IPL Day 16

Kolkata did it again. It looked like they posted a decent score, then had chance after chance after chance to polish off the Punjab innings - but dropped them all. Yet despite the repeated missed catches and run-out opportunities, somehow the game and our hopes were dragged out to the very last ball yet again. And once again we lost.

In the second game, a surprise for two reasons. Firstly a comparatively one-sided game, nearly two overs to spare in the run chase. Secondly the result itself, mighty Mumbai mauled by the resurgent Challengers who seem a different (better) side now that KP left them.

Meanwhile, leaving aside the massive subject of swine flu, the more relevant issue at least in terms of my current routine is the rather more basic question: does bacon kill you? I saw a good answer to that question today, written by Professor David Colquhoun. Worth reading. Unless you're a quack or a homeopath.

Campaign Project and Stock Management

To maintain effective marketing, or to run effective projects, both require constant cycles of prediction and measurement.

For example, first you measure the current one, use that to plan and predict the next one. Measure that one. And so on.

Having been through plenty of campaigns and plenty of projects, I know there are some things that I can predict well. Perhaps an odd predilection for stats helps. Of course we all get some estimates wrong, but the more we do the better we estimate. And I've done loads.

But there are some things that I cannot predict. I cannot predict share prices without insider knowledge. So I agree with what Stephen Dubner wrote on Freakonomics a while back:

Here are a couple of stock-market headlines I’d love to read one day:
“Stocks Surge, Reasons Unknown; May Be Nothing More Than the Random Fluctuation of a Complex System”
or:
“Stocks Dive: Three First-Movers Sold Hard and Then Everyone Else Inexplicably Followed”

Of course, in the second case, if the first-movers sold hard, if you believe that others will also sell hard, then you should sensibly follow.

So my own investment strategy, still holding all my boom-bought shares as they go bust one by one, now it looks very foolish indeed.

Wolf Work

In the last post I raised a general set of observations about the different types of human personality, but to me they seem even more applicable and relevant when applied to daily interactions in the office.

Whenever a problem arises, there are three ways to approach it:

1. It's part of my job to sort these things, all I can do is to try to fix it and put process in place to reduce future recurrence
2. It's a disaster, this is currently the most important issue in the whole organisation, stop everything and get it sorted
3. There is no such thing as a problem, only an opportunity.

Category 2 obviously belongs to those who may lose their nerve, their poise, their rationality and often their temper when put under pressure. And there may appear to be significant overlap between categories 1 and 3, but 1 acknowledges that there are things outside their control, 3 does not accept that.

Are You Lucky?

This is not a reference to last night's Apprentice, not about trucks having fun, not a reference to an old BBC study that showed how you can make your own luck anyway.

But is your glass half-full or half-empty?

First rule out one exception - clearly anyone born with a serious genetic defect or into war-ravaged famine-festering slums is less "lucky" than a broadly healthy person in the developed world. But let us just look at the different ways that people here deal with misfortune …

The usual distinction into optimists and pessimists is too obvious, there are at least three different ways that people tend to deal with the same unfortunate incident:

1. It's no big deal, we understand that it is human nature to remember the occasional disruption more than the usual smooth journey, but these things even out
2. I'm so unlucky, the breaks never go my way, my queue is the slowest, moan moan moan
3. I love the taste of defeat because it only makes me stronger. It's all of part of an almighty master plan.

There is overlap, people behave differently in different situations. Funnily enough, although category 2 (the traditional pessimists) are the pariahs of society, it is category 3 that can sometimes be the most annoying.

Appeasing the gods

I spent over a year developing a Management Information System for a large insurance company based in East Anglia. I worked with the actuaries on the business processes of matching payment schedules to claim rates, and we created systems to allow dynamic exploration of the data to accurately identify the profitable and worthless market segments.

But from Marginal Revolution this week: some people buy insurance because they think it will prevent the bad thing from happening?

