Revenue driver

Link requests went from days in an Ad Ops queue to under 60 seconds

Affinity CPA publisher portal · Associate UX Designer · Solo design with Eng and Ad Ops · 2026

< 60 sec

Programs table to live tracking link

by design

Ad Ops steps in the request path

design fact

4 levels

Drill path: Daily › Program › Geo › Sub ID

design fact

Browse Programs table. Get Link opens a side panel with the tracking link ready to copy, then confirms Copied to clipboard. — Programs table to copied tracking link. Prototype screen, illustrative data.

The problem

Publishers emailed Ad Ops for every tracking link and waited a day or more. Until it arrived, they could not promote a program or earn from it.

The gap was not a missing dashboard. It was the wait between a publisher’s intent and their first tracked revenue.

Two questions shaped the scope.

Question 1

How do I start earning?

Question 2

Can I trust what I’m paid?

Link generation

How do I start earning?

Decision 01

Compare every program in one table, then get the link from the row

Tension: Advertisers want distinct brands. Publishers need payout, geo and EPC side by side.
Rejected: A branded page per advertiser, which means a dozen tabs to compare.
Result: Three clicks: Browse Programs, Get Link, copy.

Browse Programs table with payout, geo, deep link, EPC, category and status columns. Each row has a Get Link button; the paused program’s button is disabled. — 1Get Link in the row2A dash, not a zero, when EPC has no historyPrototype screen, illustrative data.

Decision 02

The link is ready before the options

Tension: Power publishers need Sub-ID attribution. Casual ones need only a link.
Rejected: A raw query-string editor.
Result: The working link is copyable the moment the panel opens. Click ID, deep link and eight Sub-IDs sit below as optional.

Get link side panel for a program: the tracking link with a copy button, then an optional click ID checkbox, a deep link field and eight Sub-ID fields, ending in a Copy tracking link button. — 1Link ready on open2Optional: Click ID · deep link · 8 Sub-IDsPrototype screen, illustrative data.

Reports

Can I trust what I’m paid?

Decision 03

Drill down instead of rebuilding the filter

Tension: “Earnings are down” is a nested question: day, program, geo, Sub ID.
Rejected: Four separate reports, and every column shown by default.
Result: One table with a visible drill path and Clear path. Seven columns by default, the rest in Customize Columns. A pinned totals row states how many rows it sums.

Daily Stats drilled to Sub ID. A drill path reads Daily 06-Sep, Program HealthPro Premium Supplements, Geo United States, Sub ID, with a Clear path button. Below it, a totals row and per-Sub-ID rows. — 1Drill path: Daily › Program › Geo › Sub ID2Totals row states its row countPrototype screen, illustrative data.

Decision 04

Show restated payouts beside the original

Tension: One conversion carries 21 columns of facts, and trust breaks when a restated number silently replaces the old one.
Rejected: Charts first, and overwriting values in place.
Result: A row-level ledger with Conversion Date and Program frozen. Original and current earnings sit side by side.

Conversion Reports scrolled to the money columns: Total earnings and Original earnings, Total sale value and Original sale value, side by side. Conversion Date and Program stay frozen on the left. — 12 frozen columns of 212Original vs current: €36.84 → €34.10Prototype screen, illustrative data.

Also shipped

Payout-change feed

A rate cut reads $24.00 → $18.00 (−25%) before traffic is sent.

Dashboard KPIs

Every figure shows its approved, pending and declined split.

What I did not build

A full analytics suite with charts first. The bottleneck was the wait for a link, and the open question afterwards was trust in payments. So reporting shipped as two purpose-built reports instead of a general dashboard.

Result

< 60 sec

Programs table to live tracking link

by design

Ad Ops steps for standard link requests

design fact

20+

Programs in the portal

v25 prototype

Read the long version →

Next: ROAS target case study →