Assam 2026: BPN & Statscope India Exit Poll Got 99 Seats Right With 78.57% Accuracy
The 2026 Assam Assembly Election has now concluded, and with the final constituency-wise results in place, BPN & Statscope India can publicly present the performance of its Assam Exit Poll model. Out of 126 Assembly constituencies, our projection correctly predicted the winner in 99 seats, resulting in a seat-level accuracy of 78.57%.
In an era where most pollsters quietly move on after election day without publishing detailed accountability reports, BPN & Statscope India has chosen a different path. We believe data credibility is built not only by making predictions, but by openly comparing them against the final verdict seat by seat. To our knowledge, very few independent political forecasting platforms in India are currently publishing this level of post-result transparency and accountability.
Our final projection for Assam had placed the NDA in the range of 91 to 99 seats, the ASOM bloc in the range of 23 to 30 seats, and Others in the range of 4 to 5 seats. The final outcome delivered a strong NDA mandate, validating the broad direction of our model and our assessment that the BJP-led alliance remained structurally dominant across large parts of Assam.
The biggest takeaway from the Assam exercise was that the statewide mood was correctly captured. Our projection identified early that Upper Assam, large sections of Central Assam, and urban belts around Guwahati continued to heavily favour the NDA. The BJP’s organisational depth, alliance arithmetic, and consolidation of anti-opposition votes were visible in our internal assessment before counting day.
Our seat-by-seat model also correctly identified several difficult and politically sensitive contests where conventional political chatter had suggested otherwise. We successfully projected NDA holds in multiple Upper Assam battlegrounds and correctly anticipated that the BJP would continue to dominate the tea belt and urban constituencies.
At the same time, we also correctly recognised that the Congress-led opposition would remain competitive in selected minority-heavy and riverine constituencies. Our model did not project a total collapse of the opposition. Instead, it identified pockets where Congress and allied forces retained enough social and demographic strength to resist the statewide NDA wave.
The areas where our model underperformed were also instructive. Most of the incorrect projections came from highly localised contests involving minority vote fragmentation, tactical candidate effects, or unusually strong constituency-level swings that diverged from broader district trends. Seats such as Gauripur, Dhubri, Goalpara East, Mandia, and Binnakandi showed that hyper-local political behaviour can still overpower macro-level swing calculations.
Similarly, a few Upper Assam constituencies where we projected opposition resistance eventually swung more decisively toward the BJP than anticipated. Mahmora, Nazira, Jorhat, Mariani, Titabor, and Haflong demonstrated that the NDA’s late-stage consolidation in some regions was stronger than what our final swing calibration captured.
However, it is important to note that even most of our incorrect calls were not random misses. Many were marginal projection errors occurring in seats that remained highly competitive until the final stages of campaigning. The overall statewide direction of the election was correctly identified by our model.
One of the defining features of the BPN & Statscope India approach is that our forecasting system is not dependent on a single factor. We did conduct ground surveys and constituency-level voter sampling, but we deliberately did not give surveys alone overwhelming weightage. Instead, our methodology treated surveys as only one component within a larger multi-variable framework.
Our Assam model incorporated historical voting patterns, 2024 Lok Sabha voting trends, candidate strength, regional alliance arithmetic, community-based voting shifts, turnout patterns, district-level political momentum, local anti-incumbency signals, and real-time campaign energy assessment.
Each factor was given balanced consideration rather than allowing one dataset to dominate the entire projection. This hybrid approach is precisely why our model was able to broadly capture the statewide mood even in a politically complex state like Assam.
The Assam exercise has also helped us identify areas for further refinement. Minority-concentrated constituencies, tactical voting behaviour, and last-week regional swings will receive additional weighting adjustments in future models. We believe election forecasting in India is evolving rapidly, and static polling models alone are no longer enough to accurately capture voter behaviour in every constituency.
Most importantly, we are willing to publicly audit our own work. That culture of accountability is still rare in the Indian polling ecosystem. Many pollsters release headline numbers and disappear after the results. At BPN & Statscope India, we believe credibility comes from standing by the data after counting day, openly acknowledging both successes and misses, and continuously improving the methodology state after state.
Assam 2026 was another important step in building that long-term data-driven election model. We got 99 seats right, achieved 78.57% seat-level accuracy, and correctly captured the broad NDA advantage in the state. The remaining 27 seats, or 21.43%, will now become part of our internal review process so that our methodology becomes sharper, more localised, and more accurate in the next state election.