It was actually a reference to an excellent article in the NY Times by John Tierney: We may not slaughter animals anymore to ward off a plague, but we think buying health insurance will keep us from getting sick. Our brains may understand meteorology, but in our guts we still think that not carrying an umbrella will make it rain.

And one professor of psychology claims in the same article that "rationalists were just as likely as superstitious people to believe that insurance would ward off accidents".

Obviously I don't buy it.

The Polish Invasion

I tend not to put up pictures here unless they are original. So a closeup of the wall of my house.




There are four national "broadsheet" newspapers in the UK, the two biggest selling have a largely conservative anti-European leaning. I'll quote one article from each.

From The Telegraph: Almost nine in 10 new jobs created over the past decade have been taken by foreign-born workers despite a sharp increase in the number of skilled British workers, official figures show.

From The Times: Market experts estimate that about 90% of the plants on sale this Easter weekend, whether in DIY chains such as Homebase or in specialist nurseries, have come from overseas.

The Telegraph article throws out the slightly xenophobic sensationalist headline without examining it closely. Very tabloid. It suggests that you (or I) could not be both "foreign-born" and also "British".

The Times article does suggest some further explanation:

The scale of the shift is illustrated by sales of clematis. A series of strains developed in a Polish monastery now dominate the market. They were created by Brother Stefan Franczak, a Polish Jesuit monk, who started breeding clematis in the 1960s after finding seedlings in the garden of his monastery just outside Warsaw.

Franczak’s work was part of a much wider interest in plant breeding that helped create a network of renowned laboratories across eastern Europe.

By contrast in Britain, this coincided with Margaret Thatcher’s decision in the 1980s to slash Britain’s investment in such laboratories — which led to most of Britain’s plant-breeding centres being sold off or shut down.

So it is Maggie's fault?

Feedback Measures and Ideals

Yesterday I wrote about measures and feedback.

Measures. It was the UK Budget this week. Arguably the biggest economic measure decided in each budget is the proportion of revenue that is redistributed to government services such as national health and education. It seems that everybody is unhappy with it. But whereas some think it is too high, others think it is too low. So everybody may think it's wrong but maybe it's about right.

Feedback. I got some negative feedback today. Well it was actually just a quick comment from a (real world) visitor, she said that she did read the blog occasionally, but now thought the posts were getting too long. But from what I see the vast majority of posts elsewhere are far longer than these (I maintain that it is often harder to say something original concisely than to say it comprehensively) or else they are just quick "look what I've seen on the web" posts. There are at least two points of common equilibrium, and I sit between them.

So I conclude that it may be easier to determine a single ideal tax rate than a single ideal post length. But as yesterday's was long by my artificial self-imposed standards, today I'll stop here.

Who dares wins

There seems to be a developing trend here. Still one post per day, Monday to Friday tend to be about technology or marketing or science or superstition, but as evident yesterday, romance or sport can intrude at the weekend.

While we slept overnight, the England cricket team collapsed horribly to lose to New Zealand. I am sure that various commentators are going to write scathing match reports, it's reasonably easy to write something different. I am sure that the Cricinfo site will come up with pages of stats, but they have not done so yet, so I'll get in first with some almost miraculous coincidences.

One day games used to be 55 overs per side. The New Zealand second innings today was declared after exactly 55 overs, that is 330 balls. The England second innings was ended after exactly 55 overs, 330 balls. What are the chances of that?

And looking at the first innings - at the close of day one, New Zealand were 6 wickets down for 282 runs; at the close of day three, England were 6 wickets down for 286 runs (just one boundary away).

But the key bit of synchronicity is in the run rate. In the first innings, to two decimal places, England scored at only 2.00 runs per over. In an era where Australia (and even India) push for 4 or 5, that is an affront to the paying spectator. England's run rate in the second innings, exactly 2.00 runs per over.

On the same pitch in the same match, New Zealand scored at 3.3 Because England did not appear to try to win, England deserved to lose.

The Perfect Recruitment Process?

