Sudharsana.
← Back to selected work

Professional work · Data engineering · attribution · business delivery

From a new channel to a number people trust

An affiliate data pipeline built end to end, with attribution agreed before it was coded.

Walmart Global Tech · acquisition & attribution

Follow one click

A partner click travels from the channel to the view the business reads. Pick a day and watch where it goes, and where it stops.

A click passes through six stages: channel, ingest, dedupe, join to orders, attribution and a freshness check, then reaches a channel view of daily results. When a partner file is late, the freshness check holds today's value. A second view shows how last touch and first touch credit the same three step journey differently.ChannelclicksIngeston timeDedupeuniqueJoin ordersmatchedAttributionlast touchFreshnesscurrentChannel viewMonTueWedThuFriSatTodayheldLate file: view heldOne customer, one orderSearch0%Email0%Affiliate100%Orderrulelast touch
Credit rule

Clicks become a number someone trusts

Partner clicks arrive, duplicates fall out, and each click is joined to the orders that followed it. Attribution assigns the credit, and today’s bar joins the channel view.

The pipeline’s job ends where a person reads the result. That is where it gets judged.

Hold the view, raise the flag

One partner’s file has not arrived. Publishing anyway would draw a quiet dip that looks like a bad day for the channel. The freshness check holds today’s bar and tells the team the data is late.

A missing number is easier to act on than a wrong one.

Agree on the rule before the code

This customer searched, opened an email, then clicked an affiliate link and ordered. Under last touch, the affiliate click earns the credit. That rule was agreed with sales leadership before anything was built.

Switch the rule to first touch below and watch the same order change hands.

Illustrative data. The stages and checks reflect the pipeline; the clicks, days and journey are made up.

new customer visits · first production year
4.9M
GMV · first production year
$5.4M

A new channel, no data path

A new acquisition channel brings a practical question. How does a partner’s click become something the business can read on Monday morning?

I designed, built, and deployed the affiliate marketing pipeline end to end, from partner activity to the B2B data platform.

What it carried

In its first production year, the channel brought 4.9 million new customer visits and $5.4 million in GMV. Those outcomes are why the pipeline mattered: they let the business judge a channel it had just started paying for.

The decision: agree on credit first

A customer often touches several channels before ordering. Who gets the credit sounds technical, but it is a business decision. I partnered with sales leadership to define last-touch attribution across the B2B customer base before building it, so the number on the dashboard meant the same thing to everyone reading it.

Play “Who gets the credit?” above and switch the rule. The same journey tells a different story.

Late is not low

Partner data arrives on its own schedule. A file that lands late can make a healthy day look like a bad one. My platform work covers data quality, freshness, monitoring, and incident response, and it shapes one habit here: a late file should be flagged as late, not published as a dip.

A number is only useful when everyone reads it the same way.

Let’s talk ↗