Research - June 2026

From signal to decision

Sift is a B2B SaaS that turns scattered customer feedback into clear, prioritized product decisions. Every roadmap item is traceable back to the real user voices behind it. This document is the complete research foundation for the product design sprint.

The synthesis trust gap is real and undefended. No direct competitor demonstrates a trustworthy full-chain synthesis with honest confidence levels on public pages.

The "confident PM defending her roadmap" positioning is unclaimed. All HARD competitors target broad teams; no one owns the PM-specific evidence story.

The market has converged on AI capture and classification. AI synthesis that a PM can trust enough to act on is not the dominant pattern - it is the open opportunity.

Trust is earned through decomposability and inline citation, not through confident-sounding outputs. Showing the work is the mechanism, not the output quality.

The shareable evidence brief is simultaneously the acquisition asset, the retention hook, and the referral artifact. Building it well is the highest-leverage early investment.

Lean UX Canvas v2

Executive Summary

One-glance synthesis of all research findings, placed here for executive context before the detailed sections follow.

Lean UX Canvas v2

Sift - Feedback Synthesis to Defensible Decision

Framework by Jeff Gothelf, Lean UX Canvas v2 - populated from research conducted June 2026

1 Business Problem

Product teams at B2B mid-market SaaS companies cannot make confident, evidence-backed roadmap decisions because customer feedback is scattered across 3-7 disconnected tools, synthesis is manual and unreliable, and every prioritization call can be challenged by sales or leadership with no traceable evidence to counter it.

2 Business Outcomes

60% of activated users cite Sift as primary input for roadmap decisions within 90 days.

Planning session time under 25 minutes.

Free-to-paid conversion 8-12% within 60 days.

12-month NRR 110%+.

3 Users

Primary: PM at B2B SaaS, 28-38, 3-8 years experience, drowning in scattered signal, needs defensible prioritization.

Secondary: Head/VP of Product managing 3-8 PMs, presenting evidence-backed strategy to leadership.

4 User Outcomes and Benefits

Finish a prioritization session in under 25 minutes with defensible evidence, not guesswork.

Say "Sift shows this is the top priority" in a stakeholder meeting and mean it.

Close the loop with sales: here is the evidence behind why we are building this.

5 Solutions (MVP Scope)

Feedback ingestion from 3-5 sources (Intercom, Zendesk, Gong, reviews, CSV). AI synthesis into transparent theme clusters with evidence counts. Full drill-down: theme - evidence items (n=X) - raw source text. Honest confidence display (sample size visible, thin evidence flagged). Prioritization view with scoreable themes. Shareable public evidence brief. Free: 250 items/month, 1 integration, themes visible. Paid: unlimited + full traceability + team features.

6 Hypotheses

H1 (riskiest): If synthesis is transparent and traceable, PMs will trust it enough to act without re-synthesizing raw data themselves.

H2: If first synthesis uses real data and confirms something the PM already suspected, activation will complete.

H3: If evidence briefs are shareable as public links, sharing will become the primary organic acquisition channel.

H4: If upgrade prompts are context-specific, conversion will exceed a generic upgrade wall.

7 Riskiest Assumption

PMs will trust Sift's AI synthesis enough to act on a real roadmap decision without re-synthesizing the raw data themselves. If false, the core time-saving value fails.

8 Smallest Test

Show a PM their own feedback synthesized into 3-5 themes with evidence counts. Ask: "Would you present this in a roadmap review without auditing the raw data first?" A "yes" validates H1. Target: 6 out of 10 PMs in prototype interviews say yes.

