CB Gravity vs. Popularity: Doing the Math

12 replies
Hello folks,

Just some thoughts on CB Gravity, Popularity and how to pick CB products. I evaluated only vendors that have a Spotlight page so that I could get the number of subscribers. I found the following:

1) The correlation (see Pearson's correlation coefficient) between Gravity and the number of Subscribers is 0.64290, which is a fairly strong correlation. When discussing Gravity CB (and many 3rd party articles about it, including this forum) always talk about sales, timing, etc. But ... a correlation of 0.64 between Gravity and the number of subscribers indicates that Gravity is, on average, going to be high whenever the number of subscribers is high (and the reverse, of course).

2) The correlation between Gravity and Popularity is 0.81705. This is a very strong correlation. According to CB, both Gravity and Popularity use some really complicated formulas that take into consideration the number of sales, affiliates, time and whatever. But ... however complicated their formula are, the one thing that they cannot do is hide their tracks! A correlation of 0.8+ means that Popularity and Subscribers move almost in lockstep. As we saw above, Popularity and Gravity also move together, although the correlation is lower because sales count for more in the Gravity computation.

3) Gravity and Avg. Sale are almost uncorrelated (-0.15). All this means is that expensive products tend to have somewhat lower Gravity than cheaper ones, but the correlation is weak.

4) The correlation between the Number of Subscribers and Popularity is 0.52, which is moderately strong.

5) The correlation between the number of subscribers and the average sale is almost zero (-0.05). This is a bit surprising, in that it tells us that both cheap and expensive products are able to gather affiliates.

6) The correlation between Popularity and the average sale price is low and negative (-0.20). Again, all this tells us is that affiliates in general have some aversion to promoting very expensive products, probably because they think that these products will be difficult to sell, but the correlation is weak.


How to use this information
----------------------------------
Basically, we are looking at discrepancies between the actual and expected figures.

For this analysis I looked at the Cooking, Food & Wine section in CB (not my niche, I am an omnivorous geek partial to real Spanish food). Let's compare a few products. The first three are in the Paleo diet BS fad category (no, we did not evolve to eat this type of food; we evolved long legs and distance-running traits to run our prey down and be the first to scavenge a carcass. Try that ...):


Product #1: Paleo Recipe Book - Brand New Paleo Cookbook
Gravity: 128.43
Subscribers: ??? (No Vendor Spotlight, so we need to estimate, see below)
Avg.Sale: $17.5
Popularity Ranking In Category: 1
Gravity has been steady for the last 3 months around 130 or so.
% Referred hovers around 20% (Bad! This means this vendor is generating 75% of their sales from non-affiliate sales).
Given the correlation between Popularity (Rank=1), Gravity (128) and # of Subscribers our estimate for this
product needs to be 250+ affiliates.


Product #2: Paleo Cookbooks - Complete Paleo Recipe Guide To Healthy Eating
Gravity: 40.55
Subscribers: 135
Avg.Sale: $28.85
Popularity Ranking In Category: 2
Gravity has been steady for the last 3 months around 40 or so.
% Referred at 75% and growing (Good! This means the vendor generates 75% of their sales through affiliates).

Product #3: Caveman Feast: 210+ Paleo Recipes From Civilized Caveman Cooking
Gravity: 22.61
Subscribers: ??? (No Vendor Spotlight, so we need to estimate, see below)
Avg.Sale: $10.25
Popularity Ranking In Category: 4
Gravity has been hopping around between 20 and 30 for the last 3 months, and it is decreasing.
% Referred hovers around 25% (Bad! This means this vendor is generating 75% of their sales from non-affiliate sales).
Given the correlation between Popularity (Rank=4), Gravity (22.6) and # of Subscribers our estimate for this
product needs to be 50+ affiliates.


Conclusion
If you are going to jump into the Paleo Diet fad the vendor #2 is likely going to generate more and better sales for you.
The first vendor is hyped-up to the max, has a ton of affiliates (and therefore a lot of competition) but *most* of their sales
are not coming from affiliates (why?). The third vendor has the same problem in that 75% of their sales are not coming through affiliates, and in fact looks like it has been losing affiliates.
Vendor #2 looks much better, in that their popularity ranking coincides with their gravity, they have about the right number of affiliates and the majority of their sales is actually coming from affiliates.


