Having done focus groups for several years now in different countries and industries, as I reflect back on them, I think of certain aspects that have worked well and thought I'd put those down in this post:

• Get the right people: This is of course super important as you will otherwise get wrong feedback and could make costly decisions. To ensure that this is prevented, a screener becomes necessary. In this you should ask the questions that are necessary in order to maximize your chances of getting just the right target audience. Also send the screener to the research agency well in time - at least two weeks before so that they have enough time to recruit.
• Ensure that the moderator is an expert: This is equally important - the moderator needs to understand the goals of the research and should be a capable of asking probing questions to get to the bottom of certain issues. Additionally he shouldn't be opinionated or foist his opinion on the group. Recently we spent about two hours before the start of the focus group going through the entire discussion guide with the moderator - it helped quite a bit. At this time it is also useful to talk about overall goals, the business context etc - anything that will give the moderator more insight so that he can do his job better.
• Ask the right questions: This can be done through the development of a detailed discussion guide - this is what can really help you get the answers you seek so the more time spent on the discussion guide, the better the chances that you would have asked the right questions. This involves:
• Drill Down Questions: When developing the guide, anticipate the responses and ask follow up questions that help tease out more info. Common ways of doing this include questions like: If so, why/ If not, why not/ Explain why you said this/ Why don't you like this and so on.
• Ask the respondents to rank/ rate if appropriate: Often there is the misconception that since the sample sizes are small and so not representative of the audience, asking ranking/ rating questions is not useful. Nothing can be further from the truth - by asking respondents to rank/ rate we are forcing them as individual customers to think hard and make choices/ trade offs. This helps in trying to capture what the customer is really trying to say. Also once the rating/ ranking has been done, ask the customers why they ranked/ rated the way they did. This makes them defend what they have said. Recently when we did some focus groups in Japan, in addition to the discussion guide, we developed a separate feature description and ranking sheet which we gave to the attendees. This also proved quite useful.
• Ask respondents to describe/ write down features/ attributes/ benefits: This has the advantage of obtaining rich information from each respondent. Once this is done, you can ask each respondent to read out what they have written. This helps ensure that no individual dominates a group with his/ her opinions.
• Ask respondents to describe on a white board: Some rooms even have white boards so it helps to get respondents to go up there and demonstrate/ describe what they feel.
• Show pictures/ other stimulii: This also aids discussion. Also if you are discussing some concepts, the more real life you can make it, the better you can get respondents to state views/ opinions. Some years back while doing focus groups among internet shoppers in San Diego, we developed concepts with detailed descriptions as well as pictures. We were able to get respondents to understand exactly what we were talking about which proved to be quite useful.
• Localization: Make sure that the translation is accurate and accounts for /reflects contextual idiosyncracies. A good way to do it if for example you are doing a feature description is to have the description in both English and in the local language. This way you know what is being discussed, also the respondents (if they are bilingual) have two versions to look at before responding. TIP: Use the local office to help with this - more efficient that way.
• Be present at the focus group if possible: Sitting behind the one way glass, you get the opportunity to observe what the respondents are saying. You also then have the opportunity to ask follow up questions if you want to probe some more.
• Post FGD reports: Depending on agreement with the agency, you can get:
• Flash Report (Summary) : Within 3-4 days
• Full Report: Within 2 weeks (It is a good idea to ask for a formal presentation so that you have a Q&A opportunity as well) 

What about sample sizes? Focus groups are a form of qualitative research so clearly the number of people will be small (cost mainly). Thus it would be a leap of faith to assume that the entire population feels the same way. A good strategy is to conduct quantitative research based on insight/s generated at the focus group. In my experience that has been the primary benefit of focus groups - to generate insight based on drill down questioning of a small set of customers who  are part of your target audience.  

I hope this is useful.