Whilst not a generally accepted scrum principle (excellent article from Mike Cohn http://www.mountaingoatsoftware.com/blog/gasping-about-the-product-backlog) in my experience as a Scrum Master i have found that Backlog Grooming (or Backlog refinement as it is often called) is a hugely important exercise for a Scrum team.
As Mike Cohn states in the post above the backlog grooming meeting isn’t part of the core Scrum rules nor is it a Generally Accepted Scrum Practice (GASP) but more and more teams are adopting this.
I’ve found it beneficial to have two types of grooming. The first is for the very next sprint typically with the Scrum Master, Product Owner and Testers/QA. This is to ensure the backlog is ready for the next sprint, all the stories are well formed, valid and that the acceptance criteria is still accurate. Whilst it isn’t to replace or pre-empt Sprint Planning I have found that most Product Owners know what they are thinking of for the next sprint so it helps just to have a check that these stories are acceptable before Sprint Planning.
The second type of Backlog Grooming I perform is with the entire team. These are usually to discuss and estimate stories on the backlog that are either new, un-estimated or have changed scope/shape since they were previously estimated. Typically these are for stories within the next 1-4 months ahead. I’ve found that not only does this give the team members a “heads up” of what work is in the pipeline but enables them to take a collective ownership of the story by discussing it with the PO and other team members, allows them to estimate and break the story down in required. This session is also optional and have found that where the backlog requires no effort from the entire team then its best not to have it to avoid distracting the team unnecessarily (plus we know that devs hate meetings!)
For both of these grooming sessions I spend no more than one hour in each per week. This means 4 hours a sprint (two weeks) of Backlog Grooming which i have found to be more than enough.