Transforme 500 Avaliações de Clientes em um Roadmap de Produto Priorizado com Este Prompt de IA
Why this prompt matters
Without a systematic way to analyze feedback, product teams default to building what the loudest customer asked for last, not what would actually move retention or NPS. Missprioritized sprints cost engineering weeks on features that address no real pain. A single poorly-read feedback batch can push a team to solve a symptom instead of the cause, compounding both technical debt and user-experience debt for the next two quarters.
What we use it for
Your team just closed a quarterly NPS survey and has 400 open-text responses sitting in a spreadsheet alongside 200 unread App Store reviews from the past month. Sprint planning is on Friday and the team is debating what to build next, but no one has had time to read through 600 feedback items. You need clear signal from that noise before the meeting starts.
Prompt
Role: Act as a senior product analyst with expertise in customer research and data synthesis. Your job is not to summarize feedback — it is to extract signal, identify patterns, and translate customer language into actionable product decisions that a team can execute this quarter. Context: I have collected [NUMBER] pieces of customer feedback from [SOURCES: App Store reviews / Support tickets / NPS surveys / User interviews / Social media / Other]. The product is [BRIEF PRODUCT DESCRIPTION]. Our current development focus is [AREA: onboarding / core feature / performance / reliability / pricing / other]. I want to understand what customers are actually saying versus what they are asking for, identify the most painful issues versus nice-to-have requests, and prioritize what to fix or build next. Task: 1. Cluster the feedback into themes. Identify 5 to 8 meaningful themes — not generic categories like "UI" or "bugs", but specific patterns like "users cannot find the export function" or "onboarding flow loses users after step 3". 2. For each theme, estimate frequency (how many feedback items relate to this theme), sentiment intensity (how frustrated or delighted do users sound), and inferred business impact (if unresolved, what does the company lose?). 3. Identify the top 3 Quick Wins — issues that appear frequently, cause real friction, and are likely low effort to fix. 4. Identify the top 3 Strategic Priorities — themes that represent deeper product gaps with high customer impact. 5. Flag any feedback that suggests a customer segment or use case the product is not currently serving well. 6. Write one paragraph that could serve as an internal brief: what are customers really saying, in plain language? Constraints: - Do not treat feature requests at face value. Look for the underlying job the customer is trying to do. - Separate correlation (many people mentioned X) from urgency (these customers sound ready to churn over X). - Do not invent themes that are not actually supported by the feedback. - Keep technical jargon out of the internal brief — write it so a non-technical founder or executive can read it. - If the feedback volume is too small to draw reliable conclusions, say so clearly rather than speculating. Output Format: 1. Theme clusters (name, description, estimated frequency, sentiment intensity: low/medium/high, inferred business impact) 2. Quick Wins (3 items with reasoning) 3. Strategic Priorities (3 items with reasoning) 4. Underserved segments or use cases (if any) 5. Internal brief paragraph (2-4 sentences, plain language) 6. Confidence note (how confident are you in these findings given the data quality and volume?) Feedback to analyze: [PASTE YOUR CUSTOMER FEEDBACK HERE — reviews, tickets, survey responses, or interview notes]
Result
THEME CLUSTERS
1. Export and data portability Description: Users are trying to get their data out of the product into spreadsheets, other tools, or for reporting to clients. The current export button exists but produces CSV files with inconsistent column naming, no date filters, and no per-project scope. Estimated frequency: 27% of feedback items Sentiment intensity: High — several users have directly stated they are evaluating alternatives because of this Inferred business impact: Significant churn risk in the B2B segment; enterprise buyers consistently flag data ownership as a procurement requirement
2. Onboarding drop-off after step 3 Description: First-time users consistently report confusion at the workspace setup step. The term "workspace" does not match the mental model users arrive with, and the required fields feel like bureaucracy before they have seen any value. Estimated frequency: 19% of feedback items Sentiment intensity: Medium — frustration rather than rage, but correlates with low day-7 retention in support ticket volume Inferred business impact: Estimated 30-40% of trial signups do not complete setup. Improving this step is the highest-leverage acquisition funnel fix available.
