Prepared for Mr. Raghav Chadha · Confidential
Multimodal AI reads a video the way your audience does — tone, visuals, pacing, hook — and ties every choice to what actually drives engagement. We pointed this engine at your own Instagram and YouTube.
Why Galvor
The follower count told you that you lost. Here is what you kept, and who you can win back — read from your own public trace, public data only, no access to your accounts.
Multimodal AI now reads tone, visuals, pacing and hook, and ties every one of those choices to what actually drives followers and engagement. Very few people are working at this level of detail yet.
Every post, and a sample of every comment, decoded into structured attributes, so “what works” stops being a hunch and becomes a number you can act on before you publish.
Attention is the asset that compounds into everything else, and content is what earns it. The blank page after the move is yours to write, now — and this tells you exactly what to write on it.
57% of your commenters are the news & accountability crowd and 56% still back you. Your proven “name an injustice, put a number on it” formula still scores 86%. The base held; the drive-by crowd cooled.
Everything here comes from posts and comments anyone can open, gathered within Instagram’s and YouTube’s terms of service. Nothing private, nothing hacked, nothing bought.
The AI runs on our own computers, offline. Your data is never sent to any outside AI service — not in the US, not in China, nowhere abroad. It stays with us, and with you.
For the deeper read we’d take view-only access for one short session. We can never post, edit or delete anything, and you can switch it off the moment we’re done.
Step 01
This is what we learned.
03 · Context
The break starts with the AAP split (2–3 Apr) and is sealed by the BJP move (24 Apr), not creative quality: reach was rising before it, then May onward sits in a far lower band across every format tried. Posting frequency also fell ~75%.
Written comment sentiment
Hostile 13→32% · a trust problem, not a reach one.
Step 02
But GenAI dug deeper. Here’s what’s actually working for you.
05 · Insight 1
How each kind of post scores (reach × audience agreement, 0–100%)
Tap n= on any row to open real posts from that group.
03 · The recipe
Read it as a checklist. When your posts carry these ingredients, they land in your top tier this often — win rate when the trait is used.
Tap Example on any row to open a real post. Counted directly from the 32 Instagram posts we can fully score (real view/like/comment data + classified comments — the vocab-tagged political corpus; YouTube isn’t scoreable yet, see footer). Each example’s own Winning Score (reach percentile + like-rate percentile + comment positivity, 0–100, the same formula behind compute_winning.py) is shown so a huge view count never gets shown as proof of a trait it actually scored badly on.
Green = win rate when the trait is used. Red = reach penalty when the trait is present — both from the same underlying trait-delta analysis, shown as percentages.
06 · Insight 2
9,937 commenters in the sample, by segment · all eight shown, three featured below.
Where support cooled most, % supportive before → after the switch. Your news & accountability core isn’t on this list.
The sharpest drops are the casual groups who never had much at stake. Your base held better than the drive-by crowd.
Illustrative composite, not real people. Built from aggregated, non-identifying audience signals such as content engagement themes, creator/business interests, professional context, broad regional patterns, and age bands. Not intended for individual-level targeting, sensitive profiling, or political persuasion.
06 · The comeback playbook
On the data, you have handled this about as well as it can be handled. Six real repositionings show the shape of what works. What to do next comes from two places: how switchers who recovered actually did it, and your own account data.
Real party switches, India and abroad — the ones that recovered, and the ones that didn’t.
Out-delivered it: Rajya Sabha in 3 months, Union Cabinet in 16, won back his old seat by 5.4 lakh. New wins buried the old label.
Never argued the point, flipped states for BJP, became Assam CM. The doubts had faded within ~18 months. The strongest case here.
Never answered the doubts about his move, and skipped the contest himself. His party won 0 of 23 seats; dissolved by 2025.
Admitted it was survival math; the line became his opponent’s ad. Lost the primary; a 30-year career over.
Launched her own show the same day she left; “I didn’t change, the party did.” Confirmed DNI within ~2.5 years.
Made “principled outsider” the brand, so switching reads as consistent. Kingmaker for 33 years. *With real setbacks.
Six illustrative cases, not an exhaustive sample; outcomes are inferred from trajectory, not controlled comparison.
Keep doing your NEET / recharge / consumer advocacy as a BJP MP, shown as continuity of purpose. Silence on old causes reads as having quietly dropped them. Mr. Sarma never re-argued his switch and was Assam’s CM within years; Mr. Scindia had a Union Cabinet seat in 16 months. Speed and visible delivery did the talking, not silence.
Direct-to-camera and unedited: address the switch, the AAP allegations and what you’re doing now. Your audience is arguing about you without you in the room. Mrs. Gabbard launched her own show the very day she left the Democratic Party, and used it to tell her own story before anyone else could.
Make the case affirmatively and often. Quietly deleting past posts reads as concealment, not conviction. Mrs. Gabbard’s own line — “I didn’t change, the party did” — became her defining explanation; Mr. Scindia never went quiet about his reasons either.
