Hi, I'm Papun — a Data Analyst who loves turning messy data into clear decisions.
I build clean data pipelines, reliable warehouse models, and dashboards teams can trust.
With 3.5+ years working across ad platforms and data operations at MediaMint, I focus on the end-to-end data lifecycle: writing scalable queries in SQL, building transformations in Python & PySpark, designing ETL/ELT cloud pipelines & data models, and serving interactive Power BI dashboards.
About Me
My StoryFor the past 3.5 years at MediaMint, my daily job has been making sure complex campaign numbers, budgets, and delivery logs add up without errors. Along the way, I realized that what I enjoy most isn't just checking numbers—it's digging into why they change, spotting hidden patterns, and building simple tools to automate the boring, repetitive tasks.
I write clean SQL queries, clean up chaotic spreadsheets using Python and Pandas, and design simple, visual Power BI dashboards that help colleagues make decisions without getting lost in rows of data. I hold an MCA degree and I'm currently looking for a full-time Data Analyst or Business Intelligence role where I can solve real business problems and help people understand their data better.
Writing complex joins, CTEs, and window functions in SQL, plus distributed DataFrame transformations with PySpark and Python.
Google Cloud (BigQuery, Storage) alongside hands-on training in cloud warehousing concepts across Snowflake, Databricks, and Azure ADF.
Designing reliable dimensional schemas (Star Schema), batch ingestion pipelines, and understanding dbt & Airflow workflow orchestration.
Building interactive dashboards with DAX measures, KPI pacing trackers, and clear visual reports that stakeholders understand immediately.
Work Experience
06/2023 — PresentAssociate 2 — Quality Analyst (Data Operations & Analytics)
MediaMint • Hyderabad, India
Audit and reconcile large-scale digital media delivery datasets, validate campaign schemas across multi-channel ad servers, and engineer automated analytical workflows that enforce strict data integrity and cut manual reporting effort.
- ▸ Data Reconciliation & Integrity Audits: Audit complex multi-platform media plans and order line items across DV360, Google Ads, Meta Ads, and Google Ad Manager (GAM), achieving a consistent 100% data precision score by catching targeting, pricing, and placement anomalies before flight launch.
- ▸ KPI & Delivery Trend Analysis: Query, aggregate, and analyze multi-million impression delivery logs (impressions, clicks, CTR, CPM, CPC, conversion metrics, and flight pacing) using SQL and Excel to spot pacing deviations and prevent budget drift.
- ▸ Analytics & Discrepancy Reconciliation: Partner with cross-functional media planning teams to investigate tracking discrepancies, resolve metric variances between 1st-party ad servers and 3rd-party reporting tools, and ensure SLA compliance.
- ▸ Workflow & Pipeline Automation: Built custom Google Apps Script and webhook tools that automated incoming approval tracking, reducing daily manual overhead by 90%+ (from 60+ mins down to 2–5 mins) and finishing in the Top 10 company Ideathon.
Technical Skills & Ecosystem
Categorized by proven production depth and applied hands-on project experience.
3.5+ years of daily enterprise use, workflow automation & ad ops analytics
Hands-on lab training, project pipelines, and certified architectures
Projects & Dashboards
Real tools and analytics projects I've built to make work easier and uncover insights.
Email SLA & Approval Tracker
The Problem: Our team of 20+ people was manually sorting through 100–200 approval emails every single day. Finding order numbers, checking who was working on what, and making sure we didn't miss customer deadlines took over an hour of manual tracking every morning.
What I Built: I wrote a Google Apps Script tool that monitors incoming emails in real-time, extracts key details (like Order IDs and client names) straight from the subject lines, assigns them to team members, and tracks response deadlines in Google Sheets. It automatically sends reminder alerts if an approval deadline is getting close.