Product #4: Guilt Free Desserts: Gluten Free Diabetic Safe Desserts
Gravity: 19.81
Subscribers: 63
Avg.Sale: $7.37
Popularity Ranking In Category: 3
Gravity has been steady for the last 3 months between 20 and 25.
% Referred at 70% but coming down a bit lately (Good! This means the vendor generates about 70% of their sales through affiliates, but the decreasing trend might mean the vendor is also beginnig to use other channels).

Product #5: As Seen On Tv. Make Your Favorite Restaurant Dishes At Home!
Gravity: 24.75
Subscribers: 326
Avg.Sale: $14.73
Popularity Ranking In Category: 5
Gravity has been steady for the last 3 months between 25 and 30.
% Referred at about 80%, which is good, but there was a big drop to about 55% in Frebruary and then it went back up. This is worrisome because it means that this vendor had a big increase in sales from non-affiliate channels. This was probably their own promotion, but it means you are competing against your own vendor.


Conclusion
I would pick vendor #4. Steady gravity, steady % Referrals, the relationship between the number of affiliates (63) and their Gravity (24.75) looks like the Gravity is coming from a lot of sales from a lot of affiliates, rather than just one or two super-affiliates. What I mean by this is that Vendor #4 seems to have a positive discrepancy between the number of affiliates and their Gravity, and this works in your favor.
Vendor #5 also has a discrepancy between the Gravity and the # of affiliates, and a big one, but not in your favor. Given the correlation that we found between the number of affiliates and the Gravity we would not expect to see 326 affiliates for this product! Between this and the drop in % Referred (xtra-hype-promo anyone) I would stay away from this one.




Product #6: Gluten Free Low Glycemic Cookbook For Diabetics & Allergy Sufferers
Gravity: 6.27
Subscribers: 63
Avg.Sale: $7.37
Popularity Ranking In Category: 9
Gravity has jumping around a bit for the last 3 months between 4 and 9.
% Referred is almost 100%. This is excellent, as this vendor makes money exclusively through affilate sales.


Product #7: The Home Winemaker's Inner Circle
Gravity: 1.87
Subscribers: 55
Avg.Sale: $114.63
Popularity Ranking In Category: 33
Gravity has jumping around for the last 3 months between 1.7 and 4, with a decreasing trend.
% Referred is barely 25% with a very large drop from 75% three months ago (Bad!)


Conclusion
I would pick vendor #6 over vendor #7. Both are relatively low-volume sellers, but product #6 has a discrepancy between the number of subscribers (63) and their Gravity that indicates multiple sales by many affiliates. Looks like a slow-but-steady seller.

Product #7 got problems. First, there is a big discrepancy between the number of affiliates (55) and the Gravity (1.87, which is very low). Given the correlation that we saw above this product has many more affiliates than we should expect (and the discrepancy is not in our favor). The jump down from 75% in the number of % Referred indicates a big promotion by the vendor (perhaps because their affiliate sales were very slow). The 25% Referred is very, very low.


Comments, anyone?
#gravity #math #popularity
  • Profile picture of the author Alexa Smith
    Banned
    Originally Posted by mekdroid View Post

    Comments, anyone?
    Interesting.

    I'll read in more detail when I have more time, and maybe reply again. But if you'll excuse "quick, initial comments" ...

    Originally Posted by mekdroid View Post

    I evaluated only vendors that have a Spotlight page so that I could get the number of subscribers.
    Hmmmm. Well, I sort of see why, but they can have very varying numbers of additional affiliates who aren't subscribers.

    Originally Posted by mekdroid View Post

    According to CB, both Gravity and Popularity use some really complicated formulas that take into consideration the number of sales, affiliates, time and whatever.
    Ah well ... not exactly. The inputs into the Popularity algorithm are actually undisclosed. The inputs into the Gravity algorithm are fully and accurately disclosed, and there's no mystery at all about exactly how it's calculated.

    I regard it as more or less factual (based partly on conversations with ClickBank staff) that the Popularity algorithm includes a component of "the number of copies sold" and I know with certainty that the Gravity algorithm doesn't (hence so many single-figure Gravity products consistently outselling so many three-figure Gravity products, etc. etc. - this is "history" rather than "news", of course).

    I take your point about the correlation between them among vendors with Spotlight pages.