Strategy

Objectives

#	Objective	Metric	Target
1	PMs trust and act on Sift synthesis	% citing Sift as primary input for last roadmap change	60% of activated users at 90 days Hypothesis
2	Faster prioritization sessions	Avg. time from session start to decision	Under 25 minutes Hypothesis
3	Free-to-paid conversion	Conversion within 60 days of hitting limit	8-12% Hypothesis
4	Team retention and expansion	12-month net revenue retention	110%+ Hypothesis

Audience Segments

Segment	Profile	JTBD	Priority
The Overloaded PM	PM at B2B SaaS, 28-38, 3-8 years, manual synthesis today, lives in Jira/Linear	When planning sessions come up, want trustworthy traceable synthesis to make a defensible call	Primary
The Evidence-Hungry Product Lead	Head/VP Product, 35-45, manages 3-8 PMs, presents strategy to board	When presenting to leadership, want systematic evidence behind each priority	Secondary
The Signal Supplier	CS Manager or Support Lead, daily customer contact, insights lost in transit	When sharing feedback, want to see it was heard and used	Later

Primary JTBD

"When I have a roadmap planning session coming up, I want to see a trustworthy synthesis of what customers are saying - with evidence I can trace - so that I can make a confident, defensible prioritization call and not just guess."

Business Model

Free Plan

Up to 250 feedback items/month. Theme clusters visible (names and counts). 1 integration. Single-user workspace. No evidence drill-down. No team features. No sharing or export.

Goal: Let the team experience the synthesis quality before hitting the evidence-tracing wall.

Paid Plans Hypothesis

Starter (~$30-40/seat/mo): unlimited signal, full traceability, 3+ integrations. Growth (~$20-25/seat/mo, 5-20 seats): team features, roadmap sync. Enterprise: custom.

All pricing unvalidated. Willingness-to-pay research required before setting prices.

Riskiest Assumption

PMs will trust Sift's AI synthesis enough to act on a real roadmap decision without re-synthesizing the raw data themselves.

If false, the core time-saving value fails. Sift becomes another data source they process manually. See Hypothesis H1 in Section 7.

AARRR Funnel

Stage	Primary metric	Target	MVP decision
Acquisition	Organic/PLG signups/month	200/month at 6 months H	Build shareable evidence brief as acquisition asset
Activation	% completing first synthesis on own data within 7 days	40% H	CSV import + Intercom as first connections; reach synthesis fast
Retention	Week-4 retention rate (activated users)	50% H	"New signal since last session" digest as re-engagement hook
Revenue	Free-to-paid conversion within 60 days of limit	10% H	Context-specific upgrade prompt at the exact gated feature
Referral	% signups attributed to referral	25% at 12 months H	Public shareable brief with "made with Sift" attribution

The brief is everything

The evidence brief drives acquisition (shared publicly), referral (circulates in org tools), and retention (team becomes dependent). It is the single most impactful early product investment for AARRR.

Activate on real data only

Demo data does not create trust. The activation moment is "I see a theme from my own feedback that I already suspected is real." Confirmation with evidence, not discovery of something new.

Competitive Analysis

Sources: live product sites and pricing pages, fetched June 2026. Screens captured via Playwright into research/screens/.

Competitor Groups

Name	Group	Why in group
Productboard	Hard	Direct competitor: centralizes feedback, generates roadmaps, AI synthesis via Spark
Canny	Hard	Direct competitor: AI autopilot for capture, vote-based roadmap, most widely adopted mid-market
Enterpret	Hard	AI synthesis layer: adaptive taxonomy, 50+ integrations, enterprise-focused intelligence infra
Dovetail	Soft	Same JTBD (evidence-based decisions) from the research angle. "Build with facts not vibes."
Jira Product Discovery	Soft	Same JTBD via Atlassian ecosystem. Absorbed Cycle's team. Free tier with Jira integration.
Linear	Aspirational	Gold standard for craft, density, and clarity. Not a functional competitor.
Perplexity AI	Aspirational	Best-in-class citation transparency. Direct model for how Sift should show evidence.

