Designing a better traffic metric

[Coffee Pancake]

Traffic as a raw measure of actual popularity is junk and easily gamed.

As we're unlikely to get anything meaningful from LL; What would a scripted popularity meter look like, what would it measure, how would it avoid being gamed (or at least make gaming it so annoying to actual visitors as to be counter productive). Should it be interactive in anyway or entirely passive.

What behaviors should it encourage locations that place the meter to get their visitors to engage in.

What meaningful (anonymous) information could it provide to places that place such a device.

For the purposes of maintaining privacy, such a device would have to do all the processing in LSL with the goal of submitting minimal information (a couple of numbers) to an external website.

 

[Lucia Nightfire]

Replace the current attendance-only based calculation with either a search term matching method or a profile based, "I am looking for.." textbox terms matching relevance based method.

With a search term based relevance method, whenever anyone uses web or legacy search, the search term string they use becomes "bound" to that user, similar to how a scripted AO is bound.

Whenever the user enters a parcel that has Show in search active, if the search term bound to the user is found in the parcel's name or description, a traffic value for that term begins incrementing after the agent has been on the parcel for at least 1 minute and continues until either the agent leaves the parcel or uses web/legacy search with a new search term that is no longer found in the parcel's name or description.

Whenever anyone else uses web or legacy search and searches with the same term, only the traffic total from everyone that actually visited the parcel with that term bound to them shows to the person doing the search as a "relevance traffic" score.

These search term traffic "pools" keep track of each agent's start and end times as well as keep the last # hours of values alive for totals querying so current values can be referenced "live" instead of a 24 hour Pacific time with midnight cutoff value which currently leads to people going to places with no one around because the party/event happened the day before.

Let's say a parcel has "sex afk roleplay" in its description.

User A previously did a search for "sex". User previously did a search for "AFK". User C previously did a search for "sex". User D previously did a search for "puppies".

All 4 visited the parcel. User A stayed for 11 minutes. User B stayed for 6 minutes. User C stayed for 6 minutes. User D stayed for 61 minutes.

No one else visited the parcel in the last week.

Afterwards, anyone using web/legacy search searching for "sex" would see the parcel listed with a relevance score of 15. Searching for "AFK" would show a relevance score of 5. The parcel would not be listed when searching for "puppies". The attendance score, when updated the next day, would show 84 and would be just as useless as it is now.

How could land owners attempt to abuse this method? They could still log in a bunch of bots, but each bot would still only able to use one search term and each bot would have to use the same term that affects the most current and common/top search term(s) used by everyone else. They could ask everyone that shows up to change their search terms to a specific word, but they would maybe only be able to influence their regular customer base, but would they have much influence on strangers or non-regular visitors or would their regulars be sure to always change the desired search term back before visiting after previously doing searches for other places? Also, relevance traffic pool values are always in flux since user search term start/end times will always be entering/leaving them and the totals are not updated once every 24 hours.

 

With an "I am looking for.." CSV terms matching relevance based method, include an "I am looking..." textbox in the user's profile in the Interests tab. The user puts keywords in the textbox in CSV format or space delimited. This textbox is not public to other users.

Each term would be used to match terms in a parcel description as described above with single term searching. You do risk bots exploiting this more than with a single search term method.

Physical visits by users are still used with either method to increase totals instead of number of web/http queries alone to prevent a bot network from spamming searches of any/all relevant terms.

The above aims to keep traffic from being gamed and to make traffic a current/live representation of others searching for the same thing(s) and not a generic attendance record for the previous day.

Edited November 13, 2021 by Lucia Nightfire

[Mollymews]

7 hours ago, Lucia Nightfire said:

The above aims to keep traffic from being gamed and to make traffic a current/live representation of others searching for the same thing(s) and not a generic attendance record for the previous day.

as a premise i like this quite a lot

[cunomar]

What if , they aim to fix the impossible , to counteract human nature itself . By having nearby chat activity promote popularity instead of traffic ?

It will still be gamed but new players won't find themselves wandering a vast world of seemingly silent afk puppets .

Edited November 16, 2021 by cunomar

[June Starlight]

1 hour ago, cunomar said:

By having nearby chat activity promote popularity instead of traffic ?

Situation I immediately remembered is landing in an info hub on login (as my home region was offline) and seeing a bunch of bots on top of each other. Two of them had lengthy role play conversation. I was intrigued enough to stay and listen. It looked like the triggers were finding a new avatar in proximity (both greeted me) or hearing anything in local. The latter produced random Sci-fi/starfleet related piece of banter.

[Profaitchikenz Haiku]

Have a histograph showing the variations of number of avatars present and duration, most people going by traffic are looking to find other people, often more than one other person, particularly for RP sims, and again, looking for long-stayers, not just three people present for short 5-minute segments.

For shopping and adventure sms I'm not sure if traffic is significant as a destination-decider.

For sandboxes, possibly the opposite to high traffic is what is desirred?

Edited November 16, 2021 by Profaitchikenz Haiku