omenti

observation

Expensing snow tires on a company trip to California threw up a few flags.  It seems Friedolf was trying to make it up 7,500 foot Table Mountain in California in the Winter and snows were the only option.  He was a key player on the Bell Labs Telstar team responsible for its solar cells and radiation hardening.  The team was small by today's standards and little was known about how to build the world's first successful active telecommunication satellite.

 

Table Mountain happened to be the location of a Smithsonian/JPL solar observatory.  There was a question about how much power would be available for the satellite so Friedolf decided to make some calibrated measurements near sea level and on a mountain that happened to have accurate instruments.  He then used a simple model - almost back-of-the-envelope and came up with a projection that turned out to be correct within a few percent.

 

He had a fair for clever instrumentation and the late 50s he was involved with Sandia Labs testing some nuclear devices.¹  The trick was to put the instruments close enough to the device that they could record what was happening and get the signal a safe distance away before the instruments and cables were vaporized by the blast.  All of this was worked out to the nanosecond and he built a reputation for getting very close to the explosions.

 

In science observation is centrally important.  Most of your time as an experimentalist or observationalist is spent designing the observation and making sure it is doing what you think it will do.  You have to trust it to the point where anything that strikes you as being odd really is odd - that is where the real pay dirt usually lies.  Even if you don't see anything novel (which happens most of the time), you need to be certain you didn't miss anything.  Then you run the experiment or your observation session and take your data.  Finally you sort through what you saw.  The shortest amount of time spent is usually the experiment/observation phase.  In Friedolf's nuclear bomb measurements the event might last a few milliseconds, but designing and building the apparatus - thinking through exactly what it should do and verifying it works - often took several person years.  The analysis phase could also run for person years.

 

Sometimes an observation can run a very long time, but you need to be certain your observation is up to catching the important moments accurately.   All of this takes a lot of careful planning and worrying about the edge cases.  You need to fail several times to realize where the pitfalls are - people who are good at the game can tell you more than a few disaster stories from earlier in their careers.

 

I was thinking about this recently as I spent some time with a company that is probably the best mix of technology and art I've seen.  They build most of their own tools largely because they are at the leading edge in a changing world and partly because they consider understanding their tools a core competency.  At the same time their knowledge of their end user is qualitative rather than quantitate. Their understanding of how their customers relate to their product is based on emotions and any modeling they could do wouldn't be as effective as their current technique.  They currently see their process as good enough, but they continually test their assumptions.

 

I've complained about the use of the word "data" as it is often misleading.  It is important to think about how the measurements to be made, how they will be made, how trustworthy they are, how they will be analyzed and how they can be used.  If you can't do this entire chain robustly, you are probably in over your head.  So rather than terms like data-centric, data-driven, or data-whatever, I prefer something like carefully observational. 

 

In the early 90s I found myself in Mesa, AZ at an AT&T data center doing strike duty.  A few of us got curious about the process leading to some discoveries about a rather messy system.  Important corporate decisions were being made on what seemed to be straightforward call information - hundreds of millions of call detail records.²  Over the years some inaccurate assumptions were made about how CDRs were generated and recorded.  Sorting out what was really happening was a difficult slog - a few person years of work by a few physicists and a mathematician, but we were able to show that about two percent error in the billing process. This is by no means an isolated case.  I've seen it in numerous industries.  "Data" taking and assumptions about the data require a lot of thought and any changes to the flow need to be carefully thought through if these streams are important to your business.

 

Recently Pip put together a list of questions to ask companies to determine how data driven they are.  I'd be curious to see how dedicated they are to careful observation - are they observationally driven?  Do they understand where they must be qualitative and where they can be quantitate? Can they build models from their observations and use these for business decisions?  Can they use the information they gather, along with models they create and perform experiments on potential scenarios?  The possibilities are quite rich, but certainly not for all businesses.  

 

Is the work sufficiently important that the organization has an internal competence or does it or do they outsource some or all of it?  What is the background of the folks doing the work?³  Are they in a position to deeply understand what they are measuring, analyzing, modeling and interpreting?  I worry many organizations don't understand the process deeply and there are a lot of people who can make pretty charts that are - well - pretty, but not much else.  

 

I'm even more impressed by those who have thought the problem through deeply enough to realize there are parsimonious choices that can be made.  The company I visited comes to mind as does Trader Joe's - a company that makes effective use of human intelligence of its sales people.  And Friedolf...  the question came up as to how many solar cells should be on the Telstar.  He noted the craft had to be spherical given the type of stabilization it had to employ and that some surface area would be used for antennas and some instrumentation.  His answer was "cover everything else with solar cells - that will dictate everything else"...  of course he was right and no numerical measurements were required to come to the conclusion.  I'm impressed by organizations that sometimes see the purchase of snow tires as an important contribution towards understanding something deeply enough.