    "% referred" honestly meant close-to-nothing (because some vendors promote as their own affiliates, rather than just "as the vendor", and some vendors promote as multiple different affiliates, presumably in a (successful) attempt to increase their products' Gravity figures. So you can call me a skepchick regarding (m)any interpretations based on that. :p

    Again, I've looked only quickly/superficially, and your interesting post obviously "deserves better"; please excuse me.
    {{ DiscussionBoard.errors[7970263].message }}
    • Profile picture of the author mekdroid
      Hi Alexa,

      Actually, the reason why I chose to use vendors with a Spotlight is simply so that I can see the number of subscribers. That way I can calculate the correlation factors. Once I have the correlation factors then I can use them to estimate the number of subscribers for vendors without Spotlights.

      Re. the Gravity and Popularity formulas, I don't know how they are calculated, either, but they just can't help leaving a footprint! Any time you see a positive correlation like 0.8+ you know that the two series are moving in lockstep ... so whatever it is that they are doing to calculate Popularity, the number of subscribers accounts for the vast majority of the variance.

      That is all I need to find the discrepancies ...
      Signature

      Temporary occupation of some valuable piece of technical real-estate, followed by a negotiated retreat with full coffers

      {{ DiscussionBoard.errors[7970290].message }}
  • Profile picture of the author Scotty Bee
    Interesting, the only discrepency I might put in is this... If a vendor is doing all the sales and has no affiliates it may not be "bad". It may be a new product that the vendor has thrown a lot af paid traffic towards and no affiliates have come on board yet... maybe.. just by 2 cents worth
    Scotty
    Signature
    {{ DiscussionBoard.errors[7970272].message }}
    • Profile picture of the author mekdroid
      Hi Scotty,

      Yes, if the vendor has very few affiliates then the % Referred is not meaningful. The problem is when I see a vendor with a truckload of affiliates (hundreds) and a miserable 25% Referred.
      Signature

      Temporary occupation of some valuable piece of technical real-estate, followed by a negotiated retreat with full coffers

      {{ DiscussionBoard.errors[7970297].message }}
      • Profile picture of the author Alexa Smith
        Banned
        Originally Posted by mekdroid View Post

        The problem is when I see a vendor with a truckload of affiliates (hundreds) and a miserable 25% Referred.
        Depending on the niche, some/many/most of those can just be people buying one copy for their own use/assessment/whatever through their own affiliate links. It doesn't mean any of them is actually promoting the product. But they all put one point on the Gravity score and count as one affiliate, even though they're only really customers. (Which is of course exactly how so many IM/MMO products have such huge Gravity figures without many people actually promoting them at all :p ).
        {{ DiscussionBoard.errors[7970305].message }}
        • Profile picture of the author mekdroid
          LOL!!! Yes, of course. People buying through their own affiliate links will skew my calculations. The problem is that we really have very little data to go on (and it is ratty data at that ...).
          Signature

          Temporary occupation of some valuable piece of technical real-estate, followed by a negotiated retreat with full coffers

          {{ DiscussionBoard.errors[7970321].message }}
          • Profile picture of the author Alexa Smith
            Banned
            Originally Posted by mekdroid View Post

            People buying through their own affiliate links will skew my calculations.
            I'm afraid so, yes.

            I strongly suspect that this doesn't happen terribly often with hydroponic gardening products (apart from in the case of "potential affiliates" unable to get a review copy and therefore buying one to take a look with a view to possible promotion, which I sometimes do, myself - albeit not necessarily specifically in the hydroponic gardening niche, you understand!). But with IM-related and MMO-related products, of course, as the saying goes: "it only happens all the time!". :p
            {{ DiscussionBoard.errors[7970329].message }}
            • Profile picture of the author mekdroid
              Basically, what I am trying to do is to come up with a reasonable methodology for picking the products. I have been reading your posts (which are excellent!) and you have mentioned a lot of the important things:

              1) NO LEAKS
              2) NO OPT-INS
              2) Find quality products that you would use yourself
              3) No high-gravity products
              4) Low refund rates

              ... etc ... so I am putting everything together and then trying to come up with some metrics.

              In that respect, the high correlation between popularity and subscribers gives me a handle for the analysis ...
              Signature

              Temporary occupation of some valuable piece of technical real-estate, followed by a negotiated retreat with full coffers

              {{ DiscussionBoard.errors[7970374].message }}
  • Profile picture of the author SunilTanna
    Your analysis is flawed for at least 5 reasons.

    1. The number of subscribers to the spotlight is not equal to the number of affiliates.

    2. % referred tells you % referred, but nothing about the volume. For example, your vendor 3 could actually be doing more volume directly, and thru affiliates than your vendor 2. If that were the case your interpretation would be wildly inaccurate.