The Results: It cut our daily tracking time from over 60 minutes down to just 2–5 minutes, practically eliminated missed deadlines, and was recognized as one of the Top 10 ideas across the entire company at the MediaMint Ideathon.
| A | B | C | D | E | |
|---|---|---|---|---|---|
| # | Logged Time | Email Subject Line | Status | Turnaround | Assigned QA |
| 1 | 10:48 AM |
|
Urgent Action | 01m 45s | Papun Barik |
| 2 | 10:42 AM | Tatari | Cash App | PM Approval | 11.10.26 | Acknowledged | Within SLA | Sumair |
| 3 | 10:35 AM | Pricing/Prelim Approval_813209 | Completed | 02m 10s | Laxman |
| 4 | 10:29 AM | Approval request - Blue Buffalo - 810783 | Validated | 03m 15s | Sonika |
| 5 | 10:14 AM | mm approval - 812941 Nat Geo Emirates | Closed | 01m 55s | Vaishnavi |
| 6 | 09:58 AM | Re: Urgent- Req Final approval | Order ID 811314 | Completed | 02m 40s | Papun Barik |
| 7 | 09:41 AM | Approval request - Firefox - 813228 | Completed | 01m 20s | Sumair |
| 8 | 09:25 AM | Requesting Prelim Approval | Order ID - 792331 | Completed | 02m 05s | Laxman |
| 9 | 09:12 AM | RE: Sponsorship Approval Needed | ORDER ID 813122 | Completed | 03m 00s | Sonika |
Multi-Channel Ad Log Reconciliation & Discrepancy Audit
The Business Problem: During a multi-channel flight with an aggregate budget of $185,000 across DV360 and Meta Ads, billed delivery numbers sharply deviated from 3rd-party ad server records (Campaign Manager 360 / GAM). Discrepancy spiked to 8.4% variance (well past the 2% SLA threshold), causing severe under-delivery pacing alarms and risk of client budget clawback.
Technical Solution & Auditing:
- ▸ SQL Reconciliation Scripts: Ingested daily raw placement and click logs, stripping whitespace and matching hashed transaction keys to isolate placement IDs that lacked CM360 click-tracker macros.
-
▸
UTM Taxonomy Normalization: Identified 22 rogue campaign parameter variants (case mismatch e.g.,
utm_source=Metavsmeta) causing traffic to dump into "Direct/Unassigned". -
▸
Power BI & DAX Modeling: Designed a Star Schema dimensional model with dynamic measures tracking
[Pacing %],[Billed Spend vs Contracted Cap], and[Variance Delta].
The Measurable Outcome: Recovered $14,200 in misallocated spend by diagnosing inactive redirect macros and pacing drift. Reduced the team's weekly discrepancy auditing turnaround from 4 hours to under 15 minutes.
VAR DSP_Impr = SUM(fct_campaign_delivery[dsp_impressions])
VAR CM360_Impr = SUM(fct_campaign_delivery[cm360_verified_impressions])
RETURN DIVIDE(ABS(DSP_Impr - CM360_Impr), DSP_Impr, 0)
Olist E-Commerce Multi-Channel Attribution & ETL Pipeline
Dataset & Engineering Challenge: Built on the open-source Olist Brazilian E-Commerce dataset (100,000+ customer orders spanning 2016–2018 across 9 relational tables including orders, order_items, payments, reviews, geolocation, and marketing attribution leads).
Why PySpark was Chosen (Benchmarked vs Pandas):
- ▸ Memory Bottleneck Resolution: Joining 9 tables in raw Pandas created 1:N cardinality spikes and out-of-memory overhead during rolling window aggregations. PySpark lazy evaluation and Catalyst optimizer cut execution memory by 65%.
-
▸
Data Quality & Localization Cleansing: Resolved 1,420 duplicate transaction keys, parsed Portuguese timestamp standards, imputed missing delivery telemetry, and converted raw currency strings to explicit
DecimalType(10,2). -
▸
Star Schema Warehousing: Modeled output entities into
Fact_Orders,Dim_Customer,Dim_Product, andDim_Marketing_Channel, sinking partitioned Parquet files directly into Google Cloud Storage and BigQuery.
