9 ms·

The Tim Ferris way of testing ideas and how I did it.

82 points by dawie 17y ago

teej 17y ago
I've been itching to try this for some time. I would appreciate any war stories HNers could share.
- dawie 17y ago
  Like I said in my blog post, it worked pretty good for me. From here, I basically went on and created a Minimal Viable product.
- patio11 17y ago
  Email addresses don't covert on a 1 to 1 basis to sales. (If. Only.) After the autoresponder runs out (6 days in my case) they've only got only a moderately higher chance of conversion than my average trial user. (~5% vs. 2.8%) I really, really don't like the Put Up Fake AdWords Ad, See Who Gets Suckered minimum viable product. Aside from misleading customers and hurting the AdWords ecosystem (they don't actually get offered what they clicked to get offered, which Google will come down on you like a ton of bricks for if they detect, incidentally), it doesn't test your product. It tests your ability to write AdWords copy. I could get 14% conversion to stabbing toothpicks under your fingernails, but what does that tell you? The world has weird people in it, and I'm good at qualifying non-weird people out with my ad text and keyword selection? Whee? Similarly, if I delivered an AdWords campaign that was awful (low CTRs, low conversions), would that tell you the product is awful? Because I've done that before, too -- the first draft of my AdWords campaign for Bingo Card Creator was terrible. It took me almost a year before I figured out how to write AdWords decently. That was not an indication that the product was terrible (I sell it quite successfully), it was just an indication that I needed to adjust to a new way of copywriting. If it can't tell you if an idea is good, and you can't trust it if it says the idea is bad -- what is the point, again?
  - tocomment 17y ago
    Those are really good points. Is there another type of market testing you recommend?
  - sgk284 17y ago
    I think you've hit on some good points, but missed a big one. Running ad words drives traffic to a site that currently has no traffic, better yet that traffic has shown some interest in the product you're claiming to have. If some of those people try to buy (or express interest in) this non-existent product, that says something. You're not measuring how much traffic ad words brings in. You're using ad words as a traffic generator (ideally generating interested traffic), and of this traffic that comes to your site, you can determine if people are clicking through. For ~$100 you can determine if there is any interest at all. This can be powerful for people who really have no idea what the market will do. It's not perfect, but not useless.
    - patio11 17y ago
      This can be powerful for people who really have no idea what the market will do. These people should take their hands off the keyboard and not put it back on until they have spoken to at least one Real Person who has the problem that they are trying to solve. After they have done this, they will know if there is any interest at all. This produces actionable insights: if they tell you they have a problem, then at least some people have the problem. If they tell you that the industry's favorite way for solving the problem costs $1,000, that immediately establishes that there is a viable market for solving the problem. Compare this to the AdWords campaign: can you tell me what numbers you can get that say "This is a good idea -- do this" and what numbers you can get that say "This is not a good idea -- don't do this"? No -- subjectively "good" results don't imply purchasing intent, subjectively "bad" results don't imply lack of it. You'll get noise -- very precise noise, noise you can measure down to the click, but still noise.
      - sgk284 17y ago
        Suppose I came up with an idea that I absolutely knew was brilliant, let's say edible shoelaces. Unfortunately these edible shoelaces aren't cheap to produce. I'm a one man shop and have almost no money. I make a fully functional website to sell this product, only I don't have the product yet. When people browse the site and have gone through the pricing page, etc... and finally click "buy" I simply tell them that we're out of stock. Now I want to see if people are actually trying to buy this thing, so I start marketing but have a limited budget. I throw $1000 at ad words. If I get 100 people trying to buy my edible shoelaces at $20 each then I know that maybe it's worth investing in manufacturing the product. How is this not useful data? Instead of going all out and spending money on manufacturing and a website, you only pay for the website and hold off on the manufacturing expense until you're confident it'll pay off.
- jmathes 17y ago
  I worked at a software company where we should have run an experiment like this. There's an upgrade on user accounts that we charged a one-time $10 for, and people buy it all the time. We decided to try it as a subscription. We spent a month getting the subscription code ready, then deployed it to some % of users, only to discover that nobody wanted it. What we should have done was build the front-end only, get people to buy subscriptions, and then just grant them the one-time permanent version.
tocomment 17y ago
