[Data-modeling] /business/company

Jonathan W. Lowe jlowe at giswebsite.com
Tue Jul 22 21:34:46 UTC 2008


On Tue, 2008-07-22 at 22:33 +0100, Jonathan W. Lowe wrote:
> Jamie et al,
> 
> There was a point when the /business/company model changes needed for
> addition of non-business-chain companies with multiple office locations
> would become possible.  Somebody asked for a stay of execution in order
> to have more time to finish up a data load into the old schema.  I think
> that the waiting for this load turned into forgetting to change the
> schema.  Now that over two months have passed, is it safe to go ahead
> with the changes that Praveen, Jeff and Bryan have discussed and (I
> think) agreed?
> 
> Once changed the /business/company schema will then be able to receive a
> load of time-ranged Berkeley business locations...
> 
> Thanks,
> Jonathan
> 
> On Tue, 2008-05-13 at 13:42 -0700, Jamie Taylor wrote:
> > I'm very in favor of creating a location(s) property on company.   
> > Headquarters is a special location....and retail location seems like a  
> > co-type with specific properties.
> > 
> > ...splitting is a separate (more involved) issue.
> > 
> > On May 13, 2008, at 1:21 PM, Jeff Prucher wrote:
> > >
> > >
> > >> -----Original Message-----
> > >> From: Jonathan W. Lowe [mailto:jlowe at giswebsite.com]
> > >> Sent: Tuesday, May 13, 2008 11:17 AM
> > >>
> > >> The idea of having lots of properties associated with a
> > >> business type, only some of which properties apply to
> > >> particular business instances, is understandable.
> > >>
> > >> But the type breakdown of retail location, business chain and
> > >> company doesn't fit a business that has multiple addresses
> > >> for distributed offices, but is not a business chain -- e.g.
> > >> dentist's offices.  Do we have consensus on how to load this
> > >> sort of data?  Use /business_chain to hold businesses that
> > >> aren't really chains? ...or modify /company (and
> > >> /retail_location?) to hold multiple addresses? ...or...?
> > >
> > > I think Praveen's suggestion downthread is a good one. I.e., adding a
> > > property "locations" to "company", with an expected type of "business
> > > location". This avoids the semantic argument about whether a dentist  
> > > with
> > > two offices is really a "chain" or not, but provides more semantic  
> > > structure
> > > than just sticking multiple addresses on there.
> > >
> > >
> > >> Jonathan
> > >>
> > >> On Tue, 2008-05-13 at 10:51 -0700, Jeff Prucher wrote:
> > >>> My main issue with trying to split "company" into multiple types is
> > >>> that, apart from publicly-traded companies, which some
> > >> properties are
> > >>> very clearly specific to, I can't think of a way to split them that
> > >>> makes sense. This doesn't necessarily mean that there isn't
> > >> a way; I
> > >>> just haven't seen one yet. It seems to me like the properties could
> > >>> sort of be split into a couple conceptual buckets -- "big"
> > >> companies
> > >>> vs. "small" companies or "companies a lot of people care about" vs.
> > >>> "companies only locals probably care about" or even "companies we
> > >>> think we could get a lot of data for" vs "companies we
> > >> don't think we
> > >>> could get a lot of data for". All of these would probably
> > >> split along
> > >>> similar lines, and I agree that there are some good arguments for
> > >>> trying to split them; it definitely seems weird to think about the
> > >>> advisors and spin-offs for the liquor store down the block.
> > >> But if we
> > >>> took, say, the properties relating to mergers &
> > >> acquistions, revenue,
> > >>> and stocks off the company type, where would we put them?
> > >> This seems
> > >>> to be the biggest sticking point for me: What is the type for the
> > >>> "companies lots of people might care about"?  If we can
> > >> think of one that makes sense, it might be worth trying out.
