# Message test — https://survicate.com/features/research-hub/ After reading your page, 11 of 15 personas could name a reason to pick you over a similar option. - **Page tested:** https://survicate.com/features/research-hub/ - **Audience tested against:** Product, research or CX teams working at a B2C saas or digital product working with high volumes of feedback. - **Personas:** 15 simulated - **Report:** https://grader.wynter.com/r/research-hub-ai-research-repository-that-keeps-kqnZGHg > These answers are generated by AI, scored on Wynter's B2B Message > Layers framework using behaviorally-diverse simulated personas. The > methodology is real and the critique is directional. What a simulated > persona cannot have is a live budget, a renewal coming up, or a boss > asking about this quarter. --- ## 01 · The scores Every persona answered all four questions. These are four independent proportions of the same panel, not stages of a funnel. | Layer | Question | Cleared the bar | Strength | Of those who passed | | --- | --- | --- | --- | --- | | 1. Clarity | Do they understand what you do? | 15/15 | 84% | 4 without hesitation, 11 with reservations | | 2. Relevance | Can they tell what it solves, and who it's for? | 15/15 | 78% | all with reservations | | 3. Value | Do they actually want it? | 13/15 | 70% | all with reservations | | 4. Differentiation | Is there a reason to pick you over the alternatives? | 11/15 | 63% | all with reservations | **Brand alignment** (a side metric, not one of the four layers) — 14/15, 74% strength (all with reservations). Does the page read like the company you actually are? **Fix first: Differentiation.** Earliest failing layer, walking the sequence in order — not simply the lowest score. --- ## 02 · What to change, layer by layer Ordered worst-first. Specific edits, not a restatement of the score. ### Differentiation **Attach proof to each comparison-table row.** The 'Research Hub vs LLMs' table asserts 'Full dataset, no ceiling', 'Verifiable insights, backed by real data', and 'Enterprise grade security, EU Servers' with nothing standing behind them. Readers treated the whole table as marketing self-assertion. Put a concrete artefact next to at least the Volume, Control and Security rows: an actual dataset ceiling figure (e.g. 'tested on X responses per project'), a stated cross-check or verification accuracy result, and named certifications (SOC 2… *effort medium · impact high · tested against Proof next to the claim* **Lead the comparison section with traceability, not agent architecture.** The one differentiator readers could name back was linking every AI claim to the original customer quote — several said it is the only real separation from a general-purpose LLM. But the section headline 'Your research deserves more than just an LLM' and its subhead sell 'purpose-built agents, 100+ researcher workflows, and embedded research methodology', which is internal architecture. Rewrite the headline around clicking a finding through to the verbatim that produced it, and move that row… *effort low · impact high · tested against Give a reason to choose you* **Answer the 'prove it on my data' objection near the CTA.** Readers said the differentiators would only be credible after running them against their own multi-source volume, and the page leaves that objection unresolved between 'Try Research Hub free' and 'Book a demo'. Add a line beside the CTA that names what a trial actually gets them: connect your own sources, run one project, click any claim through to the source quote — the 5,000 free datapoints already exist as an offer but are stranded at the very top of the page and never framed as the way to… *effort low · impact medium · tested against Answer the live objection* ### Value **Replace unsourced time claims with sourced, named-customer numbers.** The days-to-hours and two-day figures were read as unverified marketing arithmetic. Either attribute each figure to a named customer and workload ('Wave Apps: 4,000 NPS responses across 6 sources, first sourced report in 3 hours') or drop the number and describe the work removed instead. A figure with a named source and a dataset size beside it survives scrutiny; a bare 'days to hours' does not. *effort medium · impact high · tested against Proof next to the claim* **Show hallucination prevention working instead of claiming it.** 'Control — Hallucinations, no verification layer' and 'Every conclusion linked to verifiable feedback' are the claims readers most wanted demonstrated, several because they had been burned by an AI tool before. Make the tour thumbnail 'Take a tour of Research Hub' explicitly a click-through-to-source demo, labelled as such ('See a claim clicked through to the raw quote — 40 sec'), and describe the mechanism in one plain sentence: what the agents cross-check and what happens when a claim has no… *effort medium · impact high · tested against Proof next to the claim* **Add named research leads with firm size to social proof.** 'Trusted by 2000+ digital businesses' and six unlabelled 'Company Logo' slots do no work; the Josh Litwin quote did, because it carries a role and a company. Replace the anonymous logo wall and the '2000+' framing with two or three named research or insights leads, their company, and the scale they run — sources connected, responses analysed — so a buyer can find a comparable firm to themselves. *effort medium · impact medium · tested against Specifics beat superlatives* ### Clarity **Define Research Hub in one sentence before the H1.