Competitor screens (public/pre-login pages only)

Productboard - Homepage

Productboard - Spark Pricing ($15/maker/mo billed annually)

Canny - Homepage

Canny - Pricing (Free/Pro $79/Business custom)

Enterpret - Homepage (Enterprise only, contact sales)

Dovetail - "Build with facts, not vibes"

Dovetail - Pricing (Free/Enterprise 2-tier)

Jira Product Discovery - Post-Cycle acquisition

Comparison Matrix

Product	Audience	Key mechanism	Trust signal	Pricing
Productboard Hard	Enterprise + scaling orgs, 6,000+ teams	LLM workflow prompts (Spark), feedback classification, roadmap views	[? behind login]	$19-59/maker/mo (productboard.com/pricing)
Canny Hard	Mid-market cross-functional, 100K+ companies	AI Autopilot captures from Gong/Intercom/Slack; vote aggregation; public roadmap	Vote counts + linked feedback per feature visible pre-login	Free 25 users, Pro $79/mo (canny.io/pricing)
Enterpret Hard	High-velocity enterprise: Canva, Notion, Apollo	5-level adaptive taxonomy, 50+ integrations, NL querying ("Why are customers churning?")	Positions as infrastructure, not answer layer. Honest about being the synthesis layer.	Enterprise only, contact sales
Dovetail Soft	Design/research first, expanding to product + CX	AI tags + insight docs with source clips. "Build with facts, not vibes."	Clip-level evidence - insight docs link back to specific video/text quotes	Free (1 project), Enterprise custom (dovetail.com/pricing)
Jira Product Discovery Soft	Atlassian-native PM teams, scaled to enterprise	Configurable scoring + Jira backlog link + Cycle team's approach to discovery	[? behind login]	Free 3 creators, $10-25/creator/mo (atlassian.com/jira/product-discovery/pricing)

3 Common Patterns

1. AI for capture and classification, not trusted synthesis.

2. Free tiers create desire, not full value - volume/collaboration limits are the universal conversion lever.

3. Broad multi-team positioning dilutes the PM story.

3 Key Differences

1. Infrastructure (Enterpret) vs. workflow tool (Productboard, Canny) vs. research layer (Dovetail).

2. Bottom-up PLG (Canny, JPD) vs. top-down enterprise sales (Enterpret).

3. Customer-facing (Canny: public roadmap) vs. internal-only (all others).

The Gap (Sift's Opening)

No competitor demonstrates a transparent, traceable synthesis chain (signal to theme to evidence to decision) with honest confidence levels and challengeable conclusions. This is Sift's primary differentiator.

Benchmark: Trust in the Synthesis

5 best-in-class products evaluated outside the direct competitor set, scored against 8 criteria for trust and traceability.

Benchmark product screens

Perplexity AI - Inline citation transparency benchmark

Amplitude - Drill-down from metric to raw event

Linear - Density without noise, aspirational craft

Linear Method - Design principles for product tooling

Linear Pricing - From free to $16/user/mo Business

Scores

Product	Transparency	Evidence trace	Confidence honesty	Conflict handling	Inline citation	Density	Override	Memory	Total
Perplexity AI	3	5	4	3	5	4	2	1	27/40
Amplitude	4	5	3	3	3	4	4	4	30/40
Linear	2	3	N/A	2	3	5	5	4	24/35*
Grain	2	5	2	2	4	3	3	3	24/40
Notion	2	3	1	1	3	3	5	3	21/40

*Linear scored out of 35 (Confidence honesty N/A - not a synthesis tool). Scale: 5 = best-in-class, 1 = absent or broken.

Top 3 Mechanisms to Carry into MVP

Mechanism 1

Inline numbered citation

From Perplexity AI. Every synthesis theme gets an inline citation count (n=47). Clicking opens the evidence list. The affordance of verifiability is the trust mechanism - users do not need to audit every claim, they need to know they can.

Mechanism 2

Progressive drill-down

From Amplitude. Theme - Evidence items (n=X) - Full original text. The chain is accessible without leaving the view. Decomposability is the rational foundation of trust. The user trusts the summary because they know they can audit it.

Mechanism 3

Honest confidence display

From Perplexity + Amplitude combined. Every theme shows item count, source diversity, and a confidence indicator. Thin evidence is flagged. Honesty about weak evidence increases trust in strong evidence.

