Ecuador publishes:

• Address: countryName, locality, postalCode, region and street address
• Contact point: name and URL, (but contactPoint/name is always the same as parties/name)

For address, I guess we want to compare, country, locality region and street address all together and not street address alone.

And do we want to calculate this for bidders in the same process only or in general?

@Camilamila @jpmckinney

Based on https://colab.research.google.com/drive/1q38GlyG7B_uPCsqaFBt1UvT5FnNvEtbM#scrollTo=yg8SFe-09kvD I think this indicator is within the same process only, but I haven't compared to the methodologies in the academic sources.

I think it makes sense to combine fields into a full address, yes.

We might discover that we need to do some normalization (e.g. normalize whitespace, lowercase, maybe normalize punctuation). There's more that can be done #33, but I think we'll limit to basics for now.

I think this indicator is within the same process only

True, you are right, because this one is related to detecting collusion. The example that the notebook refers to, however, is Control Ciudadano from Paraguay, and there, we did the exercise with all the bidders, not depending on whether they were bidding on the same process or not. But for this indicator, we can implement the original and documented methodology that is for the same process only.

We might discover that we need to do some normalization (e.g. normalize whitespace, lowercase, maybe normalize punctuation). There's more that can be done #33, but I think we'll limit to basics for now.

I tested and even without any normalization and with exact match comparison I got a lot of matching in Ecuador's data (at least comparing all bidders no matter the ocid)

What should be the output of the indicator? Besides flagging the OCID, do we want to output the matching bidders along with why they are similar?

We would flag the bidders like in R024, etc. (using set_result! and set_tenderer_map!). We can give a score of 1.0 for exact match. I don't think we have any way to add additional metadata about why they are similar.

I don't think we have any way to add additional metadata about why they are similar.

But should we?

Maybe open a new issue with R044 as an example, since we could add more metadata to any indicator. Right now we don’t have any user research telling us that users want more metadata.