How to read a constituency before you decide to contest it

Pull booth-wise results for the last three assembly elections and the last two general elections in the same territory. One cycle is noise — a wave year, a strong local candidate, a split opposition. Three cycles show you the floor: the vote share your party gets in this seat when nothing special happens. That floor, not the best result, is your honest starting point.

Sort every polling station into four buckets by historical performance: Strong (your side above 60% consistently), Winnable (45–60%), Swing (the 15–20% that flip between cycles), and Hostile. In a typical assembly seat this produces something like 70 strong, 85 winnable, 95 swing and 50 hostile. The entire campaign question is what happens in the swing column, and no aggregate vote-share number will ever tell you that.

In most Indian seats, the deciding factor is not the main opponent but the third and fourth candidates. Look at who took 4–12% in past cycles and where. A caste-based regional party polling 8% concentrated in eleven booths is a completely different problem from one polling 8% spread evenly. The first can be negotiated or targeted; the second usually cannot.

Electoral rolls in urban and peri-urban seats routinely carry 8–14% dead, duplicated or shifted entries. Every turnout percentage you have looked at is therefore wrong, usually in a way that flatters low-turnout areas. Before drawing conclusions about apathy, verify the roll in a sample of twenty booths. We have seen a supposed turnout problem turn out to be a roll problem.

One: is the historical floor within 8 points of the winning number? Two: are the swing booths geographically concentrated enough to be worked by a real ground team? Three: is there a single issue with over 30% unprompted salience that the incumbent cannot claim? Four: does the party structure have at least 60% of listed booth workers actually active? Two or fewer yeses, and you are funding somebody else's victory margin.