1 Mechanism That Will NOT Work

Fully automated synthesis with no visible reasoning

Observed in Canny Autopilot and early Productboard AI. Presents conclusions without showing method. The skeptical PM who needs to defend conclusions cannot say "because the AI said so" in a roadmap review. One bad synthesis experience destroys trust permanently. A tool that shows its reasoning fails gracefully; a tool that hides it fails catastrophically.

UX Patterns

Core Behavioral Patterns

#	Pattern	Implication for Sift
B1	Synthesis-first, detail-on-demand Entry Point	Synthesis view is always primary. Raw feedback is accessible but never the starting point.
B2	Evidence as a social object	Export and share are primary use cases. Brief must read independently of Sift context.
B3	Trust through spot-checking	Show the most obvious theme first. Give the PM an early win on first session.
B4	Context-switching cost aversion	Every key view delivers value in 10-15 minutes. Jira/Linear integration is retention-critical.
B5	Retrospective synthesis over real-time monitoring	Default to synthesis on demand, not live dashboard. Low-volume session-oriented notifications.

Pattern Selection

Primary choice

The Report + Evidence Brief (Patterns C + E)

The core experience is a synthesis-on-demand report. The PM opens Sift, the synthesis of their current feedback is visible (auto-updated), and they can drill from any theme to its evidence. The session output is a shareable evidence brief.

Reason 1

JTBD alignment: the PM's job is to get to a defensible answer within a planning session. Report starts with conclusion, enables verification, produces shareable output.

Reason 2

Competitor gap: no HARD competitor nails the report-with-drill-down for PM synthesis. The pattern is an open position in the market.

Reason 3

Trust mechanism fit: inline citation + progressive drill-down enable synthesis-first, spot-check to verify, then act. The trust-building progression maps directly.

Alternative under condition X

The Inbox (Pattern B)

If the target segment shifts to CS/Support Leads (Segment C), the inbox pattern becomes primary. CS teams do process feedback individually and want a triage view. Not chosen for MVP because the PM is the primary target and inbox does not scale to PM synthesis needs.

Does not fit

The Canvas (Pattern D)

Does not scale to hundreds of feedback items. Not repeatable across sessions. Trust in a canvas depends on the person who built it, not the system. Defeats the core promise of systematized, consistent synthesis that is trustworthy independently of who ran the session.

Conclusions

Gaps Table

Gap	Description	Source	Confidence
Synthesis trust chain	No competitor demonstrates full traceability from raw signal to decision with honest confidence on public pages	competitors.md, benchmark.md	High
PM-specific positioning	"Confident PM defending her roadmap" story is unclaimed by any HARD competitor	competitors.md	High
Honest confidence display	No direct competitor shows sample size and confidence level alongside synthesis conclusions	benchmark.md	High
Evidence brief as product	The shareable evidence artifact is underbuilt across all competitors	aarrr.md, competitors.md	Medium
Free tier value delivery	Current free tiers too restrictive to create genuine "aha" moments at fair volume	competitors.md	Medium
Atlassian consolidation risk	JPD + Cycle acquisition could absorb the "good enough" mid-market via distribution advantage	competitors.md, cycle.app/blog	Medium - monitor

6 Hypotheses (If / Then / Because)

Riskiest assumption - failure kills the idea

If we build synthesis that is transparent (method visible), traceable (drill-down to raw signal), and honest about confidence,

Then skeptical PMs will trust the synthesis enough to act on a real roadmap decision without re-synthesizing the raw data themselves,

Because the presence of verifiability - knowing you can audit any conclusion - is the psychological mechanism that creates calibrated trust, even when the user does not actually audit every claim. (Source: Perplexity citation pattern, benchmark.md)

Activation

If the PM's first synthesis session uses their own real feedback data and surfaces at least one theme they already suspected is real,

Then they will complete activation and not abandon during onboarding,

Because confirmation of known intuition with evidence is a lower-friction trust-building moment than discovery of something new. The PM validates "yes, that is real" and extends trust to unfamiliar themes.

