Re: Seeking overlap algorithm

On Dec 6, 2007 9:14 PM, tedd <tedd.sperling@gmail.com> wrote: > At 6:07 PM -0500 12/6/07, Nathan Nobbe wrote: > >>On Dec 6, 2007 4:59 PM, tedd <tedd@sperling.com> wrote: > >> > The problem is, could you guarantee to a "preferred" service > provider > >> > that they would receive top-listing 25 percent of the time? Keep in > >> > mind that preferred service providers will overlap. So, the question > >> > is how to accommodate for that overlap? > >> > > >> > If anyone has an algorithm, I would be interested in hearing it. > > > >i think that should be pretty easy. the algorithm is dependent upon > >2 values. the number of listings you consider as top-listings for > >each result set and the number of customers that are registered as > >preferred customers. > > > >let me illustrate. imagine you designate only the first entry of the > >resultant listings as a * > >top-listing*. in that case you can support a maximum of 4 preferred > >customers before you > >can no longer guarantee to each of the preferred customers their > >listing will be > >a top-listing 25% of the time. so, if you want more capacity you > >can increase the number > >of results that are designated as top-listing results. > > > >suppose you increase the number to 2, now you have a capacity of 8 > >customers you can guarantee top-listing status 25% of the time. > > > >what you would need to determine is what to do if ever you didnt > >want to increment the number of listings that are designated as > >top-listing customers and you were already at capacity for the > >number of customers the current capacity supports. > > Yes, that was pretty easy, but that was not the answer to the > question -- my error for not explaining it better. > > Let me rephrase the question by providing an example. > > Let's say we have a customer base that is spread-out at random over a > geographic area. Each customer has designated a 50 mile radius from > their location as being within their zone -- the map would look like > a bomb saturation map, if you know what I mean. > > Now, many of those areas overlap so that if a end-user is within that > overlap he can see all the service providers that can provide > service. It's a simple matter to pull those providers out of a > database depending upon distance and show them to him. After all, > that's the way it works, isn't it? The end-user is provided all the > service providers who are within their service range. > > However, if you are also considering that some of these service > providers should be shown as "preferred" (i.e., at the top 25 percent > of the time) then you might find yourself in a position of over > selling the top position because there may be too many "preferred" > service providers in certain areas. > > Now, what I need is a way to analyze the distribution of the current > service providers to see if a given location is open to being sold as > a "preferred" position -- do you see what I mean? > > Another example, let's say we have four "preferred" service providers > at the same location. Obviously, we could not sell another > "preferred" position within 100 miles. > > Another example, let's say we have four "preferred" service providers > 100 miles apart, clearly we can sell more "preferred" positions. But, > the number of positions available depends upon the distribution of > the original four. If they were located in a straight line, then we > could sell two positions between each one. But, if they were > distributed in a square, we could only sell one. Do you see? > > I know what solution I will be using unless someone comes up with > something different. I just want to tap this knowledgeable group > before I spin my wheels trying to solve a problem, that may be > already solved. i see the problem more clearly now. unfortunately, i dont know the solution off the top of my head :( if i think of something in the near future ill toss it on the list. -nathan