Results

Proof, measured where it counts.

Lending marketing is easy to make look good and hard to make profitable. Below are three engagements with the starting numbers, footnoted metric definitions, the sanitised artifacts behind each figure — and the parts that did not work.

  • Lending & fintech only
  • Looker Studio reporting
  • RBI-aware creative

No cost, no obligation — a written audit within 5 working days.

Funnel snapshot

Live
  • Organic sessions+184%
  • Cost per approved application−38%
  • Qualified applications / month3.1×

Illustrative full-funnel outcomes over 9 months

9 mo

Typical time to compounding organic growth

−38%

Median drop in cost per approved application

100%

Dashboards owned by the client

Engagements

Three engagements, start to current

Client names, logos and exact URLs are withheld under NDA, so each engagement is referenced by number. Every figure comes from the client's own GA4, ad accounts, call logs and lending system. Hover or tap the footnote marker beside any metric to see exactly how it was defined.

Personal loans · NBFCEngagement 019 · anonymised

From brand-only rankings to a compounding organic channel

Lender type
RBI-registered NBFC, direct lending
Book size
₹640 Cr AUM at engagement start
Ticket size
₹50k – ₹5L, 12–48 month tenure
Engagement
Oct 2024 – Jul 2025 (10 months, ongoing)

Starting position

The lender ranked for its own name and almost nothing else. 91% of applications came from Google and Meta paid, and blended cost per approved application had risen from ₹1,410 to ₹2,060 across four quarters as competitors bid up the same keywords.

Baseline measured over Jul–Sep 2024 in the client's own GA4 property and LOS export, before any work started.

Measurement stack

  • GA4
  • Google Search Console
  • Google Ads
  • Looker Studio
  • In-house LOS export

What we did, in order

  1. Month 1

    Crawl and log-file audit: 3,100 of 4,700 URLs were parameter duplicates; the application funnel was blocked from indexing by a legacy robots rule.

  2. Months 2–4

    Technical rebuild — canonical and indexation fixes, LoanOrCredit and FAQ schema, LCP on the apply flow cut from 4.8s to 1.9s on 4G.

  3. Months 3–7

    42 bilingual intent pages published against eligibility, documentation, interest-rate and city queries, reviewed by the client's compliance lead before publishing.

  4. Months 5–10

    EMI and eligibility calculators shipped as indexable pages, and paid rebuilt to stop bidding on the 61 queries organic now held in the top three.

Baseline vs. current

Engagement 019: baseline versus current values for each reported metric, with footnote references.
MetricBaselineCurrentChange
Non-brand organic sessions / month8,24023,390+184%
Approved applications / month312487+56%
Blended cost per approved application₹2,060₹1,421−31%
Organic share of approved applications9%34%2.4×
Monthly media spend₹6.4L₹6.1L−5%

Metric notes & assumptions

  1. [1]Non-brand organic sessions from Google Search Console and GA4, averaged over the last three months of the period (May–Jul 2025) against the Jul–Sep 2024 baseline average. Brand-term exclusion regex was agreed with the client and unchanged throughout.
  2. [2]Approvals are credit-approved applications recorded in the client's loan origination system, deduplicated by PAN, excluding re-applications within 30 days. Disbursal counts run roughly 8–11% lower and are reported separately in the monthly dashboard.
  3. [3]Blended cost = (paid media spend + our monthly fee) ÷ total approved applications from all channels. Because organic approvals carry no media cost, this figure falls as organic share grows; a paid-only CPA over the same window moved from ₹2,060 to ₹1,884.
  4. [4]Attribution is GA4 last non-direct click for the digital funnel, reconciled each month against the LOS source field. Roughly 6% of approvals could not be attributed to a channel and are excluded from the share calculation, not from the totals.
  5. [5]Media spend is invoiced spend excluding GST. Spend was deliberately held roughly flat so that the change in cost per approval reflects funnel and channel-mix work rather than budget change.

+184%

Non-brand organic sessions in 10 months

−31%

Blended cost per approved application

2.4×

Organic share of approvals

Evidence & artifacts (NDA-safe)

Shown on a call, sanitised — never published with client identifiers

  • Month-1 technical audit (sanitised extract)

    What it measures: 4,700 crawled URLs, duplicate-parameter clusters, robots and canonical conflicts, and the indexation status of every funnel step.