You were followed for consumer-rights and accountability, not AAP loyalty. Recast as “I fight these issues whoever’s in power,” and visibly do it. Mr. Peters branded himself the “principled outsider” above any single party, and stayed a kingmaker in New Zealand politics for 33 years on it.
Bottom line, your fastest lever. The four who recovered made the label irrelevant through fast, visible, consistent action; the two who failed stayed passive or silent. The cheapest move today needs no election cycle: owned, unfiltered, direct-to-camera content addressing the switch and your current work, aimed exactly where the damage is concentrated — your own comment sections.
Issue-driven, came for your advocacy, not fandom. The biggest slice of your audience, and 56% still back you; the rest are arguing, not gone. This is who you rebuild around.
These issues came up unprompted, again and again, across the winnable segments:
Track a scheme’s money from the announcement to the actual hospital or school, and name the shortfall with a number. Against the leak and the delay, not your party.
The citizen turned away, the Ayushman card refused, the ration short, the seat gone. Put a face and a number on the gap between a policy and its delivery.
Your audience’s No. 1 unprompted demand. Hold the delivery system to account and propose the fix — for students, not against anyone.
How much is being lost, who should be protecting people, and the one concrete fix. Classic consumer-rights ground, brought up to date.
The numbers and the thinking behind a government call, in plain language — the honest translator, not the attacker. On-side, and unmistakably you.
Where your tax money actually goes, made legible — the CA in you turning the Budget into something anyone can follow. Transparency-first, exactly your old brand.
A direction pulled from your own numbers, not a finished script — your team owns the politics. Our part is making sure whatever you post carries the levers your audience already rewards, and telling you the moment it’s working.
Keep it personal, but let the moment carry a number: “my phone bill, line by line,” or “the same RTO queue as everyone.” Real setting, your face, one everyday unfairness, one fix.
Aim the same patriotism at a specific citizen’s unfairness, with a number — a veteran’s pension, a disability tax. This is your soldiers’-pension format.
Why it lifts: not a guess — your own data already shows this exact swing on the same theme: 76–90% vs 7–22%. A 4–10× change from one adjustment.
Lead with conviction and issues, never electoral math. Rivals are already mining your old lines. Mr. Specter’s own quote admitting the switch was survival math became his opponent’s campaign ad; he lost the 2010 primary.
The Rajya Sabha committee chair is necessary, not sufficient. Pair it with direct outreach to skeptical former supporters. Mr. Specter leaned on his Senate seniority instead of rebuilding grassroots trust — and still lost the 2010 primary, ending a 30-year career.
Staying silent on the viral critical videos mirrors Mr. Azad’s silence, read by your own base as an admission. Mr. Azad never directly answered the doubts about his own move; his new party went on to win 0 of 23 seats and dissolved by 2025.
Visible, full commitment converts skeptics faster than a cautious, defensive crouch. Mr. Azad founded his own party but skipped contesting a seat himself — the kind of half-measure that reads as hedging.
Every collab, guest and borrowed cause is read as an endorsement now. A due-diligence screen reads any creator or topic in minutes — three illustrative reads:
Clean trail, stance-aligned, audience overlaps your winnable base. No controversy in reach. The kind of collab that compounds your brand.
Huge reach, but one old divisive clip and a fan-base heavy on drive-by hostility. Workable only with tight framing, never open-ended.
A live controversy and a record of attacking your current party. The association would instantly become the story, not the message.
Your lowest-scoring format on the data: studio production −50%, lifestyle framing −38%, landing around ~8% overall. Looks polished, carries no issue — and it shows in the numbers.
Among your worst-performing posts (7–22%). The same patriotism aimed at a specific citizen’s unfairness, with a number, is one of your best (see “Do · from your data”) — this version isn’t.
The auto / cricket / anime / fitness crowd is ~7% of the room, arrives via viral clips, and cooled the most of any segment since the switch. Chasing their reach pulls attention away from the majority that actually holds your coalition.
08 · Why now
You are rebuilding a public profile with a real election on the horizon. Mr. Zohran Mamdani built a mayoralty on Instagram; India’s Cockroach Janta Party built 19M followers in a week. Social media is the campaign now. The blank page after the move is yours to write, now.
09 · The intelligence layer
The bigger picture
Every figure here came from the thin public trace: your public posts and a sample of public comments. Working together, with your complete history and the internal analytics of your own accounts, the same engine goes much further.
Method & scope: 94 public posts (39 Instagram, 55 YouTube), 5 Mar–25 Jun 2026; ~22,700 comments read directly across five videos (Hinglish, Punjabi and sarcasm included), 9,937 unique authors profiled from public bios. Segments grouped on behaviour only. Nothing modelled; every figure traces to source.
10 · Next step
Galvor can execute and perfect this playbook for you.
Not just the strategy. We run the loop with you: build the calendar, vet every piece before it goes live, track what’s working, and keep sharpening it, so this keeps compounding instead of resetting.
First 30 days: demand radar + audience baseline live by week 2 · your first vetted content calendar by week 4.