HN Explorer

Ask YC: What resources do you use for your market research?

14 points by ryanspahn ↗ HN
I need data in regards to our market. A sampling of 1,000 people to various niche questions. I was wondering what creative ways others have used to gather data to present to angels and VCs? We are looking to do this on the cheap, as Google and libraries have failed us!

Things we have done or planning to do is...

1. We used a twitter search... summize.com, but Twitter's API only goes back 200 Tweets. It says total 2,000 tweets on topic, but only shows 200 :(

2. Ask friends who are pastors or teachers to ask a question here or there and do a head count.

14 comments

[ 0.34 ms ] story [ 51.5 ms ] thread
rms
You can do polls on Facebook for as low as 5 cents a vote and you can target based on facebook profile information.
LPTS
I look at medical journals with the demographic information we need. Then we ask people who could benefit from our tech and we ask them to try it and watch them use it and ask them to give us feedback.

Do you really care what 1,000 idiots on facebook (or 1,000 yahoos from anywhere on earth) with nothing to do but answer your shitty poll think?

Do you really want to work with a firm for whom a poll of a bunch of yahoos, a quarter of whom were statistically likely to be watching porn as they filled out your poll, was the thing that got you over?

Twitter is market research? There isn't a better way?

diego
For Twitter, try our tools:

http://twist.flaptor.com http://twittersearch.flaptor.com

And please let us know if you have any feature suggestions. We are trying to build a useful tool so we are all ears.

ryanspahn
yes i was using twist as i saw it on TechCrunch. I got to 200 out of 1220 tweets regarding my topic, where it said something like unable to show more then 200.

I assumed this is due to Twitter's API? If that is not the case can you code it to show all results on a topic from day one?

Though Im not sure what VCs or angels will think when I present my data being from Twitter. But to me its a cheap and solid way of gleaming demographics on topics; albeit the demographic is skewed for now.

diego
Thanks for the feedback. The 200 limit is ours, we could (and should) increase it. We don't have all the results from day one (we'd have to ask Twitter for the full archive), but we could easily display up to 1000 results without a significant performance hit.
ojbyrne
A friend of mine used an adsense campaign with ads pointing to a poll. Seemed to work well.
amobilebiz
A lot depends on the market you are trying to reach. While the concept for market research is going to be the same, the methods used to gather that research need to be tailored in order to achieve the best results. Anyone can poll 1000 people and ask them what they think, but if 900 of those people do not understand your market or would not find a use for your product/solution the chances of gathering useful data are minimal. Can you provide more specifics as to what your target demographic is or what your product is?
johnm
Polls are generally horrible. Self-selection bias, self-deception (i.e., what one perceives of one's self and what one actually does are almost always different and often wildly so), and self-disclosure bias (i.e., what one wants (others) to believe about one's self).

The entire analyst industry is basically built to do this. Of course, they tend to be ridiculously biased because they are getting paid so much for their opinions. Finding friends who work at places that already have the reports is one way to get them. :-)

In terms of actually useful data... Look for information about the users' and customers' actual behavior -- i.e., where the user/customer actually had to do something real like visit repeatedly, pay money, recommend you to other people, etc. As a number of folks mentioned at SUS, if you don't have information directly relevant to your play then look for various proxies and triangulate.

dkokelley
Polls don't have to be filled with questions about what you would do, leaving room for inaccurate self-prediction. The questions could be "have you in the past month done x" which tend to be a bit closer to reality.

Also, certain techniques (randomizing the order of the poll questions for each visitor) and statistical functions can be used to get more accurate information from those polled.

But you're right, if done improperly you can get wildly inaccurate data (just look at past or current presidential election polls)

johnm
True, you can setup the polls and question to reduce the problems but they are all still there since you still asking the user rather than measuring the actual behavior.

I.e., polls are a form of proxy for what you really want to learn but can't get directly. Your example of voting polls is a great example of this... Until the actual election, we can't know and so polling is about as good as we can get. (But I won't even bother to mention all of the games that are played with the polling to manipulate the outcomes. :-( )

dkokelley
Depending on what you're doing, you could always try government or census records. For example, I know how many people are in my city, how much they make, how old they are, etc., and for my school I know how many credits the students are taking, the age, gender, and previous education of the population.
DenisM
I was advised to use "omnibus survey" on the theory that it's cheaper that normal full-blow survey. Never gotten to use it, so just passing it on.