GL section 5. General Rules of Relevance

Rate each result individually and demote based on relevance and distance.
Always rate against the real world: If there is a better result available but it is not shown, demote the existing result(s) considering the missing one(s).
Ignore the order of the results and evaluate each one individually.
Ignore duplicate results and rate them individually. Do not demote for repetition.
If there are results that are extremely inappropriate or illegal, rate Bad. Identify such content by determining if the result is likely to be embarrassing if seen by users.

Address queries can also be just the name of a street, neighbourhood, intersection or city. Results for streets, neighbourhoods, intersection and city queries are eligible for Navigational

The expectation is that we return the requested entity with the same level of broadness that the user requested, which may not include all of the elements of a full address

 

Relevance: GL 5.14 Unexpected Results

 

The user is looking for a specific business in Sector 10 Gurugram and the result is only Sector 10. When the query contains a POI name, the user is expecting the POI as the result. Returning an address result is Unexpected and should be rated as Bad
Please review the GL examples of ‘The slanted Door’ and 10.6.1 (Gary Danko)
 

Relevance: GL 5.14 Unexpected Results

The user is looking for a specific street, but the result is a random POI that has partial query name. But it is still different form the user intent. Therefore, the correct rating should be Bad
Please review the GL examples of ‘soho’ and stevens creek blvd cupertino in section 5.14. Unexpected Results

 

Relevance: GL 5.18 Service-level Mismatch

The user intent is clear, they are looking for places that sell Gold Jewelry. Both the result sell fake jewelry or imitation jewelry. If the query only said ‘Jewelry’ it would be highly satisfying. However, since the query clearly mentions Gold Jewelry, both these results should be rated as Bad

Please review the GL examples of ‘burger king’ in section 5.18 Service-Level Mismatch

Relevance: 5.11 User inside the Fresh Viewport – Location intent is User location

If there are many results returned, zoom in your map even further to get a proper distribution of results of relevance. Many analysts rated this result as Excellent since it’s still within 1km distance from the

user. However, there are still closer results available to the user.

Since this is still further, we will give a rating of Good.

Relevance: GL 4.2 Relevance rating for Business/POI closed or does not exist

Many analysts rated this result incorrectly as Good as TryRating shows closer results available. However, when verified the closer results on the official website (https://www.myvi.in/help-support/store-locator) we can confirm that out of the 4 displayed results, only this one particular result exists. The closer should be treated as Closed. Since this result is the closest one available to the user, we should rate it as Excellent.

 

Relevance: 10.8.1 Routing Queries

Many analysts rated this as Bad as there is another Lajpat Nagar in Delhi which is not returned and considered more prominent. However, considering the user location and Viewport is nowhere near Delhi, we cannot be certain that the user meant Lajpat Nagar in Delhi. Therefore, the given result should be considered as Excellent as it’s almost equal distance from the user FVP.

10.8.1. Routing queries

When the query mentiones names of two distinct locations that are not near each other, the user was likely looking for driving directions, so returning either of the two distinct locations used for the route is the best experience Search can offer. Each locations should be rated Excellent for Relevance

Relevance: GL. 10.8.1 Routing Queries/ Lack of Connection

Many analysts rated this as Excellent. Clearly research tells us that this result is a completely different locality than the one in Query.

When the query is a routing query, results can return one of the two places mentioned in the result and it should be rated Excellent. However, if the returned place is a completely different one, rate the Relevance as Bad for not satisfying user’s intent.

 

Relevance: 10.2. Non-specific Address

Many analysts rated this result as excellent without applying GL 10.2 correctly. This result should be rated as Navigational as it is the closest result without ANY other results that are equally close.

Relevance: 10.2. Non-specific Address

The closest result is very close as compared to any other result. Therefore, the correct rating should be Navigational

 

Relevance: 10.2. Non-specific Address

Exact location returned within a fresh viewport is eligible for Navigational, unless other addresses that satisfy the user intent are also very close.

Results that are farther away from an ideal result but can potentially satisfy the user’s intent can be rated as high as Good (Distance/Prominence)

 

Relevance: GL 5.16.1 Transit Queries [Stockport station example]

When the query is a transit station that shared the name of the neighbourhood, please consider it as a query for specific station plus category query with location modifier. All stations within the same sublocality as the queried station, should be rated as Excellent.

Similar to Stockport station example in the GL, Rajouri Garden Metro station should be rated Navigational and all other metro station within Rajouri should be rated as Excellent. Tagore Garden Metro station should therefore be rated as Excellent and the metro stations in the closest next neighbourhood should be rated as Good.

Re   Relevance: GL 10.7.1 Category Query with General location

TryRating address and pin shows the result in Sector 9. While rating relevance we consider TryRating is correct and rate the relevance as Good for result being just outside sector 8

Relevance is always rated independent of any data
(Name/classification, address, pin) inaccuracies. This means that when rating relevance we always assume that the result exists (even if research reveals the location is closed) and that the data presented is correct.

 

Difference between General Location modifier & specific Location modifier 

General Location Modifier

Specific Location Modifier

GL 10.6.3.1 starbuckssanfracisco& Aldi Waco txexamples

GL 10.6.3.2

kate spade 789 Madison Ave, New York & Tacodeli burnet rd examples

Sub-Locality2/sublocality1/Locality/ Country & broader

Full Address/Building Name/Mall Name/ Street Name

If only few results within the location, results outside are still relevant

Results outside the location are not Relevant and should be rated as BAD

Relevance: 10.4. Point of Interest

.

 

The user is looking for the prominent landmark ‘Taj Mahal’ which is in Agra. However, the returned result is a restaurant named Taj Mahal VERY close to the Fresh Viewport. Therefore, according to the Guidelines example of Machu Picchu (pg189), the result should be rated as Good.

The above result is quite far from the viewport. However, this result is for the prominent POI ‘Taj Mahal’. Ac

cording to GL 10.4, “A correct location result should be rated Navigational for queries referring to a point of interest.”

 
  

 

Relevance: Chain Business – Fresh viewport with User outside

 

Query – Mcdonalds

.

Relevance: 1.3.3. Result Relevance Rating

Query – stadium Hyderabad, Telangana, India

The relevance should be rated independent of data inaccuracies. The name says Race Club; however, we cannot be sure if there is no stadium here. Therefore, we should consider this as a stadium (mentioned in the classification) while rating Relevance and then make further research regarding the data accuracy. The relevance for this should be rated as Excellent as it matches the category stadium within the location modifier, ‘Hyderabad, Telangana’.

1.3.3 Result Relevance rating

3. Hyderabad Race Club

Address

Malakpet

Hyderabad Telangana 500036

India

 

Classification

Stadiurns and Arenas

Distance to User

138.31 km

Distance to Viewport

17.698 km

Lat, Lng

17.380743, 78.507099

 

.1.3.3 Result Relevance rating

The relevance rating will take into consideration user intent as well as distance/prominence. Relevance is always rated independently of any data (name/classification, address, or pin) inaccuracies. This means that when rating relevance we always assume that the result exists (even if research reveals the location is closed) and that the data presented is correct.

Always rate against the real world: If there is a better result available but it is not shown, demote the existing result(s) while considering the missing one(s).

 

Relevance: Query for a specific POI/Result is different POI

.

 

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Post