monopoly. Takeaway food is inherently local. Strangers may want an aggregator to guide them, but locals know perfectly well the relative merits of the local suppliers. So the only question is whether it is worth the while for a local restaurant to use Just Eat rather than handle online orders itself.

If it is deeply tech-averse, it can easily hire a consultant or a third party to deal with it. It is a simple matter to redirect its (loyal) customers to another website. So I cannot see how Just Eat’s suppliers will ever get sufficiently locked-in for Just Eat to charge premium prices.

It currently charges restaurants an £850 sign-on fee and then (on average) 10.7% of orders going through its systems, plus sometimes it manages to charge customers fees for using debit/credit cards to pay online (and it waits an average of a week before passing the funds on to its restaurants) – see prospectus section 3.2 page 41. Incidentally, the prospectus pdf is bizarrely difficult to get hold of. It took me more than 20 minutes of googling to find it. I was getting close to complaining to the regulator that it was unavailable. (I suspect the company would claim that the difficulty is to avoid falling foul of SEC regulations. I have not looked closely at that issue.)

I reckon that over the next five years its margins are likely to erode faster than it gains new business. £1.4 billion (and falling fast, trades so far this morning at 240p) looks ludicrously high.

## Iran Air flight 655

It was an Airbus A300 on a flight from Tehran to Dubai shot down by USS Vincennes on 3 July 1988 using guided missiles. All 290 people on board (16 crew, 274 passengers, including 66 children) were killed.

It was a classic mix of incompetence leading to the accident, and deviousness in trying, and eventually failing, to cover up what actually happened.

At first sight the shooting was hard to believe. The flight had a stop-over at Bandar Abbas. It took off normally from there at UTC 06:47, 27 minutes late. It flew normally down the commercial air corridor Amber 59, a 20-mile wide direct route to Dubai airport. It followed the normal flight plan of climbing steadily, aiming to reach 14,000 ft, then cruise briefly, then descend to Dubai. Its transponder was broadcasting the regular civilian code (“Mode III”, easily distinguishable from the military “Mode II”). When it reached 10 miles from the Vincennes still climbing, it was shot down on the basis that it must be an Iranian F-14 descending on its final attack run.

The Vincennes was at lat 26.513056N, long 56.015833W, 10.8 miles from the nearest point of the Iranian coast (the little island of Hengam, just south of Qeshm), inside Iranian territorial waters, and was in the process of attacking small Iranian gunboats which it had lured out with a decoy “Liberian ship” the Stoval. It was neither a ship, nor Liberian, but essentially just a transmitter to fool the Iranians into coming into range of the Vincennes’ helicopters and various other US ships that were in the area.

Indeed it turned out that the US had been engaged in a secret naval war in the Gulf for some while, a war for which it did not wish to seek authorisation under the War Powers Act.

The Vincennes had all the latest kit, known as Aegis:

This was a complex computer system linked into umpteen radars, intelligence feeds and other systems, designed to allow the ship to engage up to a hundred air or surface threats simultaneously.

It performed flawlessly.

The snag was apparently that the crew did not believe the information it was giving them. They expected the plane to be a hostile Iranian plane rushing to defend the gunboats, so that is what they managed to see.

There was also a classic time-zone mix-up. The ships clocks were on UTC + 4hrs, whereas Bandar Abbas was on UTC +3.5 hrs. So although the crew knew all about the IA655, they knew it could not be the plane on the radar, because the timing was wrong.

But part of the problem was that the Vincennes had too much information. All kinds of people were intercepting, real-time, the communications between IA655 and the Bandar Abbas control tower: GCHQ and NSA (with listening stations in Oman, including Goat Island), an AWACS plane (a Grumman E2-Hawkeye) above the Gulf.

[Goat island, Oman. All maps courtesy of Google Maps]

All this information may not have been much help to the captain of the Vincennes faced with only a few minutes to make a decision as the plane closed on his position at about 6.5 miles/minute. Having said that, I am not inclined to be particularly sympathetic. There was precisely one scheduled flight out of Bandar Abbas that morning, IR 655, due to depart at 09:50 local time = 06:20 UTC. It was flying direct to Dubai, which would take it directly over the Vincennes. Clearly, avoiding downing that flight was a priority.

But the information was certainly a problem afterwards. Aegis provided a flawless audit trail. It showed that the crew had imagined things when they thought the flight was descending. It did nothing but ascend. There followed a lengthy period of giving out a mixture of flat untruths and heavily redacted truths, but the truth did emerge several years later.

Full details, and amusingly commented original documents etc, are available on Charles Harwood’s site.