Evidence brief as viral loop

If every PM can share a synthesis output as a public link (no Sift login required to view),

Then evidence brief sharing will become the primary organic acquisition channel,

Because the brief circulates in environments (Notion, Slack, Google Slides) where the audience is both receptive to PM tools and influential enough to adopt them.

Conversion timing

If the upgrade prompt appears at the specific moment the user is trying to use a gated feature (evidence drill-down, team sharing, additional integration),

Then the conversion rate will be meaningfully higher than a generic upgrade wall,

Because context-specific conversion connects the price to a concrete, immediate benefit. The user knows exactly what they are buying.

The positioning gap

If Sift positions specifically for the PM who needs to defend roadmap decisions with evidence (vs. broad "product + CS + sales + marketing" positioning of competitors),

Then PMs in the target segment will self-select into Sift over competing tools,

Because specificity creates stronger resonance with the segment that has the pain, even at the cost of reducing the addressable market on paper. Being the obvious choice for one segment outperforms being a good choice for five. (Source: competitors.md)

Retention through team dependency

If Sift's evidence briefs become a standard output for roadmap reviews (shared with the PM's team, engineering, and leadership),

Then individual PM retention becomes team-level retention,

Because shared artifacts create organizational lock-in that individual feature value cannot. The brief is the exit barrier, not the synthesis engine.

Open Questions

What is the exact baseline time PMs spend on manual feedback synthesis today?

Needed to validate Objective 2 and sharpen activation/conversion messaging. [?] - no verified primary source found for specific hours figure.

Does Productboard Spark actually provide evidence traceability, or only AI workflow prompts?

If Spark solves the trust chain problem, the differentiation argument narrows. Needs hands-on evaluation [? behind login for full assessment].

Will Jira Product Discovery absorb the synthesis market through Atlassian's distribution?

If JPD builds credible synthesis post-Cycle acquisition, the "good enough" mid-market shrinks. Monitor JPD product releases quarterly.

What is the right free plan constraint - items/month or features/tier?

Wrong limit kills activation (too low) or kills conversion (too high). A/B test in early beta.

What willingness-to-pay exists at $30-40/seat/month in the primary segment?

Pricing hypothesis entirely unvalidated. Van Westendorp price sensitivity research needed with target PMs.

Do PMs share evidence briefs externally or keep them internal?

Determines whether the brief viral loop (H3) actually activates. Observe brief sharing behavior in beta: track internal vs. external shares.

Live Research

Adversarial verification conducted June 2026. Claims from personas.md and jtbd.md were tested against live sources. The goal was to kill claims first - search for disconfirming evidence before confirming. All sources fetched in June 2026; only sources from the last 1-2 years used.

Claims tested

Sources accessed

Confirmed

Killed

No claim was directly disconfirmed

Unresolved

U1 (baseline hours), U2 (brief sharing)

Access blockers

Reddit (network security), G2 (access restricted)

Confirmed findings

PM wariness toward AI for prioritization decisions

HIGH

Multiple independent sources confirm that PMs are wary of using AI for decision support specifically. The wariness is named as the reason AI adoption is limited to lower-stakes tasks.

"Most teams are using AI today for content generation and summaries. Far fewer trust it for prioritization or decision support. The reason is skepticism about the outputs."

Sources: aipmtools.org/articles/future-of-ai-product-management, blog.ravi-mehta.com, productboard.com/blog

PMs spot-check AI outputs before acting

MEDIUM

Spot-checking (testing AI output against known-good data before extending trust) is described as standard PM practice, not exceptional behavior. Confirms behavioral pattern B3.

"Understanding the data these tools rely on, sanity-checking outputs, and using simple eval style checks before trusting an AI workflow in a critical process is a core competency for modern product managers."

Source: aipmtools.org/articles/future-of-ai-product-management

Transparent citations make AI synthesis trustworthy for stakeholder presentation

MEDIUM

When AI synthesis includes citations, PMs trust it enough to present to executives. This directly confirms the Perplexity citation benchmark finding (benchmark.md, Mechanism 1) in the PM-specific context.

