AK

Ankit
Khareshiya

Growth marketer. I find out what is actually driving growth, and measure what isn't.

Noida, India · remote ankitkha26@gmail.com LinkedIn
2,729outbound emails tracked in one dashboard
32sending domains placement-tested weekly
90mailboxes on one warmup standard
9case studies: 6 from live programmes, 3 practice builds

I audit analytics, trace where traffic and leads really come from, and fix the tracking that hides it. My work so far is in B2B outbound: lead pipelines, campaign data, and the email infrastructure underneath them. Most of what I build gets automated so the reporting runs without me.

Selected work

Reporting

Service offers opened at 55% vs 32% for tool offers

55% vs 32%open rate, service offers vs tool offers
6campaigns across 4 industries, Aug–Oct 2026
54% vs 24%ecommerce, the widest gap of any segment

Problem

Open rates for six cold-email campaigns sat in a wide sheet, one column per send day, updated weekly by a Python script. Nothing in that layout let you compare offers or industries.

Method

A reshape of the sheet into one row per campaign per send day (204 rows) with segment, offer and status split out, feeding a public Data Studio dashboard: average open rate by offer, by industry and offer, and a weekly trend.

Result

Service offers opened at 55% against 32% for tool offers. The gap held in every industry that ran both: fintech 64% vs 41%, ecommerce 54% vs 24%.

Recommended next: lead tool-offer emails with the service angle, re-measure opens

Data Studio, Google Sheets, Python · View live dashboard →

Social analytics

Median views fell 45 → 6 as posting rose 35 → 64 posts

45 → 6median Instagram views per post, Jan–May vs Aug–Oct
88%of Jan–May Facebook and Instagram views came from Reels
+340%Facebook Group member views, last 60 days
Reels, Jan–May101.5Carousels, Jan–May42Photos, Jan–May26Carousels, Aug–Oct4Photos, Aug–Oct2
Median views per post by format · Facebook Page and Instagram, Jan–Oct 2026, 121 posts

Problem

A B2B SaaS company posted across a Facebook Page, a Facebook Group, Instagram and two LinkedIn accounts, but the weekly tracking sheets only logged what went out. Nobody could say whether more posting was reaching more people.

Method

Pulled per-post views, reach and interactions from Meta Business Suite, group activity from Group Insights and LinkedIn engagement from Zoho Social, reshaped them into one row per post, and built a public Data Studio dashboard comparing periods, formats and platforms.

Result

Output went up (64 posts in Aug–Oct against 35 in Jan–May) while reach per post fell. The cause was format: Reels drove 88% of earlier views and none were posted after May, so static photos and carousels took median Instagram views from 45 to 6. The Facebook Group grew, and on LinkedIn the founder's profile earns about 15× the engagement per follower of the company page.

Recommended next: 1–2 Reels a week, re-measure in 30 days

Meta Business Suite, Zoho Social, Python, Data Studio · View live dashboard →

Reporting

One self-updating view of 2,729 emails across 6 campaigns

2,729emails tracked across 6 campaigns
141automated data pulls, twice a day
9.6%bounce on one campaign, ~3× the average

Problem

A B2B client ran six lemlist campaigns with no single view of performance. Numbers sat campaign by campaign inside lemlist, and averaging daily rates overstated quiet days.

Method

A GitHub Actions job that pulls lemlist stats into Google Sheets twice a day, feeding a Tableau dashboard: KPI tiles for leads, sent, delivered, open, reply and bounce, a campaign comparison, and trends by date. Open and reply rates are weighted by volume sent, and lead totals use a fixed-level calculation so days don't double count.

Result

Twelve weeks of data, July to September: 46% open, 1.3% reply, 3.5% bounce overall. By campaign, the three sender-named campaigns replied at 1.2–2.1% against 0.4–1.1% for the three industry campaigns, and Healthcare bounced at 9.6%.

Recommended next: re-verify the Healthcare list before its next send

Tableau, LOD calculations, GitHub Actions, lemlist, Google Sheets

Analytics

0 → 8 key events in 30 days, while sessions fell 23%

0 → 8book_demo key events from 3 users; none were tracked before
0 → 10UTM-tagged social sessions; 1 led to a key event
286 → 220sessions, −23%; marketplace referrals fell 42 → 3
Deal marketplace423−39Google organic4431−13Social, UTM-tagged010+10AI assistant (ChatGPT)02+2
Sessions by source · grey: 7 Sep – 6 Oct, orange or teal: 7 Oct – 5 Nov 2026 · GA4, all sources 286 → 220 sessions

Problem

A B2B SaaS product site had GA4 installed but nobody reading it. Over two months it logged 866 sessions and zero key events, so there was no way to tell which channel brought buyers. Event data was set to expire after two months.