> > >>>
> > >>> Jeff
> > >>>
> > >>>
> > >>>
> > >>>> -----Original Message-----
> > >>>> From: Praveen Paritosh [mailto:paritosh at metaweb.com]
> > >>>> Sent: Monday, May 12, 2008 7:24 PM
> > >>>> To: Jonathan W. Lowe
> > >>>> Cc: Jeff Prucher; bryan.cheung at metaweb.com; 'Kirrily
> > >> Robert'; 'Toby
> > >>>> Segaran'; 'Robert Cook'; 'John Giannandrea'; 'Jamie Taylor';
> > >>>> 'Christopher R. Maden'; 'Alexander Marks'
> > >>>> Subject: Re: /business/company
> > >>>>
> > >>>> Two concerns:
> > >>>>
> > >>>> 1. Detail in /business/company:
> > >>>>
> > >>>> Jeff suggested that creating a simpler company type and building
> > >>>> other types of companies by included type is not feasible, as it
> > >>>> breaks down in complexity as there are many different ways to
> > >>>> organize companies across the world.
> > >>>> Rather, have most of the meaningful properties that
> > >> people want to
> > >>>> talk about in one type and leave it at that. Jeff,
> > >> correct me if I
> > >>>> am misquoting you. It doesn't feel right to me, but maybe this is
> > >>>> the pragmatic thing to do.
> > >>>>
> > >>>> 2. Ability to talk about locations of non-chains:
> > >>>>
> > >>>> Companies have two location properties: headquarters and place
> > >>>> founded.
> > >>>> Additionally, if they were business chains, they have retail
> > >>>> locations.
> > >>>> But a non business chain (e.g., a Doctor's office in two
> > >>>> locations) has no way of talking about these. At this
> > >> point, I might
> > >>>> not know the headquarters or founding place, and it seems
> > >> better to
> > >>>> model it in the same way as business chain's retail
> > >> location. Only
> > >>>> that these are not necessarily retail locations, so
> > >> invent a type of
> > >>>> "business location"
> > >>>> and add a "locations" property to /business/company.
> > >>>> 	Jamie further suggested to make it the same property
> > >> which is used
> > >>>> by chains to point to their retail locations and change
> > >> the ECT of
> > >>>> that property to "business location".
> > >>>>
> > >>>> Meta: should this have been on the modeling list?
> > >>>>
> > >>>>
> > >>>> Thanks,
> > >>>> Praveen.
> > >>>>
> > >>>>
> > >>>>
> > >>>> Jonathan W. Lowe wrote:
> > >>>>> Hi Jeff,
> > >>>>>
> > >>>>> City of Berkeley has recently shared their business license
> > >>>> database
> > >>>>> contents -- all businesses since 1992 that have paid for a
> > >>>> license to
> > >>>>> do business in Berkeley, including geolocations and
> > >> NAICS codes.
> > >>>>> Based on the names and NAICS codes, many of these licenses
> > >>>> obviously
> > >>>>> correspond to very small businesses with fewer than 20
> > >> employees
> > >>>>> or just a sole-proprietor.  A good example is a dentist's
> > >>>> office.  There
> > >>>>> are between 10,000 and 13,000 Berkeley business licenses issued
> > >>>>> per year, most of which are this sort of "Mom & Pop"
> > >> kind of company.
> > >>>>>
> > >>>>> Most of the /business/company properties seem more
> > >> appropriate for
> > >>>>> corporate entities.  The dentists and auto-body repair
> > >>>> businesses of
> > >>>>> this world are unlikely to be eager to share market cap,
> > >>>> subsidiary,
> > >>>>> spin-off or acquisition data about their businesses
> > >> (even though
> > >>>>> they've been operating four times longer than the typical
> > >>>> startup).
> > >>>>> However, the Mom & Pop small business community seems like an
> > >>>>> appropriate target to try to hook on Freebase.  Hence the
> > >>>> suggestion
> > >>>>> for a simpler initial /business/company type that won't
> > >>>> confuse that community.
> > >>>>>
> > >>>>> Does this better contextualize the discussion?