That seems hard to believe. I did the same testing strategy with my Job applier software (http://fastjobapplier.weebly.com/ http://fastjobapplier.weebly.com/) and first of all I only got around 30 ad clicks total, and no one filled out the form. (I was thoroughly disappointed.) Granted maybe I just suck at sales copy, or perhaps there just isn't demand out there for software that helps you apply for jobs. Still though, a 10%+ conversion rates seems too high. Do you find you're still getting a similar conversion rate now that you're actually selling the product?
- utnick 17y ago
  yea 10% seems pretty high, I think it might have to do with saying the 1st month is free and you can cancel any time. Therefore, the 10% rate isn't really a 10% sales rate but more of a 10% willing to try your demo. But 10% is still pretty good for that I think.
  - dawie 17y ago
    I agree that 1st month free and cancel anytime might have something to do with it. If I do a test like this again, I will probably leave that out. I will have these terms in my live product though...
    - teej 17y ago
      The core issue isn't that you offer a month free, the issue is that your form doesn't measure purchase intent. You have no idea if 1% or 10% of your leads will convert into sales.
  - dstorrs 17y ago
    Actually, I found it pretty believable that this percentage put their email in. I ran something similar...a SurveyMonkey questionnaire about the product, where the last question was "If you'd like to be notified when this launches, put your email here." 25-30% of those who finished the survey gave their email. This is ONLY email, of course, not actual signups. I'm expecting 90+% of those people will drop out once we actually email them.
fookyong 17y ago
There is an ocean of difference between 14% of visitors committing their email address to something, and that same figure actually pulling out their credit card. A more prudent assumption would be 10% of that 14% will actually pay for your product when it is launched. I learned this first hand. Your mileage may vary of course, but it's incredibly easy to drive traffic and stimulate interest, enough that people will give you their email address or whatever other personal information you want. The moment you ask them to pay though, you're in another world in terms of conversions :) This MVP-type google ad testing doesn't really translate that well into real world data like purchase conversions. I don't think that's the point anyway. It's good for seeing what people might be interested in, which is often something that startups aren't sure of :)
- dawie 17y ago
  I agree with you. Maybe a good idea would be to ask for a credit card, but not to charge it. I think this borders on being illegal though. I am still pretty happy with 10%, since this was my cut-off for going ahead and building the product.
  - ruchi 17y ago
    Per fookyang, the conversion is just 1.4%.
  - BobbyH 17y ago
    In his book, I believe that Tim says that it's not illegal if you don't store a user's credit card information when selling a product that isn't fully-baked. But I agree that it is a bit shady...
    - cschneid 17y ago
      "This product is currently in private beta. Enter your credit card details to guarantee your spot in the next phase of release. You won't be charged a cent until you get access" Then do the authorize.net CC storage thing where you don't even need to worry about securing private details (pre-auth, don't capture payment till you have a product to sell).
abalashov 17y ago
Actually, the figures here may be pessimistic. It takes a certain kind of person to submit their e-mail address when there is quite obviously no deliverable being explicitly promised to them for immediate purchase & download. If there were one, there might have been higher conversion.
spokey 17y ago
Here's another (cheaper, although possibly less realistic) approach to this: I recently did some market research using Amazon's Mechanical Turk, and added a lead collection form as a bonus. I put up a simple survey asking users about relevant background information and their experience with the problem my application is trying to solve. Included in this survey was a field to collect the user's email address. This field was very clearly marked as optional (it even appeared after the confirmation code that allowed turkers to complete the "HIT") and the label included something like "Your answers to this survey will help us to evolve the website to better meet your needs. If you'd like to be notified when the new site launches, please enter your email address here." along with an indication that users would receive exactly one email from us due to this form. I was pretty pleased with the results of the survey--I received a lot of actionable information very quickly and cheaply (at about $0.10 per response) and I was pleasantly surprised that a little more than 10% of the respondants entered what looks like a valid email address. I don't expect many of those 10% to convert (and lead generation wasn't the point of the excercise anyway) but I was very happy with the ROI on this survey. Seeing who clicks on ads (and at with what copy) and later "soft" converts may be a more realistic test, but the MTurk approach is an order of magnitude less expensive.
joubert 17y ago
I don't get it. So you "sell a product", including a 30 day trial, but once the user signs up there's no product?

