What is a Spike in Scrum

This question comes up time and time again and the Spike is often confused with the Tracer bullet. Adam Sroka posted a great explanation of the difference between the two on the Yahoo Scrumdevelopment group . So I am quoting Adam verbatim here - thanks Adam.

"The Pragmatic Programmers described something called a "Tracer Bullet" which:

1) Is an experimental solution that cuts through all the "layers" of
the architecture.

2) Is not necessarily time-boxed.

3) Is not intended to be thrown away.

Eric Evans talks about "Thin, vertical slices," which are the same as
Tracer Bullets.

A Spike Solution:

1) Is an experimental solution that cuts through all the "layers."

2) Is necessarily time-boxed.

3) Is always intended to be thrown away.

The reason for the distinction, IMO, is that the "tracer bullet" or
"thin, vertical slice" model is how XP teams normally work. A Spike is
an exceptional way of working when we feel we don't have enough
information to give the customer realistic expectations. The goal of
the Spike is to establish those expectations."

Spikes are a really good way for teams to figure out stuff that they don't know and need to know in order to understand the complexity so that it can be properly estimated, or quoted on or simply to find out if something is technically possible or not.

Comments

[David]

Then what is the difference with a PoC?

[Darnell Brownen]

My company is developing a new, complex software platform. Much of what we are designing is so new, numerous spikes are required to investigate issues. The Product Managers are asking an important question: should we enter every spike into our backlog and/or a given sprint?

In the interest of collaboration, we want to share ‘brain storming’ with team members, but not be a distraction. The investigation requires time from Product Manager, Software Architects and Engineering. But, is there value is tracking spike? How do we balance the need to share investigations without being a distraction to sprint team members?

[Dan]

@Darnell,

First, a question: who is doing the spiking? Is it the same pool of developers that work on features (i.e. the same team?). Or is it a different group of people?

If it's the same people, then you should absolutely track spikes on the same board that the feature developers use. Since developers will split their time between features and investigation, you need to see it all at once.

If it's a different group of people, then you should absolutely not track spikes on the same board that the feature developers use. Since feature developers are never going to spend any time on the spikes, it just adds noise to their daily stand-up. Though in this case, the people who are doing the spikes should probably track their work on their own board.

(I know that you asked about the backlog, and I mentioned the status board, but really the status board is just a window into the backlog).

There's nothing wrong with multiple teams, each with their own backlog, scrum board, and stand-up meetings.