Context switching – can we really multitask?

Reading time

3 min

As we have seen, excessive Work in Progress slows the team and costs money for no return. Less obviously it also will tend to make the overall cost of the work higher. Parallelism (working on multiple items in parallel) is fundamentally inefficient because of context switching.

Many people like to believe that they can “multitask” effectively. By this they mean that they can work on several items in parallel, switching seamlessly between them. However, research suggests that switching between work is not efficient. There is a cost in moving from one piece of work to another. The “context” of the task needs to be recalled, and you need to get back to full speed on the work.

Effect of distraction

There is increasing realisation about the impact of distraction on work which requires concentration. This can be recovery time from a simple distraction when deep in a piece of work. Current research suggests that it typically takes fifteen minutes to return to the same focus level after each distraction. We also see a greater appreciation of the importance of avoiding distractions. Legislation now precludes mobile phone calls when driving a car.

It is hard to give absolute numbers about how much impact there is from parallelism. Widely quoted figures from Weinberg’s “Quality Software Management” suggest that:

Each extra work item added consumes about 20% of capacity.
If this moves to three items, we lose 40% in switching and this leaves us with only 20% of our time on each item.
By working on two items in parallel, we lose 20% of our time switching and end up with about 40% of our time on each.

The effective progress on each work item drops sharply.

How big is the effect on value delivered?

Imagine we have a team working on a piece of work which they aim to deliver in four weeks. Two more items arrive which will also take four weeks each. Should they remain on the first item or work on all three in parallel?

If the team plan the work sequentially they would deliver the first item after four weeks, the second after eight and the third after twelve.

But if they work in parallel, they will deliver all of the items together. Weinberg suggests they will lose 40% of their time in context switching. This means they will deliver everything after twenty weeks!

The contrast is stark. Where sequential work would start realising value after four weeks, parallel work gives none until twenty weeks. This is clearly the down side of high Work in Progress!

Good Practices

As an Agile leader, you can mitigate the problem by carefully watching for teams working in parallel.

Encourage teams to finish work before moving on to a new item. Measure and track Work in Progress.

I find it useful to focus teams on “Look right” thinking. This is named after the structure of Agile Boards where work starts at the left and finishes at the right. Always collaborate on closing items before you start new ones.