 

__________

¹ Bell Laboratories administered Sandia for the government.

² CDRs have start and stop time on calls, toll rate information, some switch information and not much else.

³ People from the natural scientists tend to excel - science is based on getting this right.

__________

Recipe Corner

 

A friend sent this recipe recently - I modified it a bit and made it with Cora Cora and Blood oranges.  Excellent seasonal dessert that is super easy to make. It is really good with a bit of crème fraîche.

 

Oranges with Rosemary and Honey

 

Ingredients

 

° 2 oranges at room temp

° 2 blood oranges at room temp

° 1/4 cup of honey

° 1 tbl olive oil plus some for a pan

° 1 sprig fresh rosemary, separate the leaves and mine

° a good finishing salt (Malden)

 

Technique

 

° peel the oranges removing as much pith as possible and slice about 3/8" thick.  Keep the blood oranges separate as they bleed color

° set the oven rack about 6" below the broiler and preheat.  Cover a pan with aluminum foil and brush with oil.  Put the blood oranges on one side to keep juice bleeding to a minimum.  Drizzle the oil and honey over the slices.  Sprinkle on the rosemary and salt

° broil until the oranges start to caramelize.  About 4 minutes in my oven

° serve on a platter alternating orange slices.  Pour the pan juices over the oranges. 

 

three fall vegetable tricks

In many areas late Fall vegetables are still good at this time of the year.  I like to cook and experiment a bit along with trying to understand a bit of the chemistry that goes on in the process.  There have been more than a few failures, but also a bit of insight and a few successes.  Rather than the conventional post I'm going to offer a few recipe outlines that are a bit unconventional - recipes that work with late Fall vegetables.

 

A somewhat larger than average recipe corner

 

Roasted Sweet Potatoes

 

I'm a big fan of sweet potatoes in many forms, but roasting makes them incredible - usually at least.  The problem is they are variable and often are, well, not great.  A bit of sweet potato research and experiment provided a technique that is much more reliable than regular roasting.

 

Sweet potatoes have a lot of starch and sugar in them.  Depending on their age the starch to sugar ratio varies enormously.  The great tasting ones, at least to my taste, have a lot of sugar in them that respond so well to the intense heat of roasting.  They also happen to have an enzyme that breaks starches down to maltose (a sugar composed of two glucose molecules).  The enzyme is most efficient between about 140° and 170° F.  It is somewhat active at lower temperatures, which is why some sweet potatoes are sweeter than others. All you have to do is bring it up to a temperature between those values (above 170° destroys it).

 

This is where experimentation comes into effect.  I sliced up several very different sweet potatoes and dropped them in a 150° F water bath.  The right way to do this is an immersion circulator, but keeping a large pot of water at about 150° works well too.  An hour worked well on all of them - just don't let it go above about 165° to stay away from trouble.  Now just dry them off and roast them to get a wonderful caramelization in addition to the complex chemistry that comes from the Maillard Reaction (the wonderful reaction between amino acids and reducing sugars in the food that give hundreds of complex compounds and wonderful flavors and aromas).

 

Tricked out Roasted Sweet Potatoes

 

Ingredients

 

°  3 average or 2 very large (2 kilos total) sweet potatoes quartered and cut into half inchish slices

°  6 tbl extra virgin olive oil  (you can use 3 tbl of just ok olive oil for roasting and 3 of a higher quality olive oil later)

° kosher salt and freshly ground black petter

° 1 tbl honey

 

Technique

 

° Drop the sweet potatoes in large saucepan and cover with water.  Heat to 160°F, take off heat, cover and set aside for an hour.  (this works well enough)

° put oven racks on the upper and lower middle positions and preheat the oven to 400°F

° drain the s.p. and place in a large bowl.  Toss with 3 tbl of olive oil and season with salt and pepper. 

° spread the  s.p. on two baking sheets and roast until brown - about 30 minutes.  Flip them and roast 'til the backside is brown and all is tender.

° put them in a large bowl and toss with 3 tbl olive oil and the honey.  serve!

 

 Roasted mushrooms

 

Moisture is not your friend, but not where you might suspect.  Many will tell you not to wash mushrooms - that they are clean enough with some rubbing and that the water will cause problems.  I'm not that trusting and weighed some waterlogged mushrooms - they only picked up about 5% of their weight in water and that was confined mostly to the exposed surface.  