    Sanitisation: Client domain, internal URLs and staff names removed; issue counts intact.

  • Looker Studio dashboard (demo clone)

    What it measures: Approvals, blended cost per approval, channel contribution and funnel-stage conversion, on the same layout the client sees.

    Sanitisation: Live clone shown on a call with dummied labels; no client data exported.

  • Search Console export, non-brand queries

    What it measures: Query-level clicks, impressions and average position for the 61 queries that reached the top three.

    Sanitisation: Brand-containing queries stripped; date ranges shown in full.

  • Baseline vs. current reconciliation sheet

    What it measures: GA4 sessions and conversions lined up against LOS approvals month by month, with the unattributed remainder shown.

    Sanitisation: Borrower-level rows removed; monthly aggregates only.

Request an artifact walkthrough

What didn't work

Two of the three long-form comparison hubs never ranked — the SERP was owned by aggregators with far stronger link profiles. We stopped that cluster in month 6 and moved the budget into calculators and city pages, which is where the compounding actually came from.

"The first month was mostly them telling us what was broken in our own site, which was uncomfortable and correct. What changed my mind was the monthly report reconciling against our LOS numbers instead of an ad dashboard."

Head of Digital, NBFC (name withheld under NDA)
Gold loans · Multi-branch lenderEngagement 023 · anonymised

Hyperlocal demand capture across 63 branches

Lender type
Gold loan NBFC, branch-led model
Footprint
63 branches across four states
Ticket size
₹25k – ₹2L, mostly 6–12 month tenure
Engagement
Feb 2025 – present (8 months reported)

Starting position

Footfall came from walk-ins and referrals. All digital spend sat in one national Meta campaign and a single generic 'gold loan' Google campaign, while 71% of real demand searched with a locality or 'near me' modifier and converted by phone — a path nothing in the setup could measure.

Baseline measured Nov 2024 – Jan 2025 across Google Business Profile insights and call logs from 63 branches.

Measurement stack

  • Google Business Profile
  • Call tracking
  • Meta Ads
  • Looker Studio
  • Client CRM

What we did, in order

  1. Month 1

    Audit found 19 branches with no Business Profile at all, 11 duplicates, and 26 with wrong hours or an unrouted phone number.

  2. Months 1–3

    Profiles claimed, deduplicated and standardised; per-branch review request flow handed to branch managers with a response SLA.

  3. Months 2–6

    63 branch pages plus 22 locality pages published with branch-level schema, per-gram rate transparency and directions blocks.

  4. Months 3–8

    Vernacular Meta creative split by region and ticket band, and dynamic call tracking numbers installed so branch calls landed in the CRM.

Baseline vs. current

Engagement 023: baseline versus current values for each reported metric, with footnote references.
MetricBaselineCurrentChange
Local pack impressions / month41,600129,8003.1×
Calls from Business Profiles / month1,1803,690+212%
Tracked branch enquiries / month2,2403,540+58%
Cost per qualified enquiry₹388₹295−24%
Average branch review rating3.64.4+0.8

Metric notes & assumptions

  1. [1]Nineteen branches had no profile at baseline, so part of this rise is new coverage rather than improved ranking. On the 44 branches that existed at baseline the like-for-like increase is 2.2×, which is the number we treat as the performance signal.
  2. [2]Measured as call-button interactions in Business Profile insights. Connected-call volume from the call tracking layer runs about 12% lower, mostly from taps outside branch hours; both series appear in the client dashboard.
  3. [3]An enquiry is a tracked call lasting over 45 seconds, a branch page form submission, or a redeemed walk-in coupon code, deduplicated by phone number on a 30-day window. Referral and repeat-customer walk-ins are excluded because they cannot be attributed to digital.
  4. [4]Cost = media spend plus our fee, divided by enquiries the branch manager marked qualified in the CRM. Qualification is a human judgement made at branch level, so this metric carries more variance than the volume metrics above.
  5. [5]Unweighted mean of the public rating across all 63 profiles at the end of the period versus baseline. Review volume grew from roughly 2,900 to 8,600, so newer reviews carry most of the weight in each branch's displayed rating.