I am only an advisor to the client, but I am aware of their difficulties in recruiting suitably skilled staff. When doing any recruitment, whenever you can't find the "right" candidate, you always need to ask if you are doing something fundamentally wrong, if you need to tighten or relax the criteria.

I know that the client was looking to fill a contract vacancy last year, and they utilised a recruitment agency to advertise the position and pre-screen the candidates. Sixteen people passed these initial tests. In a nonexistent ideal world, every one of those candidates would be perfect for the role. But that is only ideal for the employer. It is not necessarily ideal for a society to be in a situation where fifteen candidates are all equally capable of the job, invest equal time and effort and are equally disappointed by the outcome.

But as a practical target, if you are looking to judge if your criteria were too strict or too lax, if about half of the applications are rejected at each stage of the process, then you shouldn't be going too far wrong.

Let us look at historical (hypothetical) example. Out of those sixteen pre-screened candidate CVs, eight were considered suitable for interview. Of those eight, after initial phone discussion about the logistics of the position, four actually turned up for interview. Of those four, two were offered the position. Of those two, one accepted it.

So I conclude that, in this case, the client was doing things about right. I only wish they would do things more quickly.

Who checks your adverts?

I listened to a couple of senior marketers from Google last week. Their themes were consistent, how the developing models behind concepts such as AdWords were changing the nature of the industry. They quoted an old marketing favourite: "I know half of my advertising is wasted," John Wanamaker famously said; "I just don't know which half."

And senior management at Google are not the only people to use that and similar quotes as a sign of the changing times. I often say it too. The work-related portions of this blog constantly suggest the identical message - modern marketing should be measurable, we should know which half works.

However, yes you should learn the results, but you may not like what what you learn. The Virtual Economics site provides an excellent summary of the potentially uncomfortable results. I won't repeat everything here.

But as I see it - in the TV advertising world, those who you would most like to target with adverts for the funky new gadgets that you are trying to sell are precisely the people who are most likely to have the PVR that lets you skim past all the ads.

And even in the new web world, where these stats are more easily measurable, for example by WebTrends or Sophus3 or a myriad of other companies, still less than 1% of visitors to a site in a month click on an ad. Easy stats. Given enough visitors, still maybe ok.

But who are these visitors? I will quote directly from the first reference to the original AOL survey that I can find: Not surprisingly, they look at sweepstakes far more than any other kind of content. Yes, these are the same people that tend to open direct mail and love to talk to telemarketers.

And three conclusions about these visitors drawn from the survey here:

• More representative of lower income households than the average user.
• Less educated than the average user (or from less-educated environments in the case of minors).
• More likely to live outside of the major metro regions.

Are these the people who you most "want" to click on your banners?

Estimating Project Delivery Time

I have spent a lot of my working life doing IT Project Estimation and IT Project Proposals, it is not the most interesting subject in the world, but it is important, and I touched upon it in an earlier post.

I could throw in a few key words used to manage the estimation, words like originality, uncertainty, statistics, experience, reductionism, but it would need thick volumes to explain them properly. What is interesting is why people keep getting their estimates wrong.

It is obvious that people tend to over-estimate their own abilities. That is such basic human nature and so well-documented that we may as well just call it fact.

But many people, even educated ones, seem to have trouble with simple extrapolation, even for things as commonplace as car journeys. The British Psychology Society just published a simple study that asked:

for both pairs, just make an intuitive judgement about which jump in speed will make the largest difference to your time of arrival (i.e. save the most time):

a)Travelling at 50km/h instead of 40km/h.
b)Travelling at 130km/h instead of 80km/h.

a)Travelling at 50km/h instead of 30km/h.
b)Travelling at 130km/h instead of 60km/h.

If you're like most of the participants in Svenson's study, you will have assumed that option (b) in both pairs is the most time saving.

That guess is simply wrong. In each case, option (a) saves more time. However far the journey.

How many people will select the same option as you?