82 points by dawie 17y ago

teej 17y ago

I've been itching to try this for some time. I would appreciate any war stories HNers could share.

dawie 17y ago
Like I said in my blog post, it worked pretty good for me. From here, I basically went on and created a Minimal Viable product.
patio11 17y ago
Email addresses don't covert on a 1 to 1 basis to sales. (If. Only.) After the autoresponder runs out (6 days in my case) they've only got only a moderately higher chance of conversion than my average trial user. (~5% vs. 2.8%) I really, really don't like the Put Up Fake AdWords Ad, See Who Gets Suckered minimum viable product. Aside from misleading customers and hurting the AdWords ecosystem (they don't actually get offered what they clicked to get offered, which Google will come down on you like a ton of bricks for if they detect, incidentally), it doesn't test your product. It tests your ability to write AdWords copy. I could get 14% conversion to stabbing toothpicks under your fingernails, but what does that tell you? The world has weird people in it, and I'm good at qualifying non-weird people out with my ad text and keyword selection? Whee? Similarly, if I delivered an AdWords campaign that was awful (low CTRs, low conversions), would that tell you the product is awful? Because I've done that before, too -- the first draft of my AdWords campaign for Bingo Card Creator was terrible. It took me almost a year before I figured out how to write AdWords decently. That was not an indication that the product was terrible (I sell it quite successfully), it was just an indication that I needed to adjust to a new way of copywriting. If it can't tell you if an idea is good, and you can't trust it if it says the idea is bad -- what is the point, again?
- tocomment 17y ago
  Those are really good points. Is there another type of market testing you recommend?
- sgk284 17y ago
  I think you've hit on some good points, but missed a big one. Running ad words drives traffic to a site that currently has no traffic, better yet that traffic has shown some interest in the product you're claiming to have. If some of those people try to buy (or express interest in) this non-existent product, that says something. You're not measuring how much traffic ad words brings in. You're using ad words as a traffic generator (ideally generating interested traffic), and of this traffic that comes to your site, you can determine if people are clicking through. For ~$100 you can determine if there is any interest at all. This can be powerful for people who really have no idea what the market will do. It's not perfect, but not useless.
  - patio11 17y ago
    This can be powerful for people who really have no idea what the market will do. These people should take their hands off the keyboard and not put it back on until they have spoken to at least one Real Person who has the problem that they are trying to solve. After they have done this, they will know if there is any interest at all. This produces actionable insights: if they tell you they have a problem, then at least some people have the problem. If they tell you that the industry's favorite way for solving the problem costs $1,000, that immediately establishes that there is a viable market for solving the problem. Compare this to the AdWords campaign: can you tell me what numbers you can get that say "This is a good idea -- do this" and what numbers you can get that say "This is not a good idea -- don't do this"? No -- subjectively "good" results don't imply purchasing intent, subjectively "bad" results don't imply lack of it. You'll get noise -- very precise noise, noise you can measure down to the click, but still noise.
    - sgk284 17y ago
      Suppose I came up with an idea that I absolutely knew was brilliant, let's say edible shoelaces. Unfortunately these edible shoelaces aren't cheap to produce. I'm a one man shop and have almost no money. I make a fully functional website to sell this product, only I don't have the product yet. When people browse the site and have gone through the pricing page, etc... and finally click "buy" I simply tell them that we're out of stock. Now I want to see if people are actually trying to buy this thing, so I start marketing but have a limited budget. I throw $1000 at ad words. If I get 100 people trying to buy my edible shoelaces at $20 each then I know that maybe it's worth investing in manufacturing the product. How is this not useful data? Instead of going all out and spending money on manufacturing and a website, you only pay for the website and hold off on the manufacturing expense until you're confident it'll pay off.
jmathes 17y ago
I worked at a software company where we should have run an experiment like this. There's an upgrade on user accounts that we charged a one-time $10 for, and people buy it all the time. We decided to try it as a subscription. We spent a month getting the subscription code ready, then deployed it to some % of users, only to discover that nobody wanted it. What we should have done was build the front-end only, get people to buy subscriptions, and then just grant them the one-time permanent version.

