Tag Archives: localoptimisation

i only want a pencil

Posted on by

My pencil is blunt. And I know there are no pencil-sharpeners in this office. But I do know there’s a stationery store upstairs where I can get a new pencil. (I already have 7 blunt pencils in my bag, all provided by this stationery store. For some reason I always forget to sharpen them when I get home. So I have to treat pencils as a use-once resource.) The stationery store has gone. I ask, and it turns out that the stationery department were ‘let go’ last month.

I’m told the project office may have a secret stash of stationery, if I ask nicely. I have a contact there, so I ask her. “Ah – no, we have none; try Julia.” So I call Julia’s mobile. From wherever she is in the world, she tells me I need to see Pat. I look Pat up in the site directory and go to her end of the building. Yes, she has pencils. If you’re gasping at this rigmarole wait, it’s not over yet.

“Are you authorised to take our stationery?” I use the secret handshake and she’s happy. “Can’t be too careful these days. Not everyone is allowed access to stationery.” She walks over to a wall safe and keys in the numeric combination. From the safe she takes a set of keys. We walk over to an anonymous door, which she unlocks. We go in, and I get a pencil. “Devils!” she exclaims, “someone’s taken 8 lab-books since Thursday!” She locks everything again on the way out.

Total elapsed time from running out of lead: 30 minutes. Number of staff involved: 5. Total staff time: 60 minutes. Total cost of the exercise: around £80, plus the phonecall, plus disruption to 5 people’s flow. How much do pencils cost?

Update:
Here’s an evaporating cloud showing the conflict within the organisation:

pencils

the taxi business

Posted on by

I just had a project cancelled, because the business features it would have introduced are not sufficiently high priority right now. I’m fine with that approach. (It may be re-awakened in 6-12 months, if there’s enough money and interest.) Fortunately it was still in its infancy, so not much money has been spent on the wrong thing. Apparently. But as with everything else in this very strange organisation, when taken with the prevailing cost-accounting culture this decision may prove to be short-sighted. Let me explain…

Every project here is costed separately, and each is also judged separately for its revenue-earning ability. So infrastructure projects rarely occur; instead, the first project that needs to change the prevailing infrastructure / architecture gets to pay for that change. As one of the designers on my project put it: “The first person who needs to go somewhere buys a taxi; everyone else then gets a free ride.”

Now my project was going to be putting in place a whole raft of new infrastructure. In many respects it was seen by the techies as heralding a new dawn. And now it’s gone. And so has the taxi for all of those passengers who had planned to hitch a free ride in the next 12 months. Now they have to develop that infrastructure themselves. And pay for it. And instead of being co-ordinated by one team of designers, it will now be done in disjoint chunks, probably leading to overlap and conflict. So their costs will go up and their timescales will lengthen. The overall cost increase may well be greater than it would have cost me to buy the taxi.

So yet again, localised measures prove to be very expensive and wasteful when compared against more global measures; there has to be a better way… What would a lean manufacturing organisation do? Suppose we have a factory in which various products share the facilities; would each product carry its own production costs? What if a machine needed to be upgraded – would we charge that to the first product that needed the upgraded machine? I’m sure the Theory of Constraints muct have something to say here…

taking responsibility

Posted on by

Sometimes change can only come about when people take responsibility for problems, even when they didn’t necessarily cause the problem. Kay Pentecost tells a nice story about this in the XP group on Yahoo today, in which one group blames their problems on poor specifications received from another group.

When this happens, it seems to me to be another kind of local optimisation: someone tries to ensure that the problem lies at someone else’s door, and the result is that the overall process flow suffers. Both groups need to work together to improve communication, for the greater good…

local optimisation

Posted on by

For the last couple of days I’ve been studying the results of a local process improvement exercise. The exercise was run earlier this year, and had its own business case, complete with a financial justification. The plan was to reduce by 50% the number of defects that this department’s support team had to deal with, and the justification was that this would save around £30,000 per month in overheads. Now of course I believe that fixing defects is muda, so any reduction in this waste gets my full support. And indeed the improvement exercise was successful, as the bug statistics for recent months have shown.

So the company has saved a load of money, right? Well, certainly there are fewer bugs to fix in this one department. But no-one was laid off – instead, people now simply spend less time in support, and more time doing other work. This is a classic case of what TOC calls local optimisation: this department is now spending less of its own money, but as a result some others are probably spending more. And as I look around I find that the entire organisation – which is large – is incentivised in a similar way. Each department’s objective is to “save” costs by cross-charging its staff to other departments. But because they all ultimately work for the same company, this local accounting is obscuring the bigger picture. I’m convinced that end-to-end project costs are therefore significantly higher than they could be.

Could it be done differently? TOC says it can. The key measure of success in this company’s business is time to market. What if we could find a way for each department to somehow be measured on throughput? (and such that fixing defects is seen to reduce throughput)

I guess I’ve found a mission…