** 'Research Hub', 'Insights Hub' and 'AI Research Repository' all appear without a flat definition, and the distinction between the Hubs is never drawn. Also, 'repository' undersells the product — readers played it back as an analysis layer that synthesises and writes sourced reports, not a store. Rewrite the H1 to say what it does rather than what it is filed under, and use one product name consistently through the page. *effort low · impact medium · tested against Plain language* ### Relevance **Name research and insights teams in the hero.** The audience is only inferable from customer titles and testimonials further down. Add the role to the subhead under the H1 — a line a research or insights lead can point at, e.g. 'For research and insights teams running continuous programmes across surveys, interviews and support tickets' — so it is not left to be reconstructed from Josh Litwin's job title. *effort low · impact medium · tested against Name the audience* --- ## 03 · What is working ### The core mechanic — ingest fragmented feedback, synthesize, produce sourced reports… Five respondents played back the product accurately as an AI analysis layer that aggregates feedback from multiple sources, auto-tags it, and outputs research reports with traceable quotes. This was the most consistently clear part of the page. > AI tool that pulls customer feedback from surveys, tickets, reviews, and drafts research reports with sourced quotes. > > — Senior Product Manager, SaaS, 5000+ > That part was actually clear because they spelled out the ingestion sources and the output types (reports, dashboards, chat) rather than just saying "AI insights." > > — CX Manager, SaaS, 201-500 > a layer on top of your CRM/helpdesk/survey tools that turns scattered qualitative and quantitative feedback into synthesized, sourced research reports > > — Senior Product Manager, SaaS, 5000+ > pulls in surveys, support tickets, app reviews, interview transcripts, etc. and uses AI to draft research reports and spot patterns, with every claim supposedly traceable back to the original quote > > — CX Manager, SaaS, 201-500 > the friction was upfront, phrases like "AI Research Repository that keeps you in control" and "context" are used so often across SaaS pages now that they're near-meaningless until you get to the concrete bit about linking findings to source quotes > > — Customer Experience Manager, Digital Products, 501-1000 ### Named research-role customers were credible; generic social proof was not enough One respondent said named research-role customer proof beats a generic logo wall. Two others said the social proof falls short without named research leads from comparable firms and transparent sourcing for the metrics cited. > one Wave Apps quote and one anonymous "Head of User Research" isn't enough proof for me to take a meeting — I'd want three or four named research leads at companies my size confirming the citation-back-to-quote thing > > — Product Manager, Software, 1001-5000 > the two customer proof points — Josh Litwin at Wave Apps and the anonymous "Head of User Research" quote about turning two days of slide decks into hours — are named-enough and specific-enough (real title, real consequence) that they'd keep this on the shortlist > > — Director of Product, Digital Products, 11-50 > "2000+ digital businesses" and "97% satisfaction rate" are just numbers dropped with zero source, which makes me trust the rest of the specifics less, not more > > — CX Manager, SaaS, 201-500 ### Traceability back to source quotes is the one differentiator respondents could name and… Six respondents pointed to the ability to link AI-generated claims back to original quotes as the concrete, distinguishing feature — several called it the only real differentiator versus general-purpose LLMs. Multiple said it was the specific thing they would test in a trial or demo. > The "AI drafts it, you refine it" framing plus "verify every claim back to its source" is the bit that would actually get me to sit through a demo, because it's a specific mechanism claim, not just "AI insights" > > — CX Manager, SaaS, 201-500 > The traceability angle - "every finding links to the customer voice behind it" - is the one concrete differentiator I'd point to, since a competitor without that would just give me another black-box summary. > > — Research Manager, Software, 51-200 > pulls in surveys, support tickets, app reviews, interview transcripts, etc. and uses AI to draft research reports and spot patterns, with every claim supposedly traceable back to the original quote > > — CX Manager, SaaS, 201-500 > I'd want them to show a live example of a report where I can click a claim and see the source ticket or transcript > > — Customer Experience Manager, Digital Products, 501-1000 --- ## 04 · What the personas said ### 'Research Hub' and 'Insights Hub' are never defined, and 'repository' undersells the… Three respondents said the Hub terminology is used without a flat upfront definition and the distinction between the two remains unclear. One added that calling it a 'repository' understates the synthesis and reporting the product actually does, and one described friction from vague framing until the linking-to-quotes feature appeared. > terms like "Research Hub," "Insights Hub," and "research project" all got used close together without ever being flatly defined, so I was inferring the category from context clues > > — CX Manager, SaaS, 201-500 > the friction was upfront, phrases like "AI Research Repository that keeps you in control" and "context" are used so often across SaaS pages now that they're near-meaningless until you get to the concrete bit about linking findings to source quotes > > — Customer Experience Manager, Digital Products, 501-1000 > the phrase "AI Research Repository that keeps you in