tocomment 17y ago

That seems hard to believe. I did the same testing strategy with my Job applier software (http://fastjobapplier.weebly.com/ http://fastjobapplier.weebly.com/) and first of all I only got around 30 ad clicks total, and no one filled out the form. (I was thoroughly disappointed.) Granted maybe I just suck at sales copy, or perhaps there just isn't demand out there for software that helps you apply for jobs. Still though, a 10%+ conversion rates seems too high. Do you find you're still getting a similar conversion rate now that you're actually selling the product?

utnick 17y ago
yea 10% seems pretty high, I think it might have to do with saying the 1st month is free and you can cancel any time. Therefore, the 10% rate isn't really a 10% sales rate but more of a 10% willing to try your demo. But 10% is still pretty good for that I think.
- dawie 17y ago
  I agree that 1st month free and cancel anytime might have something to do with it. If I do a test like this again, I will probably leave that out. I will have these terms in my live product though...
  - teej 17y ago
    The core issue isn't that you offer a month free, the issue is that your form doesn't measure purchase intent. You have no idea if 1% or 10% of your leads will convert into sales.
- dstorrs 17y ago
  Actually, I found it pretty believable that this percentage put their email in. I ran something similar...a SurveyMonkey questionnaire about the product, where the last question was "If you'd like to be notified when this launches, put your email here." 25-30% of those who finished the survey gave their email. This is ONLY email, of course, not actual signups. I'm expecting 90+% of those people will drop out once we actually email them.

fookyong 17y ago

There is an ocean of difference between 14% of visitors committing their email address to something, and that same figure actually pulling out their credit card. A more prudent assumption would be 10% of that 14% will actually pay for your product when it is launched. I learned this first hand. Your mileage may vary of course, but it's incredibly easy to drive traffic and stimulate interest, enough that people will give you their email address or whatever other personal information you want. The moment you ask them to pay though, you're in another world in terms of conversions :) This MVP-type google ad testing doesn't really translate that well into real world data like purchase conversions. I don't think that's the point anyway. It's good for seeing what people might be interested in, which is often something that startups aren't sure of :)

dawie 17y ago
I agree with you. Maybe a good idea would be to ask for a credit card, but not to charge it. I think this borders on being illegal though. I am still pretty happy with 10%, since this was my cut-off for going ahead and building the product.
- ruchi 17y ago
  Per fookyang, the conversion is just 1.4%.
- BobbyH 17y ago
  In his book, I believe that Tim says that it's not illegal if you don't store a user's credit card information when selling a product that isn't fully-baked. But I agree that it is a bit shady...
  - cschneid 17y ago
    "This product is currently in private beta. Enter your credit card details to guarantee your spot in the next phase of release. You won't be charged a cent until you get access" Then do the authorize.net CC storage thing where you don't even need to worry about securing private details (pre-auth, don't capture payment till you have a product to sell).

abalashov 17y ago

Actually, the figures here may be pessimistic. It takes a certain kind of person to submit their e-mail address when there is quite obviously no deliverable being explicitly promised to them for immediate purchase & download. If there were one, there might have been higher conversion.

spokey 17y ago

Here's another (cheaper, although possibly less realistic) approach to this: I recently did some market research using Amazon's Mechanical Turk, and added a lead collection form as a bonus. I put up a simple survey asking users about relevant background information and their experience with the problem my application is trying to solve. Included in this survey was a field to collect the user's email address. This field was very clearly marked as optional (it even appeared after the confirmation code that allowed turkers to complete the "HIT") and the label included something like "Your answers to this survey will help us to evolve the website to better meet your needs. If you'd like to be notified when the new site launches, please enter your email address here." along with an indication that users would receive exactly one email from us due to this form. I was pretty pleased with the results of the survey--I received a lot of actionable information very quickly and cheaply (at about $0.10 per response) and I was pleasantly surprised that a little more than 10% of the respondants entered what looks like a valid email address. I don't expect many of those 10% to convert (and lead generation wasn't the point of the excercise anyway) but I was very happy with the ROI on this survey. Seeing who clicks on ads (and at with what copy) and later "soft" converts may be a more realistic test, but the MTurk approach is an order of magnitude less expensive.

joubert 17y ago

I don't get it. So you "sell a product", including a 30 day trial, but once the user signs up there's no product?