    For example if vendor 2 were selling $25 direct per week and $75 thru affiliates, but vendor 3 was doing $750 direct, and $250 thru affiliates. Vendor 3 doesn't look so crappy now?

    3. Lots of vendors are their own affiliate, either for tracking reasons or in a bid to increase gravity. It's trivial for a vendor to make every sale, or many sales, count as referred, when in reality only a small % might be referred by outside affiliates.

    4. Given 99% of affiliates are pretty much useless, 1 or 2 good affiliates can completely swing the % referred. If you've got a vendor with 100 crappy affiliates, the % referred will be low maybe 10%. If the vendor recruits just 1 good affiliate, who is equally good at selling as the vendor, suddenly the % referred would shoot up to 55%.

    5. The correlation between popularity and gravity probably doesn't tell you what you think it does.

    The most likely reason they are highly correlated is that they mostly measure the same underlying factors, and mostly weight these factors in highly similar ways. As gravity does not include sales volume, if popularity is highly correlated with gravity, it suggest that variability in sales volume can only account for a small percentage of the variability in popularity. In other words, a vendor massively increased their sales volume, would only see a small increase in popularity. In contrast, a vendor who signifcantly increased their gravity (even if sales would was constant or went down) would expect to see an increase in popularity.
    Signature
    ClickBank Vendor?
    - Protect Your Thank You Pages & Downloads
    - Give Your Affiliates Multiple Landing Pages (Video Demo)
    - Killer Graphics for Your Site
    SPECIAL WSO PRICES FOR WARRIORS + GET THE "CLICKBANK DISCOUNT" TOO!
    {{ DiscussionBoard.errors[7970515].message }}
    • Profile picture of the author mekdroid
      Yes, Anil, but:

      1) we do not have access to the number of affiliates, all we have is the number of subscribers, and then only for the vendors with a Spotlight. It seems sensible to use the number of subscribers as a proxy for the number of affiliates, particularly since the correlation between subscribers and popularity is so strong ...

      2) True, % Referred does not tell us the volume, but nothing else tells us the volume, either. Can't use data that I don't have!

      So ... between making my decisions based on no data and using whatever data I do have (no matter how poor) it seems logical to take the second option.

      The worst that can happen is that I will go and exclude people with high gravity, a ton of subscribers and very low % Referred ... which is probably not that bad an idea in the first place!
      Signature

      Temporary occupation of some valuable piece of technical real-estate, followed by a negotiated retreat with full coffers

      {{ DiscussionBoard.errors[7970555].message }}
      • Profile picture of the author SunilTanna
        Yes you can't rely on data you don't have. And, as you admit you have basically no data about the volume.

        (Likewise the correlation between popularity and subscribers could simply be that subscribers is included in popularity. There is no reason to think that they are independently measuring something else, in fact they can't be, because subscribers is it's own measure).

        But that doesn't mean making up data, which is what you are effectively doing, is the way to go.

        When you make up data, you may or may or may not get the right answer (in this case the best vendor(s)). If you do get the best vendor(s) it will be by chance. But the worst thing is you will never know if you have the right/best vendor(s), because you have no real data underlying your decision.

        Since cb does not provide the data (and I suppose even if they were to,, it would be unclear if you could extrapolate from network average to individual affiliate), you need to generate your own data. The only way that I can see to do that is by testing.

        The cb numbers might give you some ideas which vendors to test first, but more importantly, you can actually see every vendors sales pitch, so you can make a judgement about which vendors are worth testing.
        Signature
        ClickBank Vendor?
        - Protect Your Thank You Pages & Downloads
        - Give Your Affiliates Multiple Landing Pages (Video Demo)
        - Killer Graphics for Your Site
        SPECIAL WSO PRICES FOR WARRIORS + GET THE "CLICKBANK DISCOUNT" TOO!
        {{ DiscussionBoard.errors[7970635].message }}
        • Profile picture of the author Vivette
          I just recently start promoting CB products and I will be promoting #4.
          When I entered The CB Hoplink that I created into a new window it does not have my CB ID attached to the link.

          I have promoted 1 other vendor and I noticed that my CB ID was attached
          and I also take a look at some other vendors and my CB ID was attached

          So should I be promoting this vendor? How will sales be tracked?
          {{ DiscussionBoard.errors[8130137].message }}

Trending Topics