> > >>>>>
> > >>>>> Jonathan
> > >>>>>
> > >>>>> On Mon, 2008-05-12 at 10:39 -0700, Jeff Prucher wrote:
> > >>>>>> Apart from ticker-symbol and major shareholders, which
> > >>>> properties do
> > >>>>>> you think are specific only to a single type of company? I
> > >>>>>> suspect that abstracting these out (I guess this is
> > >> actually the opposite:
> > >>>>>> de-abstracting? Concretizing?) would be a false economy.
> > >>>>>>
> > >>>>>> Jeff
> > >>>>>>
> > >>>>>>> -----Original Message-----
> > >>>>>>> From: Praveen Paritosh [mailto:paritosh at metaweb.com]
> > >>>>>>> Sent: Saturday, May 10, 2008 4:38 PM
> > >>>>>>> To: Jonathan W. Lowe; bryan.cheung at metaweb.com; Kirrily
> > >>>> Robert; Toby
> > >>>>>>> Segaran; Robert Cook; John Giannandrea; Jeff Prucher;
> > >>>> Jamie Taylor;
> > >>>>>>> Christopher R. Maden; Alexander Marks
> > >>>>>>> Subject: /business/company
> > >>>>>>>
> > >>>>>>>
> > >>>>>>> The /business/company schema is quite complicated, with the
> > >>>>>>> following
> > >>>>>>> properties:
> > >>>>>>>
> > >>>>>>>         "/business/company/founded",
> > >>>>>>>         "/business/company/slogan",
> > >>>>>>>         "/business/company/company_type",
> > >>>>>>>         "/business/company/place_founded",
> > >>>>>>>         "/business/company/headquarters",
> > >>>>>>>         "/business/company/industry",
> > >>>>>>>         "/business/company/revenue",
> > >>>>>>>         "/business/company/operating_income",
> > >>>>>>>         "/business/company/net_income",
> > >>>>>>>         "/business/company/number_of_employees",
> > >>>>>>>         "/business/company/parent_company",
> > >>>>>>>         "/business/company/ticker_symbol",
> > >>>>>>>         "/business/company/subsidiary_companies",
> > >>>>>>>         "/business/company/founders",
> > >>>>>>>         "/business/company/board_members",
> > >>>>>>>         "/business/company/advisors",
> > >>>>>>>         "/business/company/previous_names",
> > >>>>>>>         "/business/company/spun_off_from",
> > >>>>>>>         "/business/company/spin_offs",
> > >>>>>>>         "/business/company/companies_acquired",
> > >>>>>>>         "/business/company/acquired_by",
> > >>>>>>>         "/business/company/product_of_merger",
> > >>>>>>>         "/business/company/involved_in_merger",
> > >>>>>>>         "/business/company/market_capitalization",
> > >>>>>>>         "/business/company/major_shareholders"
> > >>>>>>>
> > >>>>>>>
> > >>>>>>> The /business/company/company_type takes values like
> > >>>> "Privately held
> > >>>>>>> company", "Partnership", etc.
> > >>>>>>>
> > >>>>>>> Some of the the properties do not apply to all
> > >> companies, e.g.,
> > >>>>>>> ticker_symbol does not make sense for a privately
> > >> held company
> > >>>>>>> or non-profit organization (right?).
> > >>>>>>>
> > >>>>>>> Wouldn't it make sense to have a very simple
> > >>>> /business/company type
> > >>>>>>> with a few properties, and then build each of the
> > >>>>>>> /business/company_types by including this simple company type
> > >>>>>>> and adding properties that make sense to that type of company?
> > >>>>>>>
> > >>>>>>> Thanks,
> > >>>>>>> Praveen
> > >>>>>>>
> > >>>>>>
> > >>>>>
> > >>>>
> > >>>
> > >>>
> > >>>
> > >>
> > >>
> > >
> > >
> > 
> > 



More information about the Data-modeling mailing list