Method

A channel and source breakdown of all traffic, and an exploration of outbound clicks. Thirty days after the fixes, a comparison of 7 Oct – 5 Nov against the 30 days before, by source, landing page and key event.

Result

19 demo-booking clicks had gone untracked, five social channels sent about 10 sessions in 90 days, and 59% of traffic arrived as Direct. Shipped a book_demo key event on the demo-booking click, event retention raised from 2 to 14 months, and UTM-tagged social profile links. Testing the key event end to end exposed a broken demo-booking flow, flagged the same day.

30 days later

Tracking worked: book_demo logged 8 events from 3 users, and tagged links brought 10 social sessions (Instagram 4, LinkedIn 4, Facebook 2), one ending in a key event. Traffic fell: sessions 286 to 220, users 215 to 186. Marketplace referrals (42 to 3) and Google organic (44 to 31) account for 52 of the 66 lost sessions. With the marketplace, the audit's most engaged source, nearly gone, engagement rate fell from 34.3% to 24.6% and pricing-page landings from 19 to 6.

Recommended next: fix the booking flow, find out why marketplace referrals stopped

Google Analytics 4, GA4 Explorations, key events, UTM tagging

Behaviour analytics

87% of visitors left from the homepage

87%of homepage visitors leave from the homepage
52%leave before scrolling 20% of the page
3%of desktop views click the hero CTA
Homepage visits165Left from homepage144Reached pricing44Left from pricing37
Where sessions end · Microsoft Clarity, last 30 days, 198 sessions
5%10%15%20%25%30%40%50%55%Visitors still on page, by scroll depth48% at 20%
Homepage scroll depth · 88 desktop page views

Problem

The Clarity project everyone checked showed zero sessions. The site's tag pointed to a different Clarity project, so 30 days of behaviour data sat unread.

Method

Matched the site's tag to the live project, then traced exits, scroll depth and click heatmaps.

Result

The homepage is the drop-off: most visitors leave before the first proof section, and the hero email form barely gets used. Recommended replacing the hero email form with a single "Start free trial" button and A/B testing it for 30 days, measuring CTA clicks per homepage view.

Recommended next: A/B test a single-CTA hero for 30 days

Microsoft Clarity, heatmaps, scroll maps, exit-page analysis

Experimentation

50.7% vs 44.7% opens, too close to call

50.7% vs 44.7%open rate, A (control) vs B, last 30 days with both live
p = 0.22two-proportion z-test, z = −1.23, not significant
207 / 390sends per variant vs the 390+ planned (B: 206)
Intervals overlapA · control50.7%B · new subject44.7%30%40%50%60%
Step 3 open rate, dot = observed rate, line = 95% confidence interval · B2B fintech cold-email campaign, last 30 days with both variants live, 413 sends

Problem

Step 3, the follow-up email in a B2B fintech cold-email campaign, had run on one subject line. The question was whether a new subject line would lift opens.

Method

Split Step 3 sends between the current subject line (A, the control) and a new one (B). B only ran in the last 30 days, so I compared that window alone, when both were live, rather than A's full 90 days. The gap was tested with a two-proportion z-test against a plan of 390+ sends per variant.

Result

A opened at 50.7% (105 of 207), B at 44.7% (92 of 206): 6.1 points lower, 12% relative, with z = −1.23 and p = 0.22. Not significant. Across A's full 90 days (367 sends, 47.4%) the gap was smaller still, p = 0.53. Neither variant reached 390 sends; confirming a 6-point gap would take about 1,090 per variant.

Learning

A 6-point gap looks decisive in a report. At this sample size it was noise, so I did not call a winner on it.

Decision: no winner, keep A (control)

Cold-email platform A/B split, two-proportion z-test, sample-size planning

Practice builds

Built on public or generated data to practise lifecycle, automation and segmentation. No client data used.