 

There is a problem with water though ..  Mushrooms like roasting at a somewhat lower temperature - about 350 - 375 F°.  They stay below the boiling point of water until the water has evaporated - the energy from the oven is going into boiling water rather than roasting mushrooms.  As you cook them a pool of water forms on the bottom of your (hopefully) rimmed roasting pan.  The trick is to pour the excess liquid off about 15 minutes into the roasting process.  You'll find they brown much better and the process is faster.

 

Faster and Browner Roasted Mushrooms

 

Ingredients

 

° 1 kilo (about 2 pounds) of cremini or button mushrooms.  Wash and quarter them

° 1/4 cup of extra virgin olive oil

° kosher salt and freshly ground black pepper

° about a dozen springs of thyme or rosemary (both are *great*, but I wouldn't mix here)

° 2 tbl chopped fresh herbs - chives, parsley or whatever you like

 

Technique

 

° oven rack to the center position and preheat to 375°F

° toss mushrooms in the olive oil, salt and pepper

° spread evenly on a foil lined rimmed baking sheet and sprinkle the thyme or rosemary sprigs over the mushrooms

° roast for about 15 minutes and then drain off the mushroom liquid (you might want to keep it as it is flavorful)

° return to roasting until brown but tender - maybe 25 or 30 minutes more.

° place in a bowl and toss with herbs.  serve

 

Roasted carrots

 

Carrots can be sweet to begin with and are wonderful with a bit of caramelization.  The problem is they shrivel at the temperature needed to roast.  Unlike mushrooms you want to hold some water in.   The trick is to not peel them and par-boil them first (the same is true with parsnips, which can be spectacular roasted)

The recipe I give has a few non-standard ingredients.  Try everything - carrots like spices.  Toss them with olive oil and some sort of spice blend.  Curries are a nice place to start.

 

Roasted Spiced Carrots

 

Ingredients

 

° 1 kilo medium carrots with the ends trimmed. Quarter lengthwise and cut into 2 or 3 inch segments

° kosher salt

° 1/4 cup crème fraîche 

° 2 tbl harissa paste (a North African chili paste.  Find it in speciality stores 

° 1/2 tsp ground cumin

° 1/2 tsp freshly ground black pepper

° 1/4 cup extra virgin olive oil

° 2 tbl freshly chopped cilantro

 

Technique

 

° oven rack in the center and preheat to 375°F

° put carrots in a large pot, cover with water and season heavily with salt.  Bring to boil, reduce heat and simmer until the carrots just start to become tender - 4 or 5 minutes.  Drain and let dry - set aside.

° Season crème fraîche with a bit of salt and refrigerate until ready to use

° Combine harissa, cumin, pepper and olive oil in a large bowl. Season with a bit of salt.  Set half aside in a small bowl.  Add the carrots to the rest and toss the mixture.  Spread evenly on a foil lined rimmed baking sheet and pop into the oven

° Roast until caramelized - about 40 to 45 minutes turning at about 25 minutes.  

° Transfer to a large bowl and toss with cilantro

° spread the crème fraîche on a serving plate and artfully:-) arrange the carrots on top and drizzle the remaining harissa mixture over everything.  serve

 

__________

Finally a tool recommendation.  

Roasting is wonderful, but there are times when you need a bit of portable and or local roasting.  A torch is a lovely tool, but they can fall over with unfortunate results and and leave a torch taste.  The later is more a product of burning the food with heat that was too intense rather than unburned gas products.  Dave Arnold (an all out crazy guy and thinker about cooking) has a great gadget that addresses both of these problems - the Booker and Dax (named after his sons) Searzall.  Think of it as a portable mini salamander for under a hundred bucks.  It is a kickstarter project.

 

 

metastable states, richard milhous nixon and the apparel business

Curiosity about apparel and fashion have become something of a hobby and I'm putting together some thoughts for a talk to people with a lot of background knowledge.  In thinking about possible trajectories, I've been spending some time looking at how we got to where we are - not only has their been a dramatic move to mass produced apparel, but there has been a dramatic shift towards very low cost production, low quality clothing and shrinking delays between design and delivery.  I won't go into much of that, but something really interesting happened during the 70s - a period when several forces conspired to rapidly accelerate change in both the industry and the consumer.  But first a bit on metastable states.

