We Didn’t Have a Product Problem. We Had a Memory Problem.

MULIKI SIRINI — Tue, 29 Sep 2026 16:03:44 +0000

Building PRATHIDHWANI: A Memory-Powered Customer Feedback Intelligence System Using Hindsight

A few months ago, I was sitting in a product discussion when someone asked a deceptively simple question:
“Why did we decide not to build this feature?”

The room went silent.

One person vaguely remembered the discussion. Another thought it had something to do with customer feedback. Someone started searching through old messages. Ten minutes later, we still didn't have a clear answer.
The information probably existed somewhere.

The problem was that nobody could quickly reconstruct what happened, why the decision was made, and whether the reasoning was still relevant.

That moment made us realize something:
We didn't have a product problem. We had a memory problem.

That realization became the foundation for
PRATHIDHWANI — Customer Feedback Intelligence.

The Knowledge Is There. The Context Isn't.
Modern product teams generate enormous amounts of knowledge every day.

Customer interviews reveal pain points. Support tickets expose recurring problems. Roadmap meetings create strategic decisions. Engineering releases introduce product changes. Retrospectives capture lessons learned.
But this information is rarely connected.

Customer feedback might live in one system. Product decisions might exist in meeting notes. Release information might be somewhere else. Historical discussions may be buried inside chats.

Eventually, someone asks:

Why was this feature rejected?

Why did the roadmap change?

Did the latest release actually solve the problem?

Why are customers still complaining about this?

The information exists.

Finding the relationship between the information is the difficult part.

This is the problem PRATHIDHWANI is designed to address.

When Smart Teams Repeat the Same Conversation

One of the most frustrating patterns we noticed was how often teams unknowingly revisited problems they had already discussed.

A feature request arrives.

The team investigates it.

Customers are interviewed.

Tradeoffs are analyzed.

Engineering constraints are considered.

A decision is made.

Then everyone moves forward.

Six months later, another person sees the same feature request.

The original context is gone.

The conversation begins again.

This isn't because the team lacks intelligence.

It's because the organization's memory is fragmented.

PRATHIDHWANI approaches this differently by connecting customer feedback, product changes, and historical context.

Instead of simply asking what customers are saying today, we can begin asking:

What changed after we listened to them?

From Feedback to Product Decisions

The central idea behind PRATHIDHWANI is simple:

Customer feedback becomes more valuable when it is connected to what the product actually did about it.

Suppose hundreds of customers report problems with checkout.

Looking at the feedback tells us there is a problem.

But that isn't the end of the story.

The product team might release a fix.

The next release might show fewer complaints.

Another issue might emerge.

Six weeks later, the team needs to know whether the original problem was actually resolved.

Without historical context, the team is looking at separate snapshots.

With historical memory, the team can connect:

Feedback → Product Change → Outcome → Historical Context → Next Decision

That is where PRATHIDHWANI becomes more than a feedback dashboard.

It becomes a product intelligence system.

What the PRATHIDHWANI Dashboard Shows

The current PRATHIDHWANI dashboard brings together feedback trends and historical signals in one place.

Across the January, February, and March releases, the system has analyzed 1,248 pieces of feedback.

The overview shows:

61% positive sentiment, representing a +14% change versus the baseline.

At the same time:

27% negative sentiment, with Mobile OTP concentration identified as an important source of negative feedback.

The system also surfaces 4 emerging issues requiring attention.

These numbers are useful because they give the product team an immediate overview.

But the more interesting information appears when we look at how individual issues are moving over time.

Not Every Trend Means the Same Thing

The Top Customer Issues section highlights several important trajectories.

Mobile OTP (Gateway) shows a +280% trajectory, associated with token delivery failures.

Checkout (Apple Pay) shows a +94% trajectory involving voucher reset callbacks.

Dashboard navigation shows +42%, associated with hierarchy friction.

But there is another important signal:

Export performance — -76% (Resolved), associated with asynchronous report workers.

That last signal demonstrates an important principle.

A product intelligence system shouldn't only identify problems.

It should help teams understand whether their interventions worked.

A decreasing complaint trajectory after a product change can provide evidence that the change had an impact.

That leads to a much more useful question:

“We released the fix. Did it actually fix the problem?”

Why Historical Memory Matters

One of the key sections in PRATHIDHWANI explicitly compares the same inquiry:

Without Hindsight — Current Data

versus

With Hindsight — Continuous Memory

Without historical memory, an AI may produce a generic present-state observation:

“Export performance is currently good. March feedback indicates that exports are running quickly.”