Segmentation · real public data

RFM segments predicted who bought again

68.5%

of revenue came from the top 22.7% of customers. Scored at June 2011, six-month repurchase was 90.2% for Champions and 20.9% for Hibernating. 1.07M rows cleaned in SQL, including 22,523 duplicated rows from overlapping sheets. 815 customers on the win-back list; about 41% came back with no email, so the test needs a holdout.

SQL, Python, pandas · UCI Online Retail II (CC BY 4.0)

Automation · generated data

A lead pipeline that survived a flaky CRM

71 / 71

invalid and duplicate leads caught out of 500 submissions. 93 CRM errors absorbed by retries, 1 lead lost after retries and recovered by replay. Alerts, dead-letter table, runbook, 10 tests. The CRM is a local mock.

Python, webhooks, REST API, SQLite

Lifecycle · generated data

An 8-email trial flow, built and QA'd

8 / 8

emails pass a QA gate for UTMs, unsubscribe link, size and alt text. Splits on one activation event; stops on payment or unsubscribe. A 5-point open-rate lift needs 1,534 recipients per variant. Not sent to real users, so no open or conversion results.

Python, HTML email, n8n, UTM tagging

Also shipped

Automation

Automated the weekly inbox placement test

40 → 9 min

Hands-on time per run, on one Playwright script covering 32 domains across three clients.

Python, Playwright, seed mailboxes

Pipeline

Built the outbound lead pipeline from nothing

27%

Reply rate: 16 qualified leads in four weeks, 11 contacted, 3 replied, with zero sales tooling.

Claude in Chrome, Google Sheets, Apps Script, Unibox

Execution

Rolled out a 30-domain campaign, deliverability-checked first

150 / 150

Gmail seeds landed in the inbox across 30 domains before launch.

Woodpecker, Instantly, deliverability audit tooling

Monitoring

Caught live domains slipping into spam, and rotated in the backup

60 / day

Full volume restored on the backup domain after the weekly placement test caught the drop.

Inbox placement testing, Instantly

Content operations

Ran five social channels as the only person on content

21

Posts a week planned across 5 channels, scheduled ahead through 16 Oct.

Zoho Social, Meta Business Suite, Canva

Also: authentication for six new sending domains (2 live DNS faults fixed at publish) · warmup runs for six clients' sending accounts · warmup standardised across 90 mailboxes.

Field notes

Deliverability · 6 min read

The cold email deliverability checklist I run before every launch

Domains, authentication, warmup, lists, placement testing and weekly monitoring, with the checks that caught real problems.

Read →

How I work

  1. 01

    Audit the tracking

    Before trusting any number, I check it is real: tags firing, key events defined, sources tagged, data kept long enough to compare.

  2. 02

    Find the real driver

    I split results by offer, format, channel or campaign until one variable explains the gap, then weight it by volume so quiet days don't skew it.

  3. 03

    Automate the reporting

    Scheduled scripts feed a live dashboard, so the team sees the number every week without asking and I can move to the next question.

About

I make email-led growth measurable, then make the measuring run itself.

Most of my work starts with a number nobody trusts: an analytics property with no key events, a dashboard averaging the wrong way, a channel that looks dead because its traffic lands in Direct. I fix the tracking, find what is actually moving opens, replies and pipeline, and automate the reporting so the team sees it every week without asking.

I work across campaigns, deliverability and data for B2B outbound programmes, so I know what can break between a sequence launching and a reply arriving, and I check that first.

Focus
Lifecycle, email-led growth, marketing ops
Strongest at
Analytics audits, campaign reporting, deliverability
Builds with
Python, Apps Script, GitHub Actions, browser agents
Based in
Noida, India · works remote
Looking for
Growth, lifecycle and marketing ops roles

What I work in

Analytics

  • Google Analytics 4
  • SQL
  • Microsoft Clarity
  • Tableau
  • Data Studio
  • Campaign reporting

Automation

  • Python, Playwright
  • Browser agents
  • Google Apps Script
  • GitHub Actions

Sending

  • Instantly
  • Woodpecker
  • lemlist
  • Mailcow
  • Google Workspace, Microsoft 365
  • Campaign setup and monitoring

Deliverability

  • SPF, DKIM, DMARC
  • Domain and mailbox warmup
  • Inbox placement testing
  • Blacklist and reputation monitoring
  • Google Postmaster Tools
  • DMARC reporting (TDMARC)

Hiring a growth marketer?

I am looking for growth marketing roles. If you need someone to find what is really driving your numbers, and fix the tracking that hides it, I would like to hear from you.