Performance Versus Cost

Dear Gemba Coach

I am the continuous improvement director of a large hospital. We invested two years ago in a central machine to perform all the blood tests.  Unfortunately, we have many technical issues with the equipment and we don’t know how to get the manufacturer to solve the problems. The wards complain about our lack of responsiveness as well, and I don’t know where to start. What would you suggest?

Ouch – It sounds like you’re stuck, and I’m not sure I’ve got anything helpful to say. You’re experiencing a problem common in industry: concentration of means makes accounting sense in terms of cost per unit, but not necessarily overall performance sense it terms of delivering at the lowest real cost. Complex centralized equipment has three generic problems which I’ve never seen resolved satisfactorily:

1. Service: By centralizing one operation, you create a “monument”: all flows now have to pass through this massive toll gate. It’s like creating a central border crossing checkpoint where all roads have to converge. Not surprisingly, this creates long queues. On top of which the monument is rarely flexible so it will batch work. Which means that one specific test from one specific ward might be stuck far in the queue and wait a long time.
2. Cost: Although cost by unit is clearly lower, the cost of use of the machine is generally higher than expected. Complex equipment has complex breakdowns, and requires sophisticated problem solving done by top dollar engineers. To evaluate the cost performance of a centralized investment, one needs to include all the exceptional babysitting costs of keeping the machine up and running. If you’ve got to fly engineers in from Frankfurt to check the software, you might be in for a pretty penny.
3. Quality: On average the quality performance of a central machine tends to be better than of multiple spread-out operations – and yet, it’s also stable. What this means is that with smaller delocalized operations one can hope to improve quality if we try, as it’s often a case of training and attention. With the machine, however, the quality parameters tend to be mysterious, even to the machine manufacturer. Understanding how to go beyond 90% quality with a large complex equipment means doing serious capability analysis that often requires technical skills beyond what we currently have: although we know how to improve blood analyses (it’s what we do), we rarely know how to improve the automation of blood analyses (what the machine does).

Balancing Act

I’m not arguing against concentration of means – this is still a sensible thing to do in many cases. I’m ranting about the over-enthusiastic investment in such solutions (senior execs love nothing more than signing big checks on big toys). In practice, large concentrated investments get the job done in conditions of massive overcapacity – to cover up for the previous issues, which also means we’ve spent more money upfront than necessary.

This might sound very theoretical and far from your gemba problem now that we’ve got the machine, but please bear with me. This is an issue you’re going to have to master fully because the few countermeasures I know of will be unpalatable to management as they will seem to add more cost to an investment that already broke the bank. Before you go charging in about fixing the problem, you need to frame the debate in the right way to have a chance of winning the upcoming political battle (no executive takes kindly to be shown he’s made a bad mistake and has egg all over his face).

The common frame of thinking for industrial decisions is cost. The reasoning is simple – simplistic in fact – control and reduce cost per unit, and so improve your profitability. “Figures are figures,” they say, so sign with the supplier proposing the lowest price and pocket the difference. The trouble is that we also need to perform for customers. If the lowest price supplier has quality issues, this will appear in our processes and products – as extra costs. We need to organize entry quality checking (cost!), we need to rework the bad parts we make because of their components (cost!) we need to check our own products more carefully in case a part has gone through (cost!), doing all of this distracts us from producing as planned (cost!), and if we don’t do any of this, our customers will walk away and we lose sales. Although accountants have ROI calculations to evaluate an investment, they seldom know how to discount the R in terms of the real costs incurred in making good products with so-so components. Still, most managers hold on to the magical belief that if you systematically go for cheaper you’ll get better. And that’s just with quality. If the supplier has delivery issues as well, double the cost.

Lean has taught us that cost management (as opposed to cost killing) requires a better understanding of performance. Performance is traditionally expressed in terms of quality, cost and delivery (lead-time), to which I’d add safety and morale, but for the sake of the bean counting argument, let’s stick with quality, cost and lead-time to start with. Performance is not as satisfying to accountants as just cost because it needs an element of human judgment (God forbid managers actually would have to know what they’re doing!). The overall estimate of performance cannot be estimated by a calculus, it has to be a personal estimate of the balance between quality performance, cost performance, and lead-time performance.

Frame the Debate

Toyota’s standard is to run its assembly plants in two shifts as opposed to the three shifts in most other OEMs. The downtime between the two shifts allows Toyota engineers to (1) work on the equipment to keep it in tip-top shape and (2) do overtime if necessary to stick to the sacred production plan. When Toyota’s French plant started operating in three shifts, all over the world, lean aficionados started rolling their eyes and hollering that Toyota was losing its way. That’s silly. Let’s face it, running a third shift radically reduces the cost per car – if we can perform as well with three shifts as with two. Toyota would have been daft not to try it. And, as expected, the French plant worked hard, became the most productive plant of the lot (no surprise), but never quite performed at Toyota standard. It never was a bad plant, but, the French engineers argued, the site was never designed for the volume of three shifts, so of course they had a bit more inventory and of course they had more frequent breakdowns and so on.

After ten years of running the French experiment (back to two shifts because of volume reductions), (1) Toyota has not generalized three shifts to its other facilities and (2) the French plant kept receiving visits from senior Japanese managers asking them: show me how this three shift system works. Convince me. The point of the story is that the overall estimate of performance between quality, cost, and lead-time is just that: an estimate, and as such subject to debate. The second point of the story is that it’s key for senior execs to have this debate constantly amongst themselves because this is precisely how they’ll form a common vision of how to run the company profitably over the long term.

This is precisely the debate you must bring to your management boardroom if you want to have a chance of solving this problem. To transform others, first transform yourself and try to master the simple technique of evaluating decisions in terms of performance. This means sketching the following table at the back of an envelope:

The point of this exercise is NOT to make the best decision – outcomes are, after all, uncertain, BUT to have a sensible way of discussing this uncertainty within the management committee without being fixated on cost only.

Back to your gemba. You’re going to have to change the focus from the machine to the wards: what is the delivery contract you have with the wards? What do you have to do to honor it, no matter what?

You’ll probably find that not all blood test requests are the same. An emergency sample because the patient is in the operating theatre and, while running the pre-op checklist, the surgeon suddenly has a massive doubt, is not the same issue as a patience being monitored during the recovery phase. The first step is understanding customer usage and customer requirements, and stop trying to fit it all within the machine’s constraints.

Secondly, and this is not going to go down well if the machine costs a lot to purchase and maintain, you’re going to have to set up parallel operations to actually perform to contract. Usually this means handmade analysis of urgent requests. This might seem costly, but is necessary not just to fix your problem but to understand it better.

Looking at performance will lead-you to distinguish cases and progressively to learn how to better balance machine use and human use. As you take away all the extreme cases from the machine’s workload, you will also be able to narrow down the machine issues to fewer, more machine specific problems, which can now be focused on and resolved.

To answer your question directly, first specify your performance contracts with your customers in order to understand better their real demands. One usually finds that an “average” performance measure does nothing for customers as they have different expectations in different cases, which is precisely what we need to understand. Second, forget the cost fixation for a second and put in the local means necessary to fulfill customer contracts. Third, optimize the overall resource to reduce the need for the extra costs, by focusing on the real problems, which you have now broken down in manageable chunks.  I realize my answer might not help your right there and then, because once the investment has been paid for in full, most deciders are busy justifying the investment rather than solving the problem. But having misplaced my magic wand, I don’t know how else to go about it. Good luck!