[projectvrm] value of market research (was: Re: Open Loyalty)

[Henri Asseily < >]

On May 11, 2011, at 5:01 PM, Katherine Warman Kern wrote:

> Really interesting Henri.  I'm curious about what you think about this.  In 
> the model below, you performed research exclusively fir one company 
> designed to satisfy their objectives, right?   What factors effected 
> pricing?  I suspect that sample size was an important factor?

The research was across all companies participating in the bizrate.com 
ratings program. Data acquisition was done at 2 specific points of the 
purchase process: just after the purchase event at the merchant 
(point-of-purchase info), and via an email that was sent a few days after the 
expected date of delivery (post-purchase info).
A very unique feature of the bizrate.com surveys was that the surveys were 
24/7, had no end date, and the questions could be changed at any time. They 
were highly customizable, with a fixed core of questions that were the base 
data across all companies, and unique questions that the company would setup 
for its own interests. The result for merchants was the ability to compare 
themselves to their industry as well as to request whatever they felt they 
should know of their customers via a trusted 3rd party for anonymity.

The whole pricing model for the research at first was based upon how much 
customization the merchants wanted to do, and whether they wanted more 
advanced research. But later on (2001 onwards) I completely automated 
everything which allowed us to give all the above for free to the merchants 
via an extranet. They could do queries on their data in real time over the 
previous 13 months. If another entity wanted industry data, they'd pay for 
the report.

Regarding sample size, that was our strength: we weren't going for sampling, 
we were trying to do a census. I think (memory is fuzzy by now) that response 
rates stabilized between 10% and 30%, which is absolutely massive considering 
that the surveying was done 24/7. With such a "sample size", all sorts of 
subgroups could be analyzed without worrying about sample size.

> 
> Now if our model flips the purpose of the research from being to meet one 
> company's needs to meet the needs of groups of individuals, I wonder what 
> the minimum "sample" size would need to be to be valued by business.

It's just statistics. As a rule of thumb, when you want to analyze a group, 
you need at least 30 full data sets to get some confidence. With 100 you're 
good. The problem is that you often want to compare for example households 
with and without kids. In that case you want 30 sets in each group.
> 
> Do you think they would pay more or less than your model of custom, 
> proprietary research if this info would be sold to any business who wanted 
> the information?  In other words, my assumption is that they may pay less 
> than they'd pay for custom research, but multiple companies would buy the 
> info.  In fact, just the need to know what the competition knows would 
> compel all competitors to buy - sort of like Nielsen market share data.  

There is no real business in pure market research online because companies 
look at the very short term online (think high volatility). It needs to show 
direct actionable value, such as reducing exit rates or increasing 
click-through rates, and then it's as valuable as the verified increase in 
profits. Another use of the data as I said earlier could be to create buying 
groups of similar individuals who want to pool their purchasing power. Then 
the value is to the individuals, and not the companies. I don't think it's 
been done yet at this level. There's stuff done based on behavior such as 
amazon's "people like you bought x" but nothing as deep as pooling clusters 
of demographics or psychographics as entered by the individuals, not inferred 
by the corporations.

So the data acquired for the market research from the customers with the 
customer's active participation (as opposed to the data being inferred from 
his behavior) could be used to provide buying power back for those customers. 
The thing to be careful about is that you still need the merchants for the 
very valuable point-of-sale acquisition entrypoint but as bizrate.com showed, 
they're very willing to do that when also given back their own aggregated 
customer data for analysis, and aggregated industry/competition data for a 
small fee.