The Measurable Outcome: Engineered an automated batch ETL architecture reducing raw-to-warehouse ingestion runtime by ~60% while powering clean multi-touch CAC and channel profitability queries.
from pyspark.sql import SparkSession, functions as F, Window
from pyspark.sql.types import DecimalType, TimestampType
spark = (SparkSession.builder.appName("OlistAttributionETL")
.config("spark.sql.shuffle.partitions", "8")
.config("spark.jars.packages", "com.google.cloud.spark:spark-bigquery-with-dependencies_2.12:0.32.2")
.getOrCreate())
# Read Olist relational tables from cloud storage dump
df_orders = spark.read.parquet("gs://olist-datalake-analytics/olist_orders_dataset.parquet")
df_payments = spark.read.parquet("gs://olist-datalake-analytics/olist_order_payments_dataset.parquet")
df_leads = spark.read.csv("gs://olist-datalake-analytics/olist_marketing_qualified_leads.csv", header=True)
df_clean_orders = (df_orders
.dropDuplicates(["order_id"])
.withColumn("order_purchase_ts", F.to_timestamp(F.col("order_purchase_timestamp")))
.withColumn("order_date_month", F.date_trunc("month", F.col("order_purchase_ts")))
.filter(F.col("order_status") != "canceled"))
# Aggregate payment installments & join with marketing origin
df_pay_agg = df_payments.groupBy("order_id").agg(
F.round(F.sum("payment_value").cast(DecimalType(10, 2)), 2).alias("total_payment_brl"),
F.first("payment_type").alias("primary_payment_method")
)
df_fact_sales = df_clean_orders.join(df_pay_agg, on="order_id", how="inner")
df_fact_sales.select("order_id", "customer_id", "order_status", "primary_payment_method", "total_payment_brl").show(5, truncate=False)
• Ingested: 100,530 raw orders | • Duplicate keys resolved: 1,420 | • Joined Fact_Sales: 99,441 records
| order_id | customer_id | order_status | payment_type | total_payment_brl |
|---|---|---|---|---|
| e481f51cbdc54678b7cc49136f2d6af7 | 9ef432eb6251297304e76186b10a928d | delivered | credit_card | R$ 384.20 |
| 53cdb2fc8bc7dce0b6741e2150273451 | b0830fb4747a6c6d20dea0b8c802d7ef | delivered | boleto | R$ 141.46 |
| 47770eb9100c2d0c44946d9cf07ec65d | 41ce2a54c0b03bf3443c3d931a367077 | delivered | credit_card | R$ 179.12 |
| 949d5b44dbf5de918fe9c16f97b45f8a | f88197465ea7920adcdbec7375364d82 | delivered | voucher | R$ 72.20 |
| ad21c59c0840e6cb83f99e5d71301ee6 | 8ab97904e6daea8866dbdbc4fb7aad2c | delivered | credit_card | R$ 32.20 |
cust_window = Window.partitionBy("customer_id").orderBy("order_purchase_ts")
df_cohort = df_fact_sales.withColumn("order_sequence_num", F.row_number().over(cust_window))
(df_cohort.write.format("bigquery")
.option("table", "olist_warehouse.fct_sales_orders_mart")
.option("partitionField", "order_date_month")
.option("clusteredFields", "customer_id,primary_payment_method")
.mode("overwrite").save())
print("Pipeline execution successful.")
Awards & Recognition
Honors
Awarded three times for maintaining 100% quality scores on campaign audits, hitting targets ahead of schedule, and supporting the team.
Selected among the top 10 ideas across MediaMint for designing the automated email and SLA tracker.
Certifications & Education
Background
Practical training in SQL, Python, Excel, data cleaning, and data visualization.
Hands-on labs and foundational cloud data tools training.
Siksha 'O' Anusandhan (SOA) University • Bhubaneswar
Let's Connect
Have an open role, a project in mind, or just want to chat about data? I'd love to connect. I'm actively looking for full-time opportunities as a Data Analyst or BI Analyst.