The slashdots site had a recent "survey" that prompted a lot of discussion. I won't try to collect your response here, that would count as one of those reading "interruptions" that I try to avoid, but it had just one simple question in it:

How many people will select the same option as you?

• 0%
• 1-25%
  
• 26-50%
• 51-75%
• 76-99%
• 100%

It looks very obvious. Cognitive Daily summed up the basic proposition:

At first pass, you might figure that there are 6 possible responses, so if people respond randomly then about 16 percent will choose each answer, so the correct response would be 1-25 percent. But of course, if everyone used that same logic, then many more than 25 percent of respondents would choose that answer. The next logical response would be to move on to 26-50 percent. But once again, if everyone used that logic, the correct response would be higher still.

Very sensible analysis. If over-analysed then it might seem paradoxical. But basically common sense. I do have one further point, for as with my dislike of interruptive multimedia, another blog rule is that I do not just post stuff without adding an original thought. Incidentally, I guessed correctly. Even more incidentally, I work in marketing.

In marketing, we cannot stock every alternative, so we regularly have to predict the "most popular option". Of course that is not necessarily the option that we personally would choose, despite overblown arguments for "you must have personal experience of what you are selling".

For we try to predict the option that the consumer will choose, but that is not necessarily the option that the consumer should choose. We also need to account for the influence of other people's selections upon that choice. We need to guess the proportion of people who will behave rationally, the proportions who will act in isolation or in collusion. We should determine the evolving Nash equilibria. We should resolve the prisoner's dilemma.

The fine art of marketing

I work in marketing. But as with psychology, is it art or is it science?

Marketing is not about award-winning advertising or sophisticated value propositions or refinement of sales techniques. It is not about "the creation of something original and beautiful", although that should be a required by-product. It is the creation of something effective. And it is only effective if we can measure the results.

A few years I worked with a great character at a nearby client site. Although very creative and nominally a business marketing guru, he was extremely capable of delving into technical detail, for example to verify suspected anomalies in reports. However he was also perfectly aware of the time and effort required to get the precise answers, and the consequent practical necessity of making broad yet inaccurate estimates instead. And he would often justify his approximations with the same phrase: "this is an art not a science".

I completely understand his sentiment. I fight the same questions every single working day. But marketing is not art. It is just that we rarely have sufficient time, resource, skills or commitment to do the required scientific measurement and to properly evaluate its effectiveness. But marketing is science.

Interactive Marketing and Patient Choice

When we prepare direct marketing, in order to enable calculations of incrementality and return on investment, we often use the principle of a control group - that is a randomly selected sample of the population chosen using exactly the same criteria as used in the main selection. However as targeting improves then target populations get smaller, and then the control group can become statistically insignificant. It is a significant problem.

There are various statistical methods that can help to estimate result reliability more accurately, from simple chi-square tests to complex Fisher analysis. Every day we need to make decisions about balancing the time and effort required for further predictive analysis with the expected impact on statistical result validity and campaign execution timescales.

Sometimes these same issues crop up in far more critical arenas than in the marketing of optional consumer goods. I have suggested that you are more likely to die in a hospital than anywhere else. I still suggest that the marketing of patient choice is over-hyped.

I recently learned of research published in an American medical journal that compared mortality rates in different hospitals. The article looked at very similar operations, and questioned whether those hospitals with a zero mortality rate did better than those with a non-zero mortality rate. Even just judged on that single measure, the subsequent evidence concluded that hospitals with a history of zero mortality subsequently experience mortality rates that are the same or higher than those of other hospitals. Yes, choose a hospital with historically zero mortality and you are more likely to die there.

This is relevant to public health policy in England too. There is regular shouting for more patient choice, particularly by the main political opposition party, but the usefulness of those choices needs to be considered in the light of this kind of statistical evidence. More importantly, you personally may need to make life changing decisions if you ever have to choose a school or hospital.