3.1×

Local pack impressions

+58%

Tracked branch enquiries

−24%

Cost per qualified enquiry

Evidence & artifacts (NDA-safe)

Shown on a call, sanitised — never published with client identifiers

  • Branch profile audit matrix (sanitised)

    What it measures: Per-branch profile status, duplicates, hours accuracy, phone routing and category setup across all 63 locations.

    Sanitisation: Branch names replaced with codes (B-01…B-63); state-level grouping kept.

  • Call tracking summary export

    What it measures: Call volume, duration bands, missed-call rate and branch routing, by month and by state.

    Sanitisation: Caller numbers removed entirely; only counts and duration bands shared.

  • Review programme log

    What it measures: Review requests sent, reviews received, response time against the SLA, and rating movement per branch.

    Sanitisation: Reviewer names and review text excluded; ratings and counts shown.

  • Vernacular creative test board

    What it measures: Creative variants by language and ticket band with spend, enquiries and cost per enquiry for each.

    Sanitisation: Client logo and rate cards masked in the shared version.

Request an artifact walkthrough

What didn't work

Hindi creative lifted enquiries in two states and flatlined in a third, where borrowers responded better to the regional language. We had assumed Hindi would generalise across the north; it did not, and the fix cost us a month of testing.

"Nineteen of our branches did not exist on Google. We had been paying for a national campaign for two years without knowing that."

Regional Business Head, gold loan NBFC (name withheld under NDA)
MSME loans · Digital lenderEngagement 027 · anonymised

Fixing the funnel before spending more on traffic

Lender type
Digital-first MSME lender, co-lending model
Book size
₹210 Cr disbursed annually
Ticket size
₹2L – ₹25L, working capital
Engagement
Jan 2025 – Sep 2025 (9 months)

Starting position

Media spend had grown 40% in two quarters while approvals stayed flat. Channel reporting looked healthy because it stopped at 'lead'. The real loss sat between form start and GST document upload, where 68% of applicants dropped out and were never contacted again.

Baseline measured Oct–Dec 2024 using event-level funnel data instrumented in week 2 of the engagement, backfilled from server logs.

Measurement stack

  • GA4
  • Google Ads
  • Meta Ads
  • WhatsApp Business API
  • Looker Studio
  • Client LOS

What we did, in order

  1. Month 1

    Click-to-disbursal instrumentation: 11 funnel events defined and reconciled against the lending system, exposing the document-upload cliff.

  2. Months 2–4

    Single 23-field form rebuilt into four saved steps with resume links, mobile document capture and inline eligibility feedback.

  3. Months 3–7

    Opt-in WhatsApp and email recovery journeys for stalled applications, triggered at 2h, 24h and 72h after drop-off.

  4. Months 4–9

    Search terms, placements and audiences pruned against approved-application data instead of lead volume — 214 negative keywords added.

Baseline vs. current

Engagement 027: baseline versus current values for each reported metric, with footnote references.
MetricBaselineCurrentChange
Application completion rate32%47%+46%
Approved applications / month141236+67%
Cost per approved application₹4,880₹3,020−38%
Applications recovered by journeys / month0218new
Monthly media budget₹11.5L₹11.5L0%

Metric notes & assumptions

  1. [1]Completion rate is applications reaching document submission ÷ applications started, measured on event-level data. The move is 32% to 47% — a 15 percentage-point gain, or +46% in relative terms. Both framings appear in the client's dashboard.
  2. [2]Credit-approved applications in the lending system. The client confirmed in writing that underwriting policy and score cut-offs were unchanged between the baseline and reporting periods, so the rise is funnel volume and quality rather than looser approval.
  3. [3]Cost = (media spend + our fee) ÷ approved applications. Media was held at ₹11.5L a month by agreement, which is why this metric moves almost entirely with approval volume.
  4. [4]Recovered applications are stalled applications that resumed within 7 days of an opt-in nudge and reached document submission. Where a borrower also clicked a paid ad in that window, the application is credited to the journey only if the resume link was the last interaction.
  5. [5]Budget was intentionally frozen for the engagement. This is the control that makes the cost-per-approval movement meaningful; scaling spend afterwards would change the figure in both directions.

