Deciding where to campaign
The political scientist Fernando Limongi seems to be the Folha's darling of the weekend. He's also been interviewed (briefly) by the paper regarding the nature of the vote in Sao Paulo and the general prospects for campaigning by the candidates in the city.
According to him the October mayoral poll won't reflect the federal elections, but at the same time neither is it a blank canvas (good hedging of bets there!). The centre of the city is the most politicised and likely to be stable in terms of decided voters. So for candidates it may be a good idea to get out to the periphery (which for that also read lower income) where the opportunity to influence the vote may be stronger. But again he also reserves his comments by suggesting that work can be done in the centre.
In the end what he and other analysts do know is that Sao Paulo voters don't cross from left to right or vice versa easily. In this then the city is probably unlike parts of the country in the north where such ideological affiliation is much weaker and tied to a personality. Certainly in this respect Sao Paulo's poll will be quite unlike the rest of Brazil in October.
Of course if this was the UK and we were talking about the Lib Dems then there would be talk of past election results and canvassing returns. But Brazilian elections don't tend to involve canvassing (not only is it too difficult to get to people's homes, in come cases it may be too dangerous). Chances are this analysis will be used not to identify individual voters, but target publicity material and ad campaigns in particular areas of the city where potential supporters live.