Here are some quick tips to keep in mind about forecasting questions:
- Can be binary (Yes/No) or multiple
Must have objective and verifiable answers (so that
we can determine forecasting accuracy)
Good rule of thumb: do not start a question with
“should", or anything that elicits a subjective
Information should be readily available via research
of materials or interactions with people
Ask enough questions, but don't overwhelm. Many
people ask how many questions there should be to start. There is no exact
answer, but we have seen that starting with about 3-4 business questions
and 2 culturally-relevant "just for fun" questions is a good
start to get people intrigued and forecasting. Ideally, you want at
least 15-20 forecasters per question for valid results.
Publish new questions on an ongoing basis (weekly or
biweekly). After the launch, you should continue to have about
one new question each week or biweekly to keep people coming to the
Vary the question timeframes, have a good
"mix". Just as important as the number of questions is
the question timeframe. For greater participation, you must shoot for a
mix of short-term (1-3 months) and longer-term (4-8 months) questions.
Short timeframes are important as people want to see how they forecasted
- this gets them excited and drives more participation. If you're
finding that your business questions are long-term, figure out if there
are shorter timeframe milestones you can ask about and create a cluster
of questions that can guide predictions around a more long term event. As
a last resort, include short-term "fun" or more tangential
business questions (e.g. Who will win the X sporting event? What will be
the highest price of oil from now through the end of the month?).
Go back to the Support site.