EPAM Systems · Piplsay
7 months. 15k respondents. The honest story of Piplsay — a quantitative research platform that shipped, gathered real data, and got shelved.
The Problem
The platform had to serve a seasoned market researcher who lives in data AND a brand manager who just wants answers by Friday. Building for both without compromising either was the central design challenge.
Alex Rivera
Senior Market Research Analyst
"I need efficient tools to quickly gather & analyse data."
Uses advanced tools for data analysis. Conducts market research frequently.
Hover / focus to reveal pain points
Jordan Lee
Brand Manager, Tech Startup
"How can I maximise quality insights with minimal time investment?"
Limited experience with market research tools. Needs answers, not a PhD.
Hover / focus to reveal pain points
The Market
Every competitor chose between ease of use OR analytical depth. None of them had built the translation layer between business questions and research methodology — that was Piplsay's opening.
— competitors | — Piplsay vision
Piplsay's differentiation wasn't in the survey format — it was in the translation layer between business question and research methodology that no competitor had built.
The Innovation
Step 1 of 3 — Tell us what you'd like to know
I want to hear from smokers about insights on
Enter in one topic (company, event, person etc.) that you would like insights on
The system behind one sentence
Entity
AstraZenica
Product / Person / Event
→ Overall Satisfaction
→ Satisfaction by rec.
→ Satisfaction by attr.
→ Brand Attitude
→ Competence
→ Trustworthiness
→ Intent to Buy
→ Maximum Price
→ Price vs Value
Insight Recipe
Brand Health
Study
Reusable template
The metaphor was the product. Anyone with a business question could now access professional-grade research methodology — like picking up a recipe and cooking with it.
The Experience
The entire survey creation experience condensed into three steps — each one a single plain-English action. The complexity lived in the system; the interface made it feel simple.
A conversational sentence builder — type your target audience, your topic, and your goal in plain English. Trending social topics surfaced to reduce the blank-canvas problem.
Design question we were solving
"How do we restrict users to typing ONE entity? Do we show classification feedback?"
A split-panel recipe builder: browse constructs on the left, configure your research questions on the right. Popular Recipes tab surfaces pre-validated templates.
Design question we were solving
"Dependency based on previous construct output — how does the replaceable entity fit into the sentence?"
Cover page editor with live mobile preview. Credits system shows cost transparency before launch. Insight Study Summary shows the full problem statement back to the researcher.
Design question we were solving
"From which face do we position configuration options — to respondent or to client?"
The Build
Three parallel tracks — strategy, visual language, and respondent UX — all converging into a single coherent platform. The team deliberately treated the survey respondent as a first-class user.
Sep 2021
Oct
Nov
Dec
Jan 2022
Feb
Mar
Avadhesh
Lead / Strategy / Hi-Fi
Bharti
Personas / Visual Language
Rajeshwari
Respondent UX
The Moment
This wasn't a prototype. Piplsay shipped, launched live studies, and returned real demographic insights across a global respondent pool.
0k
Total Respondents
Macbook Pro insight example
0
Countries Reached
Global panel distribution
0%
Completion Rate
3,268 / 15,000 in progress
Respondent Demographics — Macbook Pro Insight
Real-time demographic breakdown by age group and gender as respondents completed the study.
Measured insights: Affect, Like/Dislike, Satisfaction, Purchase Intent — live, real-time as responses came in.
The Reality
Two numbers that changed everything. Not failures of design — failures of the ML maturity needed to make the design work at the quality the vision demanded.
0%
Quotas Detection Accuracy
The system struggled to reliably detect when demographic quotas had been filled — creating unpredictable data quality.
0%
Recipe Relevance Score
Suggested constructs didn't match user intent closely enough. The ML engine needed more training data than existed at beta.
Homepage
ShippedSign Up & Login
ShippedDashboard
ShippedNLP Targeting
Scaled back significantly
PartialCreate Insight (3-step)
ShippedLaunch + In-Progress Page
ShippedData Visualization
ShippedSurvey Preview
0% — never implemented
0%Onboarding Flow
0% — never shipped
0%Payment Details
0% — not built
0%Clone / Rename Insight
Incomplete
PartialUser Notifications
0%
0%The design was production-quality. The ML engine that needed to power entity classification, quota detection, and recipe relevance wasn't ready. Demos hid this gap. Beta exposed it.
The Retrospective
Not every project ships to production and stays there. The Piplsay research platform was shelved after beta. These are the real reasons — and what the team learned.
The four reasons it was shelved
ML maturity gap: entity classification & quota detection below acceptable quality threshold
Missing onboarding left new users without context on how the recipe system worked
Incomplete payment flow meant the product couldn't generate revenue to fund further development
Engineering bandwidth shifted to core Piplsay mobile app after beta results
What worked
The 3-step sentence builder — every user test confirmed it lowered barrier to entry
Construct-based recipe system was genuinely novel and impressed all stakeholder reviews
Respondent-first design track ensured survey UX was never an afterthought
Hi-fi quality reached production standard before beta — rare at this stage
The visual language held consistent from wireframe through to live screens
What we'd change
Ship a prototype of the ML layer earlier — discover quota detection gaps in usability testing, not beta
Define "done" for NLP entity classification before designing the UI that depends on it
Onboarding flow should have been non-negotiable scope — first-run experience was never built
Payment flow needed to exist before live user testing (credibility gap with real users)
More frequent design reviews of the respondent side with the engineering team
What shipped — the hi-fi quality that was never held back by design
Design lesson: Validate technology feasibility during exploration, not after hi-fi is shipped. A brilliant UX backed by an immature ML system is still a broken product.
The Principles
Four design principles extracted from seven months of work, real user data, and the experience of watching a well-designed product get shelved for non-design reasons.
Hover each card to see annotation highlights
When a product serves two users at opposite ends of a spectrum, the design job is to build a translation layer between them — not to find a middle ground that satisfies neither. The construct recipe system was that
The "recipe" metaphor wasn't just copy — it was the UX architecture. It determined the mental model, the information structure, the interaction flows, and how users described the product to each other. When a metaphor works this well, protect it.
Most B2B research tools treat the survey respondent as an afterthought. Dedicating a full design track to the respondent experience was unusual — and it was the right call. Respondent experience directly affects data quality, which is the product's core value.
The biggest mistake was building production-quality hi-fi designs on top of ML systems that hadn't been stress-tested. Entity classification and quota detection errors only became visible after 15,000 respondents. The floor check belongs in exploration, not beta.
Final thought
Piplsay didn't ship to the market permanently. But the design decisions made during those seven months — the recipe metaphor, the dual-audience architecture, the respondent-first approach — are the kind of thinking that compounds across projects.
The question at the top of this page was: "Why was the Quant research platform shelved?" The honest answer is: the design was right. The infrastructure wasn't ready. That distinction matters.
Applied in this case study
More work
Real-Time Insights Platform
Product Design Lead · 2023
Design System at Scale
Systems Design · 2022
Let's design something
worth keeping.
I'm Avadhesh Kumar Singh — Product Design Lead with 7+ years building products that ship, fail productively, and make teams better.
Get in touchPiplsay · EPAM Systems · Sept 2021 – March 2022