+46%

Application completion rate

−38%

Cost per approved application

₹0

Added to monthly media budget

Evidence & artifacts (NDA-safe)

Shown on a call, sanitised — never published with client identifiers

  • Funnel event map (sanitised)

    What it measures: All 11 events from ad click to disbursal, with the drop-off percentage at each step before and after the rebuild.

    Sanitisation: Internal event names generalised; client property IDs removed.

  • Form rebuild before/after wireframes

    What it measures: The 23-field single form against the four-step saved flow, with field-level abandonment marked.

    Sanitisation: Client branding stripped; layout and field logic preserved.

  • WhatsApp journey spec and opt-out log

    What it measures: Message timing, template copy, consent capture points, opt-out rate before and after the sequence was shortened.

    Sanitisation: Borrower identifiers removed; only aggregate rates shared.

  • Negative keyword and audience prune list

    What it measures: The 214 terms removed, with spend and approval counts that justified each removal.

    Sanitisation: Shared as a term-level list without account IDs or client-specific bids.

Request an artifact walkthrough

What didn't work

The first WhatsApp sequence sent three messages in 24 hours and pushed opt-outs to 9%. Cutting it to one message plus a single reminder recovered nearly the same volume of applications with opt-outs back under 2%.

"We were about to sign off another ₹5L a month in spend. The instrumentation work showed the problem was on page three of our own form."

Co-founder, MSME lending platform (name withheld under NDA)

Verification checklist

How to validate an anonymised case study

Anonymised results deserve scepticism — including ours. Use this checklist on us, and on every other agency you shortlist. Each point is something we can answer on a single call.

  1. 01

    Check the reporting window against the baseline window

    Every figure names both periods. Ask whether the baseline is a single month or an average — ours are three-month averages, which hides less. A case study that names only an end date is hiding the start.

  2. 02

    Ask which system the conversion number came from

    Ad platforms count leads; loan origination systems count approvals. Ask us to show which system each number was pulled from, and where the two disagree. We report the LOS number when they differ.

  3. 03

    Test the cost metric definition

    Ask whether cost per approval includes agency fees, GST, and organic conversions. Ours includes fees, excludes GST, and counts all approvals including organic — which is a harder number than a paid-only CPA.

  4. 04

    Look for the like-for-like number

    Growth from new coverage is not the same as growth from better performance. Where new branches or new pages inflate a figure, we publish the like-for-like comparison next to it — ask for it if you do not see it.

  5. 05

    Ask what did not work

    Every engagement has abandoned experiments. Each case study above names one. If an agency cannot tell you what they stopped doing and why, the numbers are marketing, not measurement.

  6. 06

    Ask to see the live dashboard, not a screenshot

    On a call we screen-share a working Looker Studio dashboard with dummied labels so you can change date ranges yourself. Static screenshots are the easiest artifact in the world to fabricate.

  7. 07

    Ask for a reference conversation

    Where the client has agreed in advance, we introduce you directly. NDAs stop us naming clients unprompted; they do not stop a client volunteering to vouch for the work.

Book a verification call

30 minutes, screen-shared dashboard, no pitch deck.

Methodology

How we measure

One metric of record

We agree on your blended cost per approved application before work starts, and report against it every month.

Attribution you can audit

GA4, call tracking, and CRM or LOS stages wired together so numbers reconcile with your own systems — not just an ad dashboard.

No vanity reporting

Impressions and rankings appear as inputs. Approvals, cost, and channel contribution lead the report.

What we will and won't share

We will screen-share a live Looker Studio dashboard with dummied labels, show the date ranges each baseline was taken from, walk through the sanitised artifacts listed under each engagement, and arrange a reference conversation where the client has agreed in advance. We will not publish client names, logos, domains, screenshots containing identifiers, or any borrower-level data.

Ask for a reference call

Figures are actuals from the engagements described, rounded, and reported for the periods stated, under the definitions given in each metric's footnote. Past performance in other engagements does not guarantee future results. Outcomes depend on your product, pricing, underwriting, market, and budget.

Next step

See what your own numbers could look like.

The Growth Audit gives you a channel-by-channel gap analysis and a realistic view of the upside before you commit to anything.

  • Free of charge
  • Delivered in 5 working days
  • No lock-in