 

Consider a landscape that is flat on the left moving towards some hills as you move to the right.  An object under the influence of gravity will move to the lowest point.  A ball that finds itself on the plain won't magically move up the hills.  If the ball is in the valley labeled b it won't move to the plain or the valley c unless you give it a kick.  Even though  being on the plain is theoretically more stable, it is stuck in  a stable stable.  If I get it a kick to the left it can roll up the little hill and then zip downward.  Likewise a harder kick to the right could land it in valley c.  a, b and c are all stable - the overall system has three metastable states.  The ball can only move from one to the other if some outside kick is applied.

 

In the early 70s Richard Nixon was faced with several trade issues.  Cheap imported clothing from Taiwan, Japan, Korea and a few other countries were on the radar screen as American apparel companies folded under intense competition.  Nixon had positioned himself as a hero of the American worker and ginned up multinational support with European countries that having similar problems and a massive trade agreement - the Multifiber Agreement or MFA - was signed.  

 

South Korea and the others suddenly had a big problem.  Enormous factories and production capacities far beyond the trade limits.  The MFA was killing their business.  But it was a very complex and targeted agreement and, in that focus, came a way out.

 

About this time a few people in Pakistan were frantically looking for new businesses to pull the country out of total poverty. A chance meeting took place with Daewoo in South Korea and an agreement was hammered out.  The MFA quota was specific to South Korea, but didn't say anything about Pakistan - after all - Pakistan was a basket case.  

 

Daewoo set up a facility in Pakistan and trained Pakistani managers.  Problem solved.  Similar agreements were made between other pairs of countries and the spiral to the lowest cost producer ramped up.  

 

Nixon, in trying to focus so tightly, had supplied just enough energy to kick the ball up the hill so it could roll into another valley.  The system had been perturbed and settled into a new metastable state. And other influenes were being felt.

 

Container shipping was emerging as as a way to move massive amounts of goods to ports around the world at very low costs.  It was often more expensive to ship clothing within the America than from Pakistan to an American port.  People began to get used to cheap apparel - it wasn't made very well, but cheap trumped all and closets began to fill.

 

The impact of the shift started to take its toll on American firms, but now Reagan President and he had little appetite for protecting union workers.  Buying patterns were changing.   Companies like Daewoo moved upscale, but did so with high production and lower quality.  What was being marketed as mid-quality wasn't built to the standards of a decade earlier.  Vintage boffin Jheri notes you don't find much that is newer than 1980 - the quality had dropped so much that little of it is worth repairing.  

 

Clothing had become an expendable.  There is an interesting connection between this and the notion of what fashion is.  Fashion was now being dictated more than assembled and marketing changed instep.¹  This continues to accelerate and now we are in the era of fast fashion - where a design can move to production in a few weeks and to the shelfs in six weeks by boat or almost immediately by air.  There are so many rich threads (pardon the pun) and now, with entry of a few new technologies and rich "word of eye" communications on the Internet, we may be on the edge of a new major change.  It will take time to move a $1.7T industry ($350 to $400B in the US), but there has been massive change in the recent past.  

 

Perhaps the industry will get excited to the point where a new metastable state (or states) will be found...

__________

¹ The history is much more complex and interesting than this - consider this a 100 mile view.  There are a number of fascinating books on the subject.  For the current state I recommend Overdressed by Elizabeth Cline and for the current state of production and a bit of history try The Travels of a T-Shirt in the Global Economy by Petra Rivoli.  Jheri thinks about this quite a bit and has a guest post here.

__________

Recipe corner

 

A winning pecan pie that isn't as supersweet as those made with karo syrup.  I realize good pecans are nearly unobtainum this year.  A friend sent me a nice supply, so I'm set for the season.  I love to bake, but tend to give things like this away as I don't want too much of this sort of indulgence.

 

Maple Pecan Pie

 

Ingredients

 

° an unbaked pie shell (use whatever technique you prefer.  I like a good graham cracker crust or a convention flour/butter crust for this one)

° 215 g (about 2 cups) of coarsely chopped pecans 

° 55 g dark brown sugar

° 25 g tapioca

° 3/4 tsp salt

° 240 g (about 1 cup) grade B high quality maple syrup (the dark brown grade)

° 60 g heavy cream

° 1 tsp vanilla extract

° 2 large eggs

 

Technique

 

° preheat oven to 425°F

° lay the pecans in the pie shell

° whisk together the sugar through salt in a medium sized bowl.  Then whisk in the maple syrup and cream.  Finally whisk in the eggs one at a time.

° pour over the pecans and pop into the oven

° bake for 20 minutes and then drop the heat to 350°F for another 30 minutes.  

° let the pie cool on a rack for at least an hour before serving

° top with a good ice cream or whipped cream if you want something over the top