control" is a bit of marketing shorthand that doesn't say what it does until you hit the "editable report where every finding links to the customer voice behind it" line further down, so the category name itself ("repository") undersold the synthesis/reporting function > > — Director of Product, Digital Products, 11-50 > "Research Hub" versus "Insights Hub" get used in the same breath without a clean definition of the split (there's even an FAQ item asking that exact question, which tells me even they know it's confusing) > > — Customer Experience Manager, Digital Products, 501-1000 ### Differentiation claims are asserted without third-party proof, benchmarks, or security… Three respondents said the comparison table carries no third-party validation or before/after benchmark, and one flagged the absence of SOC2/ISO citations given the product ingests sensitive feedback data. Several said the differentiators would only be credible after testing against their own data volume. > it's still just their own claims with no benchmark or third-party validation, so it wouldn't beat out a competitor who could show me an actual before/after report on messy, multi-source data > > — CX Manager, SaaS, 201-500 > I'd want them to prove the "no ceiling" claim with our actual data volume in a trial, because right now it's just a table cell, not evidence > > — Customer Experience Manager, Digital Products, 501-1000 > the security row just says "Enterprise grade security, EU Servers" with no SOC2/ISO citation despite the footer badges — for a tool ingesting support tickets and interview transcripts, that's the line I'd push on > > — Product Manager, Software, 1001-5000 ### Respondents recognised research teams as the audience, but had to infer it from job… Four respondents said the page conveys quickly that it is for research teams rather than generic marketers or PMs. Four others said that audience is only inferable from customer titles and testimonials and should be named directly, with one asking for product-buyer proof rather than research titles alone. > The intended reader wasn't spelled out with a job title, but I inferred it from the quotes — "Senior Manager, Research and Insights" at Wave Apps and "Head of User Research" — so this is clearly aimed at research/insights people > > — CX Manager, SaaS, 201-500 > A line naming the role or team directly — "built for CX and research teams" or similar — instead of making me infer it from a job title in a testimonial; right now I have to do the work of matching myself to the audience rather than being told. > > — Customer Experience Manager, Digital Products, 501-1000 > I'd want a line naming the role directly — something like "Built for Product and Insights leaders who need to prove a hypothesis before a roadmap decision" — plus a customer logo or quote from a Director of Product specifically, not just Research/Insights titles, since that's a slightly different buyer than me > > — Director of Product, Digital Products, 11-50 > "researcher-controlled" and quotes from a "Senior Manager, Research and Insights" and "Head of User Research" make it obvious this is for research/insights teams, not generic marketers. > > — Research Manager, Software, 51-200 > The reader is implicitly a research/insights person — "Josh Litwin, Senior Manager, Research and Insights" and "Head of User Research" testimonials make that explicit > > — Product Manager, Software, 1001-5000 > The reader isn't named outright as "CX Manager" or "researcher," but the testimonials ("Senior Manager, Research and Insights," "Head of User Research") and the LLM comparison table make it obvious this is aimed at research/insights people > > — CX Manager, SaaS, 201-500 > "2000+ digital businesses" and "97% satisfaction rate" are just numbers dropped with zero source, which makes me trust the rest of the specifics less, not more > > — CX Manager, SaaS, 201-500 ### The time-savings claim is believed as a premise but treated as unverified Five respondents accepted that eliminating manual data stitching and report synthesis is where the value sits, and two said it could justify a pilot. But the days-to-hours and two-day figures were repeatedly flagged as unquantified and requiring verification on their own multi-source data or a live demo. > Saves report-writing time, quotes traceable to source - decent if true. > > — Senior Product Manager, SaaS, 5000+ > I'd want to see it run against our actual messy multi-source data (not a demo dataset), get a real number on hallucination/error rate, and know who signs off on 15+ integrations and data residency before I'd take it past a first call. > > — Customer Experience Manager, Digital Products, 501-1000 > that would cut the days I currently spend manually stitching together survey data, support tickets, and interview notes down to maybe hours > > — Product Manager, Software, 1001-5000 > That's a concrete time saving I'd want quantified before committing. But "full dataset, no ceiling" and "hallucinations, no verification layer" vs "verifiable insights" are just assertions in a table with no methodology or numbers behind them > > — Research Manager, Software, 51-200 > two days on slide decks down to hours, tweaking an AI draft > > — Senior Product Manager, SaaS, 5000+ > I'd want them to show a live example of a report where I can click a claim and see the source ticket or transcript > > — Customer Experience Manager, Digital Products, 501-1000 ### Hallucination prevention is claimed but unproven, which is where verification demands… Respondents said the verification and hallucination-prevention mechanisms lack proof on real data, and asked for a live demo of