"Citations make it trustworthy enough to present to execs, and citations make it verifiable when presenting data to stakeholders or executives."

Sources: aipmtools.org/articles/future-of-ai-product-management, aipmtools.org/articles/ai-changing-product-management

H1 reformulated - riskiest assumption

Conditional H1: With transparency, PMs act on AI synthesis without re-synthesizing

MEDIUM

H1 is conditionally confirmed. When synthesis is transparent and traceable, PMs DO act on it without re-reading all raw data. They spot-check a subset (trust calibration), then act. Black-box synthesis is not trusted. The condition is transparency - inline citations, evidence counts, drill-down.

"There is no translation layer, no export step, no 'now what?' moment. The insight arrives already oriented toward action." (Productboard on Spark intent)

Sources: productboard.com/blog, aipmtools.org, blog.ravi-mehta.com, capterra.com (Canny reviews - Freeman C., Senior PM, Nov 2025). See strategy.md Section 5.

The synthesis gap at market level is real and current

HIGH

Real user reviews and industry analysis independently confirm that competitors have solved collection, not synthesis. The PM voice from Canny reviews names this gap explicitly.

"What is the real value of automatically adding 1000s of feature requests into Canny if there's no way of using AI to query or filter within those posts? I would have rather had some type of AI that could read ticket context and auto-assign tags or teams or categories or add to roadmap." - Freeman C., Senior PM (Capterra, Nov 2025)

Sources: capterra.com/p/161103/Canny/reviews/, aipmtools.org/articles/ai-changing-product-management

PM feedback pain - volume and scatter - is real

MEDIUM

Multiple independent sources confirm the baseline pain. The pattern is consistent even though no single source confirms the exact 3-5 hours per planning cycle figure. "Hours per week" is confirmed as the order of magnitude.

"One PM reported that summarizing feedback with an AI tool saves them hours every week, replacing the tedious process of reading through hundreds of tickets."

Sources: enterpret.com/blog/2025-product-planning-tips, productboard.com/blog, aipmtools.org

Killed claims

None. No claim from personas.md or jtbd.md was directly disconfirmed by this research. Killed count: 0. This is shown, not hidden.

The adversarial search was conducted: Reddit via browser (network security block), G2 (access temporarily restricted), direct search for PM abandonment of AI synthesis tools attempted. The unconditional reading of H1 (PMs trust ANY AI synthesis without verification) was tested and does not hold - but H1 as stated in strategy.md (trust transparent synthesis to act without re-synthesizing raw data) was not killed - it was conditionally confirmed.

What we still do not know

U1 - Exact baseline time PMs spend on manual synthesis per planning cycle

No primary source confirmed the 3-5 hours figure from research.md. Order of magnitude (hours) is supported. The specific figure requires user interviews. To close: ask target PMs in user interviews to time their next pre-planning synthesis session.

U2 - Whether PMs share evidence briefs externally or only internally

No source addressed this directly. All sources about sharing focused on internal stakeholder alignment, not external sharing. The referral viral loop (H3) depends on external sharing. To close: beta observation - track internal vs. external links in shared brief URLs.

Key implication

Sift's trust design is not optional polish - it is what makes the core job possible. The research confirms the conditional form of H1 and sharpens the design requirement:

PMs are wary of AI synthesis for decision support (confirmed, HIGH). That wall comes down with transparency: citations, evidence counts, drill-down, and the ability to spot-check (confirmed, MEDIUM). When those mechanisms are present, PMs DO act on synthesis without re-synthesizing everything themselves (confirmed, MEDIUM).

The first post-launch signal to watch: whether activated users open the evidence drill-down at least once in their first session. If they do, the spot-checking pattern is active and the trust chain is working. If they never open it, either synthesis quality is already trusted (good) or the drill-down is too hidden (bad). One behavior distinguishes between two very different product situations.

Full verification record: research/docs/live-research.md