> 
> 
> Katherine Warman Kern
> www.comradity.com
> @comradity
> 203-918-2617
> 
> On May 10, 2011, at 2:49 PM, Henri Asseily 
> <
 >
>  wrote:
> 
>> On May 10, 2011, at 3:30 PM, Katherine Warman Kern wrote:
>> 
>>> What about a cooperative as the business model?  But instead of forming a 
>>> purchasing co-op to leverage clout for better pricing, this is a 
>>> cooperative leveraging the clout of the demand of many to get better 
>>> stuff.  
>>> 
>>> We use revenues to pay for operating costs and share profits with members 
>>> - like dividends- to motivate people to take the time to share thoughtful 
>>> info and be honest and share rheir info exclusively with us.  The better 
>>> info we generate and the better we handle it, the more value will be 
>>> created at three links in the marketing chain:
>>> 1)Better than insights than business gets from proprietary consumer 
>>> research (even if it is shared with competitors).
>>> 2)We get more relevant offers in response to our info and business gets 
>>> more receptivity to their offers.
>>> 3)retail transactions are managed in our trusted ecosystem which gives 
>>> customers the control over what info they contribute back to the co-op to 
>>> encourage continuous improvement from business.
>>> 
>>> The implication is that there's are three differrent sources of revenue:
>>> 1) from businesses interested in the info we generate as a group (In my 
>>> opinion the output will have more value if it is not one rfp (which will 
>>> be hard to reach consensus on anyway) but more like a cluster map - but 
>>> this is getting ahead of where we are now). 
>>> 2) from businesses interested in presenting responses to those our info 
>>> predicts will be most receptive and influence others
>>> 3) from customers who want us to handle the transaction (because they 
>>> trust our personal data ecosystem and want the option of contributing 
>>> their data exclusively back into our data analysis to signal needs for 
>>> continuous improvement)
>>> 
>>> I don't know if the legal definition is a non-profit or for profit.  
>>> Would be great to hear the pros and cons from those in the know.  The 
>>> principle is that the customers who generate the info are rewarded for 
>>> the value they contribute after the cost of making sense of that info is 
>>> paid for.   
>>> 
>>> Thoughts?
>> 
>> At a glance, this gets very close to a large chunk of the business model 
>> of my previous company, Shopzilla.com. When we started, we wanted to 
>> reduce friction between customers and vendors by doing a few things:
>> - help customers find the best vendors for them
>> - help the vendors better understand their customer needs
>> - help vendors acquire customers through their stellar service ratings
>> 
>> And then later on we added the following:
>> - help customers find the best products for them
>> - help vendors directly acquire leads through a hybrid PPC auction system 
>> in the product listing
>> 
>> The first part was the basis of the BizRate point-of-sale ratings, which 
>> is interesting because we did the ratings before the marketplace, while 
>> eBay (for example) had to build ratings to help its existing marketplace.
>> 
>> So we used to (I don't know if it's still done, I left some years ago) do 
>> a lot of market research for the vendors, and we would daily process tens 
>> of thousands of in-depth surveys, aggregating, correlating and building 
>> matrices. The surveys were significantly more interesting than the current 
>> cookie-based behavioral stuff, in that they asked pointy questions that 
>> were relevant to each vendor and touched upon demographics and 
>> psychographics. For example, babystyle.com had a question about the # of 
>> children in the household. The end result would be that vendors had a very 
>> good idea who their customers were, what they needed to do to improve, and 
>> where they had an opportunity to impact their bottom line.
>> 
>> If I were to do it all again today, I would certainly look for a way to 
>> allow customers who filled in surveys to automatically find similar, 
>> like-minded and like-behaviored people for potential group buying 
>> opportunities. Now that I think about it, it should have been pretty damn 
>> obvious, but once we started down the path of a PPC marketplace, that took 
>> precedence.
>> 
>> The weird thing is that people did give us a HUGE amount of personal data, 
>> simply because they saw obvious, instant value to it: helping a merchant 
>> improve, giving one's opinion of a purchase, while using a (trusted) 3rd 
>> party that would aggregate, normalize and anonymize the data.
>> We struggled for years to find a better revenue stream than pure market 
>> research for merchants which doesn't pay much, but I think that 
>> algorithmic clusters for group buying power would have been a killer app 
>> for all this data.
>> 
>> H.