That might describe the current situation.

But it doesn't explain the journey.

What was happening before?

How serious was the original issue?

What changed?

When did the improvement happen?

Did customers actually respond differently after the release?

Historical memory allows those questions to become part of the answer.

This is the difference between knowing what is happening and understanding what happened.

Why Hindsight?

The difficult part of this project wasn't generating text.

Large language models are already extremely capable at generating responses.

The difficult part was remembering.

We used Hindsight as the memory layer behind PRATHIDHWANI.

The idea is straightforward:

Customer Feedback
↓
Product Changes
↓
Historical Context
↓
Hindsight
Memory Layer
↓
Relevant Memories
↓
AI Reasoning
↓
Product Intelligence

Hindsight provides the persistent memory layer.
The AI provides reasoning and response generation.

PRATHIDHWANI connects the two to the product workflow.

The result is an AI system that can reason using organizational history rather than treating every interaction as an isolated event.
More Than a Dashboard
The system is also organized around the questions product teams repeatedly face.

Dashboard Overview provides key metrics, issue trends, and comparative synthesis.

Fix Verification Loop helps check whether released fixes actually resolved customer complaints.

Prioritization Split supports side-by-side evaluation of potential next-sprint priorities.

Account Revenue Risk focuses on enterprise accounts experiencing friction across multiple releases.

Weekly PM Briefing provides an automatically generated view of executive sentiment and aging issues.

Identity Stitching connects complaints across channels to account context.

Track Product Changes records release events so teams can evaluate post-release impact.

And Memory Explorer provides a way to explore the historical context behind the information being surfaced.

These modules all point toward the same goal:

turning fragmented feedback into usable product knowledge.

What We Learned

Building PRATHIDHWANI changed our perspective on AI.

At the beginning, we asked:

“How intelligent can we make the assistant?”

Eventually, we realized the more important question was:

“What can the assistant remember?”

A model can generate an excellent answer.

But an answer without historical context can still be incomplete.

A customer complaint is useful.

A product change is useful.

A sentiment score is useful.

A release metric is useful.

But connecting them creates something more valuable.

It creates understanding.

The goal isn't to replace product managers or engineers.

It's to make sure that the knowledge those people create doesn't disappear when the conversation ends.

Memory Should Be Infrastructure

One of our biggest lessons was that memory shouldn't be treated as a small feature added to an AI assistant.

It should be treated as infrastructure.

Every customer interaction can become useful historical context.

Every product decision can preserve its reasoning.

Every release can become evidence.

Every resolved issue can become a lesson.

Over time, these individual events create something much more valuable than a database of feedback.

They create a memory of how the product evolved.

That memory can help teams make better-informed decisions without repeatedly rediscovering the same information.

Where PRATHIDHWANI Can Go Next

The long-term vision extends beyond the current dashboard.

PRATHIDHWANI can connect organizational memory with project-management tools, support platforms, collaboration systems, meeting transcripts, product analytics, customer feedback, and roadmap systems.

Imagine asking:

“Why did we prioritize this feature?”

and receiving an answer connected to the customer evidence and historical discussions behind the decision.

Or asking:

“Did the last release solve the problem?”

and seeing how customer feedback changed before and after the release.

Or asking:

“What needs attention next?”

and receiving a response based not only on today's data, but on how issues have evolved over time.

That's the real promise of persistent product memory.

The Bigger Idea

Product teams don't suffer from a lack of data.

They suffer from fragmented context and forgotten decisions.

PRATHIDHWANI was built around a simple idea:

Don't just analyze what customers are saying. Remember what happened after the team listened.

Hindsight provides the memory layer.

AI provides the reasoning layer.

PRATHIDHWANI connects them into a continuous product intelligence workflow.

Because the most valuable AI answer isn't always the one generated by the smartest model.

Sometimes it's the answer that remembers what your team already learned.

Without memory, AI starts over every time.

With memory, AI can build on where you've already been.

And that is the idea behind PRATHIDHWANI — Customer Feedback Intelligence:

Turn customer feedback into product decisions.

DEV Community: MULIKI SIRINI

We Didn’t Have a Product Problem. We Had a Memory Problem.

Building PRATHIDHWANI: A Memory-Powered Customer Feedback Intelligence System Using Hindsight