And a very similar issue affects how we interpret football results :)

Random numbers and letters

Measurable marketing requires the use of control groups. Control groups require the selection of random samples. Ensuring statistical validity across overlapping groups, even with normal distributions, can be difficult. But randomization is easy.

The New Hampshire Democratic primary has just given Hillary Clinton a shock victory, whereas Barack Obama was predicted to win by the pollsters. I am not going to comment on policies, just on the most pathetic randomization that I have ever heard seen. Not in the poll sample selection, which can be difficult. But in the ballot itself, which is ridiculous.

It has been suggested, and tests have proved, that a name listed at the top of the ballot is more likely to be picked than when placed at the bottom. The effect could be up to 3%, but I think we still do not have enough data yet to quantify exactly.

As ABC News reported: This year, the secretary of state changed the procedure so the names were alphabetical starting with a randomly selected letter, in all precincts. The randomly selected letter this year was Z.

Now that is just ridiculous. So given the lack of surnames ending in Z, the order was simply alphabetical. The fact that "this year" it was Z implies they are going to maintain this crazy idea. So a name like Miller is nearly always going to be above a name like Nader, in fact about 96% of the time. How is that random? And a name like Cleasby is always going to be above a name like Courtney.

Life is unfair, but ballots should not be.

Maradona and Statto

I usually play football twice per week. One of those games involves a group of us who have been playing at the same time each week for nearly a decade. The sides vary, but we have records of the match results, and also on how often each player wins. But in today's lesson on how stats can mislead, a reminder that win ratio does not correlate to talent or ability.

It should be obvious, but to help understand why, assume that one player (for the sake of argument let's call him X) is much better than everyone else. Fair enough, X would be expected to have a decent win ratio. But look at the effect on everyone else.

When we are choosing sides before the games, we make a conscious decision to balance them. And if we have said that X should expect a high win ratio, then those players who are almost as good will naturally have a correspondingly low win ratio. Yet we have just implied that they are amongst the best players that we have!

Common sense? Or just an excuse for poor results?

Size matters

At one time, every large human group was a small human group. What are factors that could have made it grow faster than other groups?

While in a minority in a region, it could encourage isolation from competing ideas. Within families, the group could discourage contraception and encourage large numbers. It could encourage conversion into the group.

It could discourage dissent. Alternative groups could be suppressed by removing competing ideas or by removing competing individuals.

But total numbers in the group will not rise if people leave at the same rate that they are joining. So it could be made difficult to leave the group. Once any group has taken control of land and business, there will inevitably be social pressure on others to conform. Or penalties for leaving the group could be made even more explicit.

Many of these factors could apply to any group that is growing. But does the fact that something is growing mean that we should encourage it?

Paramedics

This is a blog for common sense (and for unnecessary puns) - not for discussion of metaphysics or formal philosophy.

In a detective story, the initial diagnosis may look like it is causing a paradox. But once you have watched or read through to the end, it is always the case that there was no paradox, just an incorrect deduction.

In my example yesterday, we had two conclusions, one drawn from each line of data. Those two were fair enough. But it was then just lazy or presumptuous to draw another conclusion from those two conclusions, instead of from the data.

So once we understand the data then there is no paradox.

That only leaves quantum theory ...

Touching the Data

This note is not really new material for those who work with me, it is one of those examples that I have been dredging out for year after year. But as I've just mentioned the reverend Simpson, and as this blog is holistic, it seems appropriate to head back to the analysis and evaluation of marketing activity.

If programme 1 performs better than programme 2 in a first phase, with a significantly higher response rate, and if programme 1 also performs better than programme 2 in a second phase, - then that does NOT necessarily mean that programme 1 has performed better than programme 2 overall. Even if programme 1 "wins" subsequent phases.

Nothing to do with creative or external factors, just really simple maths. Hence common sense.

But most people seem to understand this better when related to football than to marketing. With my invented numbers:





Easy.

I just need to recognise, manage and control this effect in real campaigns every single day.