Connecting the dots
“You can’t connect the dots looking forward; you can only connect them looking backward.”
— Steve Jobs
Hardi Jain · AI and SaaS product manager
B2C · STORY 01
SRT optimization through bot detection
A platform serving more than 10 million users saw its Success Rate of Transactions fall from 65% to 60%.
01 · THE SIGNAL
A platform serving more than 10 million users saw its Success Rate of Transactions fall from 65% to 60%.
AWS monitoring flagged the deviation once it passed two percentage points.
02 · INVESTIGATION
I segmented the decline across merchant level, industry segment and transaction source to identify where the SRT movement was concentrated.
03 · THE FIRST HYPOTHESIS
My initial hypothesis was that NGO merchants were driving the decline, because publicly shared payment links could attract users with low purchase intent.
The data disproved my first hypothesis. That is part of the story, not something to hide: hypothesis, evidence, revision.
04 · THE ANOMALY
Transaction-level logs revealed approximately 500 transactions initiated through a single email ID within a five-minute window, originating from outside India during non-standard business hours.
The metric wasn’t telling the whole story.
05 · FOLLOW THE TRANSACTION
I worked with Backend Engineering to trace the abnormal transactions back to their origin.
The investigation confirmed that automated scripts were generating fraudulent bot transactions through a Shopify source integration.
06 · THE PRODUCT PROBLEM
How do we stop them without disrupting legitimate users?
This is where the work stopped being analysis and became a product decision.
07 · OPTION A
What it solves: it creates an immediate barrier for automated scripts.
The product concern: it introduces an additional interaction into the checkout experience for legitimate users.
08 · OPTION B
Detect abnormal script execution patterns in the background and mitigate suspicious transactions without requiring an interaction from genuine users.
09 · THE TRADE-OFF
Both options stop automated scripts. They differ in who pays for it.
CAPTCHA puts the cost on every legitimate user at checkout. Detection puts it on the traffic that looks abnormal.
10 · THE DECISION
CAPTCHA was the initial proposed solution. After evaluating the trade-off between mitigation and user friction, I pivoted toward implementing bot detection first.
11 · FROM DECISION TO DELIVERY
I led problem definition, PRDs, tracking metrics, stakeholder alignment, sprint prioritization, cross-functional testing and post-launch monitoring.
12 · THE RESULT
Within one month of launch, SRT improved from 60% to 67%.
The bot traffic was mitigated without introducing an upfront CAPTCHA challenge into the checkout experience.
Product Manager · AI and SaaS
Connecting the dots
“You can’t connect the dots looking forward; you can only connect them looking backward.”
— Steve Jobs