A ₹42L/year skincare client caught a ROAS calculation that overstated their performance by 22% and threatened to walk. Two analysts quit in Q2 citing burnout. Pitches kept slipping.
₹42L
contract on the brink
How a Bangalore agency stopped burning 2,100 analyst hours a quarter on weekly client reports.
Northpoint Performance runs paid acquisition for 47 D2C brands across India and SEA — fashion, beauty, consumer health — on Meta, Google, and Amazon. 62 employees, roughly ₹18 Cr ARR, known internally for sharp creative and aggressive ROAS targets.
Our protagonist is Megha Iyer, Head of Client Services, who owns retention. What kept her up every Sunday night was the same thing: would Monday's reports actually go out on Monday.
Every client expected a weekly report by Monday 10am. Eight analysts spent Tuesday through Thursday pulling data from Meta Ads Manager, Google Ads, GA4, Shopify, and Klaviyo, then stitching it in Sheets and Slides. Each report took 3 hours 25 minutes on average. Across 47 clients, that's 160+ hours a week — just to produce decks.
Internal audits flagged a 6.2% error rate, mostly attribution-window mismatches and copy-paste mistakes. Senior strategists were doing junior data work, and new-business pitches kept slipping by a week.
A ₹42L/year skincare client caught a ROAS calculation that overstated their performance by 22% and threatened to walk. Two analysts quit in Q2 citing burnout. Pitches kept slipping.
Two senior strategists were doing junior data work. Sunday-night Slacks had become a ritual. Something had to change.
Vynara built a unified data layer on Postgres, with nightly syncs from all five platforms via Airbyte connectors and custom pulls. Each client's data lands in a normalized schema. An LLM-driven reporting layer generates a draft narrative every Sunday night — anomalies, week-on-week swings, and underperforming campaigns flagged inline.
7 weeks, build to deployed. Analysts open the pre-built draft Monday morning, add client-specific context the model can't know (creative refreshes, market events, internal launches), and push. Data pulling, joining, charting, and baseline commentary became automated. Client narrative, recommendations, and the relationship stayed human.
The 5 analysts freed up moved to strategy + new business. Q4 pitches won went from 4 to 11.
The first Monday it actually worked, two analysts asked if they could leave early. I said no — go work on the Delhi cohort pitch. That's the real win.
I was ready to quit. I came here to run campaigns, not to be a copy-paste machine. Now I'm actually building optimization playbooks again.
Most agencies don't have a headcount problem — they have a workflow problem. We can replicate this for any agency running 20+ clients across 3+ ad and analytics platforms.