PPC and Performance Marketing: Paid Media Wired to the Database, Not the Platform Pixel
ScaleGrowth Digital runs paid search, paid social, and YouTube as a single engineering practice. The work starts with the unit economics pulled from your POS, CRM, or ledger, and ends with a campaign architecture that the bid platforms cannot quietly overspend. Engagements suit Indian fintech, multi-location F&B, healthcare, NBFC, and B2B SaaS where the founder no longer trusts the dashboard ROAS, and someone needs to reconcile the ad account to the bank account.
The problem we solve
Most performance accounts at INR 5 lakh per month and above are doing two things at once. They are buying high-intent traffic, which is the job. They are also paying retargeting and brand-pixel taxes on demand that the brand was going to convert anyway, which is not the job.
The pattern shows up in the same shape across categories. A founder-stated CAC of INR 800 turns into INR 2,200 once the post-iOS-14 pixel drop and the view-through window are stripped out. A “1.4x ROAS” Meta account turns into 0.7x once the database is asked which orders were net-new versus repeat. A Google Ads account that “drove” 100 leads last month turns out to have driven 28, with the other 72 sitting in Direct, branded, or organic in GA4 because the pixel was greedy on attribution.
The inverse pattern is more common than founders expect. An instant-loan fintech with USD 6 million in funding ran 100 percent of acquisition on paid, holding a 28-keyword paid footprint driving 1.1 million monthly impressions, while the organic side ranked on 526 keywords (roughly 470 of them branded). The category leader sat on 63,352 organic keywords. PPC was not the problem. The absence of a paid-to-organic compounding loop was.
How ScaleGrowth approaches it
Paid media is treated as a controlled experiment against a known margin envelope. The methodology has four named pieces, run as a single engagement.
1. Margin extraction from the source of truth. Before any audit slide is built, per-customer or per-store revenue gets pulled directly from the operating database (Laravel, Shopify Admin, Salesforce, Rista POS, NBFC core banking, whatever the system of record is). On an 86-store F&B engagement, the founder-stated baseline was INR 4 lakh per store per month. The DB pull surfaced INR 1.58 lakh on the four pilot stores being modelled, a 2.5x correction that re-shaped the entire Q2 envelope from INR 51 to 78 lakh down to INR 6 to 9 lakh. Quarterly spend math gets locked against this number, not the pitch deck number.
2. Account audit against a defended bid floor. Search, Meta, and YouTube accounts get audited against a per-keyword and per-audience profitable bid ceiling derived from step 1. Search-term reports get parsed for waste (broad-match contamination, brand-term cannibalisation by the same account on Performance Max, dynamic-search-ads pulling junk landing pages). Negative-keyword and placement-exclusion lists are rebuilt against actual transaction data, not platform “recommendation” suggestions. The audit is paired with a technical SEO baseline when the brief includes paid-to-organic shift.
3. Attribution sanity, not attribution theatre. Server-side conversion APIs (CAPI for Meta, Enhanced Conversions for Google, Click ID stitching for B2B) get deployed so the platforms train on database events, not browser-pixel ghosts. The reporting tier merges platform data against the operating DB pull. When the platform reports 50 conversions and the CRM sees 20 qualified leads, the campaign learns on 20, not 50. Incrementality tests (geo-holdouts, intent-keyword pauses) get run to prove the campaign is producing net-new revenue rather than cannibalising organic demand.
4. The paid-to-organic compounding play. Where the data supports it, paid is run as a controlled feeder for category-defining organic keywords. The fintech case (1.1 million paid impressions, 526 organic keywords) is the canonical example. A 12-month plan re-routes a portion of paid budget into the content and entity work documented on the SEO services page, with the explicit goal that 18 months out, the same demand is being captured at a fraction of the CPC.
Proof
Two engagements, two different category economics, two corrected envelopes.
A multi-location F&B brand with 86 active stores. The brand had been internally reported as a 199-store network until a DB pull surfaced the truth: 86 active stores (51 COCO plus 35 FOFO), the rest closed and inflating the count. The proposed Q2 paid envelope had been INR 51 to 78 lakh, anchored on a founder-claimed INR 4 lakh per store per month. The Laravel ops-DB pull corrected per-store revenue to INR 1.58 lakh, and the Q2 cash investment landed at INR 5.93 to 8.23 lakh, an 8x compression. Inside that envelope: Pack 1 Meta at INR 3 to 4.5 lakh, modelled at 1.3 to 1.4x direct ROAS; Pack 3 Google at INR 2.25 to 3 lakh, modelled at 1.4 to 1.5x. Both numbers sit at the EBITDA-positive threshold for Indian QSR variable cost, derived backward from menu-mix economics. Background sits on the retail and multi-location methodology page.