click-through-to-source before committing. One tied this directly to a previous AI failure they had experienced. > the rest (report quality, hallucination prevention specifics) I'd still need proof on before I'd trust it > > — Product Manager, Software, 1001-5000 > The "AI drafts it, you refine it" framing plus "verify every claim back to its source" is the bit that would actually get me to sit through a demo, because it's a specific mechanism claim, not just "AI insights" > > — CX Manager, SaaS, 201-500 > the "every finding links to the customer voice behind it" and the LLM-vs-Research Hub table (full dataset not a sample, verifiable insights, multi-agent cross-check) are exactly the two things that burned me last time > > — Customer Experience Manager, Digital Products, 501-1000 > I'd want them to show a live example of a report where I can click a claim and see the source ticket or transcript > > — Customer Experience Manager, Digital Products, 501-1000 --- ## 05 · The hardest read An adversarial pass over the findings. Every claim below was checked against the panel's own answers; unsupported ones were dropped. - **Every number on the page is treated as an unverified assertion, so the value proposition cannot close anyone without a demo.** *(high)* 6 of 15 flagged the days-to-hours and two-day figures as unquantified and requiring verification on their own data; 3 said the comparison table carries no third-party validation or before/after benchmark; 3 said metrics lack transparent sourcing. The page's numeric claims consistently push the decision off-page instead of advancing it. - **The one differentiator that landed is also the one respondents refuse to believe until they test it, so the page's strongest asset generates a demo dependency rather than conviction.** *(high)* 6 of 15 named traceability to source quotes as the only real differentiator versus general-purpose LLMs, and multiple said it was the specific thing they would test in a trial. Separately, 3 said the verification and hallucination-prevention mechanisms lack proof on real data and asked for live click-through-to-source, one citing a prior AI failure. The page asserts the exact capability that carries the most buyer skepticism with zero evidence attached. - **Comprehension is not the problem; credibility is — and the page invested in the thing that was already working.** *(high)* 5 of 15 played back the core mechanic accurately, the most consistently clear part of the page. Against that, 6 flagged unverified time savings, 3 flagged unproven differentiation and missing SOC2/ISO citations, 3 flagged thin social proof, and 3 flagged unproven hallucination prevention. Fifteen negative credibility mentions against clear comprehension means added explanation buys nothing. - **The page produces pilot interest, not purchase intent, and that ceiling is self-imposed.** *(high)* Only 2 of 15 said the value could justify a pilot, and that was conditioned on verification against their own multi-source data or a live demo; multiple others said differentiators would only be credible after testing against their own data volume, and the traceability differentiator was named as the thing they would test in a trial. Nothing on the page converts without a subsequent evaluation step. - **Naming the audience is left to the reader, so relevance depends on scanning testimonials rather than on the copy.** *(medium)* 7 of 15 engaged the audience question and 4 of them said research teams are only inferable from customer titles and testimonials and should be stated directly, with one asking for product-buyer proof rather than research titles alone. The page is outsourcing its core targeting work to social proof. - **The product's own vocabulary works against it: undefined 'Hub' labels and the word 'repository' shrink the product below what respondents said it actually does.** *(medium)* 4 of 15 said 'Research Hub' and 'Insights Hub' are never defined and the distinction between them stays unclear, one said 'repository' understates the synthesis and reporting, and one described friction from vague framing until the traceability feature appeared. The naming actively suppresses the value the rest of the page is trying to establish. --- ## 06 · Who answered | # | Role | Industry | Company size | | --- | --- | --- | --- | | 1 | CX Manager | SaaS | 201-500 | | 2 | Customer Experience Manager | Digital Products | 501-1000 | | 3 | Product Manager | Software | 1001-5000 | | 4 | Senior Product Manager | SaaS | 5000+ | | 5 | Director of Product | Digital Products | 11-50 | | 6 | Research Manager | Software | 51-200 | | 7 | CX Manager | SaaS | 201-500 | | 8 | Customer Experience Manager | Digital Products | 501-1000 | | 9 | Product Manager | Software | 1001-5000 | | 10 | Senior Product Manager | SaaS | 5000+ | | 11 | Director of Product | Digital Products | 11-50 | | 12 | Research Manager | Software | 51-200 | | 13 | CX Manager | SaaS | 201-500 | | 14 | Customer Experience Manager | Digital Products | 501-1000 | | 15 | Product Manager | Software | 1001-5000 | --- ## 07 · Before you act on this The methodology is real, and the critique is directional. What a simulated persona cannot have is a live budget, a renewal coming up, or a boss asking about this quarter. **Validate anything you're betting on with real ICPs who are actually in-market.** Being wrong is more expensive than you think. Finding out is cheaper than you'd guess. Wynter runs message testing with verified B2B professionals — trusted by HubSpot, RingCentral, Shopify, Cognism, Paddle, Veeam, Rippling and Miro. This report is kept for 60 days from 2026-08-25, then deleted along with the personas and their answers.