3 ms·

I’ve seen a lot of people promising to sell (aggregated, cleaned, private, inferred) data on Snowflake — using the database as a platform. Snowflake themselves

by bertil 3y ago

I’ve seen a lot of people promising to sell (aggregated, cleaned, private, inferred) data on Snowflake — using the database as a platform. Snowflake themselves have been listing this idea as a key value proposition. This is the first that I’ve seen “in the wild” (i.e. outside of an interview with a start-up that yet had to find a lot of paying customers). Is that common? Do you know of other companies doing that? Do they typically offer additional individual data (say: lead generation), inferred insights (say: likely spending by post-code based on local shops and disposable income), or something else (custom dashboard)?

izyda 3y ago
Most of the large data providers (S&P, Factset, Nielsen, ZoomInfo) have started to do so. I would say that they have partial solutions -- they do not list their entire catalog of assets on the platform. Cybersyn is obviously a startup explicitly trying to do just this -- AFAIK, we're the first "Snowflake Native" startup but it is hard to tell, there are more than 2200 data products in Snowflake Marketplace already so I easily could be wrong.
- bertil 3y ago
  The top of the free listing is repackaged open-source datasets (economic data, Covid). The top of the paid listing is the usual suspects: weather, geographic (GeoIP and matching elements), and you, CyberSyn, but nothing really exotic yet.
  - izyda 3y ago
    That's true but you would sort of expect those datasets to be the most popular just given the addressable market for those is by far the biggest (essentially every enterprise probably requires some degree of weather, Covid, and economic data). What sort of exotic data would be useful to you, out of curiosity?
    - bertil 3y ago
      I personally think opinionated high-value data would be more helpful, a bit like classifying US voters into seven groups has proven transformational for electoral campaigns. I’ve seen B2B sales desperately trying to sling their solutions to so many companies that are not at the right stage to value or understand it, so things like “companies with one level of engineering management” vs. “two” vs. “more.” “Companies disillusioned by terrible attempts at Agile,” or “… where the security team is a candle they forgot to light,” or “Companies with a violently dysfunctional Product leadership,” or “Companies with an inexperienced/checked-out CEO” or “…looking for a buyer” might be relevant — assuming you manage to rephrase those into less hurtful categories. Details gained from LinkedIn, etc. would be helpful to train our funnel model to say: “We seem more likely to sell our LLMaaS to companies with a certain number of technical employees or CTOs that like open-source.” Same for customers: I receive so many coupons for things I clearly do not need or care for. But there are things that I would be considering (my Amazon basket is a good indicator). You’d need to include relevant ideas and exclude things I’ve likely already bought, though… Given how Amazon Prime is unable to not insistently recommend TV series that I’ve finished and will hide what I’m actively watching even when I search for it by name, that could be a hard problem to fix. Matching that with the political thing: how to convince people to buy certain things. There’s been some controversy in trying to use the OCEAN model to get people to vote for certain people, but I think there’s room for an ethical way of telling marketers: that person would subscribe to HelloFresh because it’s cheaper than GrubHub, that person because they need structure in their life, that person because they want to eat healthy, that person because they like the idea of learning something, that person because they need to have something nice to serve their dates and that person, they already subscribe, stop being weird.

by bertil 3y ago

izyda 3y ago

Most of the large data providers (S&P, Factset, Nielsen, ZoomInfo) have started to do so. I would say that they have partial solutions -- they do not list their entire catalog of assets on the platform. Cybersyn is obviously a startup explicitly trying to do just this -- AFAIK, we're the first "Snowflake Native" startup but it is hard to tell, there are more than 2200 data products in Snowflake Marketplace already so I easily could be wrong.

bertil 3y ago
The top of the free listing is repackaged open-source datasets (economic data, Covid). The top of the paid listing is the usual suspects: weather, geographic (GeoIP and matching elements), and you, CyberSyn, but nothing really exotic yet.
- izyda 3y ago
  That's true but you would sort of expect those datasets to be the most popular just given the addressable market for those is by far the biggest (essentially every enterprise probably requires some degree of weather, Covid, and economic data). What sort of exotic data would be useful to you, out of curiosity?
  - bertil 3y ago
    I personally think opinionated high-value data would be more helpful, a bit like classifying US voters into seven groups has proven transformational for electoral campaigns. I’ve seen B2B sales desperately trying to sling their solutions to so many companies that are not at the right stage to value or understand it, so things like “companies with one level of engineering management” vs. “two” vs. “more.” “Companies disillusioned by terrible attempts at Agile,” or “… where the security team is a candle they forgot to light,” or “Companies with a violently dysfunctional Product leadership,” or “Companies with an inexperienced/checked-out CEO” or “…looking for a buyer” might be relevant — assuming you manage to rephrase those into less hurtful categories. Details gained from LinkedIn, etc. would be helpful to train our funnel model to say: “We seem more likely to sell our LLMaaS to companies with a certain number of technical employees or CTOs that like open-source.” Same for customers: I receive so many coupons for things I clearly do not need or care for. But there are things that I would be considering (my Amazon basket is a good indicator). You’d need to include relevant ideas and exclude things I’ve likely already bought, though… Given how Amazon Prime is unable to not insistently recommend TV series that I’ve finished and will hide what I’m actively watching even when I search for it by name, that could be a hard problem to fix. Matching that with the political thing: how to convince people to buy certain things. There’s been some controversy in trying to use the OCEAN model to get people to vote for certain people, but I think there’s room for an ethical way of telling marketers: that person would subscribe to HelloFresh because it’s cheaper than GrubHub, that person because they need structure in their life, that person because they want to eat healthy, that person because they like the idea of learning something, that person because they need to have something nice to serve their dates and that person, they already subscribe, stop being weird.