An instant-loan fintech with USD 6 million in funding. The brand ran 100 percent paid acquisition with a 218-person team and four NBFC lending partners. A SEMrush adwords export plus a DataForSEO organic pull surfaced 28 paid keywords driving 1.1 million monthly impressions against 526 organic keywords (around 470 of them branded). The category leader sat on 63,352 organic keywords. Every paid impression was being taxed because no organic floor existed underneath. The fix sequence specified a 12-month plan to shift 20 to 30 percent of paid spend into content and entity work, with the paid account simultaneously rebuilt on Enhanced Conversions and server-side CAPI.
Process: what working with us looks like
Engagements run on a fixed 90-day commercial structure. The first 14 days are diagnostic. Week one is data plumbing (read access to ads accounts, GA4, CRM or POS export, ledger or P&L for variable-cost math). Week two is the margin pull and the search-term and placement audit. End of week two ships a written diagnosis with the corrected CAC ceiling, the wasted-spend tally, and the recommended account restructure.
Days 15 to 45 cover the rebuild. Conversion tracking moves to server-side (CAPI, Enhanced Conversions, Click ID logging). Account structure gets rewritten: SKAG or single-theme ad-group structure where Smart Bidding is fed clean signals, brand and non-brand split, Performance Max constrained with audience exclusions so it stops cannibalising organic search. Creative production runs in parallel: static and video assets briefed against the audience cohorts the database actually identified as profitable.
Days 46 to 90 are the controlled scale period. Geo or audience holdouts run alongside the live account so incrementality is measurable. The reporting layer pulls daily from the operating DB and the ad platforms; any divergence above 15 percent triggers a manual reconciliation. Reviews happen against the merged dataset. By day 90, the engagement renews on a maintenance fee against a defended ROAS target or hands ownership back with documentation.
Pricing
Diagnostic engagements (the 14-day audit and corrected CAC envelope, deliverable as a written brief plus a working ROAS calculator) are priced at INR 2,50,000. The 90-day rebuild is INR 8,50,000 to INR 14,00,000 depending on monthly media volume, ad platforms in scope, and whether server-side tracking and CRM stitching are included. Ongoing management is a flat retainer of INR 3,50,000 to INR 6,50,000 per month against a written ROAS or CAC target. Percentage-of-spend pricing is not offered. Multi-country accounts and accounts above INR 50 lakh monthly media spend are quoted separately.
FAQ
Why won’t you charge percentage of ad spend like other agencies?
Percentage-of-spend incentivises the agency to grow the media budget, not the profit margin. The 86-store F&B engagement is the cleanest example: a percentage-of-spend agency on that account would have been pushing for the INR 51 to 78 lakh envelope, not the corrected INR 6 to 9 lakh one. Flat retainers tied to a written ROAS or CAC target are the only structure that lets the recommendation honestly include “spend less.”
How do you handle attribution when iOS, Safari ITP, and ad blockers eat 30 to 50 percent of pixel data?
Server-side. Meta Conversions API, Google Enhanced Conversions, and Click ID logging at the form or POS layer move the conversion signal off the browser pixel and onto the database. The bidder is trained on the events the operating system of record actually saw, not the events that survived the cookie cull. Where the CRM or POS has lag (offline conversions in a B2B funnel, in-store sales for retail), a daily offline-conversion upload pipeline gets wired in.
Do you run incrementality tests, or just trust platform ROAS?
Platform ROAS is the starting hypothesis, never the answer. Geo-holdout tests (a matched market is paused for 21 days, the delta is measured against the running market) and intent-keyword pauses (the brand keywords are paused for two weeks to read the organic floor underneath) are part of the 90-day rebuild. Most accounts above INR 10 lakh monthly spend turn out to have 20 to 40 percent of “paid” revenue that was happening anyway. That number is what the renewal conversation gets anchored to.
Can the engagement bridge paid and organic, or do we hire two agencies?
One engagement. The fintech case in Proof is the worked example: the paid account audit and the organic gap analysis were run by the same pod, with the 12-month plan explicitly routing budget across the two channels against a single blended-CAC target. Sites where the organic floor is too thin to ever compound (very young domains, regulated categories with editorial restrictions) get told that on the diagnosis call.
We have an in-house performance team. What changes?
Engagements pair with in-house teams more often than they replace them. In-house owns daily campaign operations and creative production. ScaleGrowth owns the margin extraction, the server-side tracking architecture, the search-term and Performance Max audits, and the incrementality testing protocol. Where the in-house team is already running clean server-side tracking, the engagement compresses to a senior-review model at a lower retainer.
An engineer walks the live account, names the likely waste and tracking gaps, and quotes the diagnostic before any contract.