3. Notification overload Description: Users are receiving too many emails and in-app pings for low-priority events. Several long-term customers report that they have disabled all notifications, which means they are also missing important alerts. Estimated frequency: 14% of feedback items Sentiment intensity: Medium — annoyance that is eroding trust with power users Inferred business impact: Notification fatigue is a known precursor to disengagement; if power users are tuning out, they are also missing important product updates and re-engagement touchpoints
4. Mobile app stability on Android Description: Android users specifically report crashes on the reports screen when viewing projects with more than 50 items. This is a distinct pattern from general performance complaints. Estimated frequency: 11% of feedback items Sentiment intensity: High — these users are blocked, not inconvenienced Inferred business impact: Android accounts for an estimated 38% of mobile sessions; crashes on a core screen are a direct retention risk
5. Pricing and tier confusion Description: Users do not understand what the Pro plan includes versus Free. Several churned users mentioned they paid for Pro expecting a feature that is actually on Enterprise only. Estimated frequency: 9% of feedback items Sentiment intensity: High — these users feel misled Inferred business impact: Pricing page clarity could recover failed upgrades and reduce churn in the $29/month tier
QUICK WINS 1. Fix the export CSV column naming and add a date range filter. Low engineering effort, directly addresses the most urgent churn signal. 2. Rename "workspace" to "project" in the onboarding flow and remove the two required fields that users skip or fill with placeholder text. 3. Add a notification frequency control (daily digest vs. real-time) as a single toggle. Existing notification infrastructure supports this; it is primarily a UI addition.
STRATEGIC PRIORITIES 1. Export and data portability — build a full export center with scoped exports, date filters, and PDF report output for client-ready deliverables. 2. Onboarding redesign — the step-3 confusion is a symptom of a deeper mismatch between product language and user mental models. A short Jobs-To-Be-Done research sprint before the next redesign would pay for itself. 3. Mobile reliability — Android crash patterns suggest the reports screen has an unhandled edge case at scale. This needs a dedicated bug sprint before the next mobile release.
UNDERSERVED SEGMENTS Freelancers managing multiple clients appear in roughly 15% of feedback but the product currently treats all projects as belonging to a single workspace owner. This segment is hitting limits the product was not designed for, and several are asking for client-portal or sharing features that would require a different product model.
INTERNAL BRIEF Customers who are churning or at risk are not unhappy with the product's core concept — they are blocked by execution gaps: they cannot get their data out easily, they cannot set up the product without hitting confusing terminology, and on Android they cannot reliably use a core screen. The clearest near-term wins are operational fixes, not new features. The most important strategic gap is that a meaningful segment of freelancers is using the product in a way it was not designed for, and either the product adapts or that segment churns to a competitor built for them.
CONFIDENCE NOTE Moderate to high. The sample of 600 feedback items is sufficient to identify the top themes with reasonable confidence, but frequency estimates should be treated as directional, not precise. The Android crash pattern and export frustration are robust findings. The pricing confusion theme is based on a smaller subset and should be validated with a targeted survey before committing significant design work.
A maioria dos times de produto passa mais tempo debatendo prioridades do que lendo o feedback que resolveria o debate. Os dados existem — avaliações da App Store, respostas abertas de NPS, tickets de suporte, anotações de entrevistas com usuários — mas processar centenas de comentários brutos em uma decisão de roadmap defensável leva um tempo que a maioria dos times não tem antes do sprint planning.
Este prompt dá à IA um trabalho estruturado para fazer com esse feedback. Em vez de pedir um resumo, ele pede extração de sinal: agrupar o feedback em temas significativos, pontuar cada um por frequência e urgência, separar correlação de risco de churn, e produzir três coisas que a maioria das análises de feedback perde — Quick Wins, Prioridades Estratégicas e um sinalizador para segmentos de usuários mal atendidos.
O que torna este prompt diferente
Prompts genéricos do tipo 'analise meu feedback' devolvem resumos genéricos. Este força o modelo a separar o que os clientes dizem do que eles querem dizer. Um usuário que diz 'adicione um modo escuro' pode na verdade estar frustrado com o cansaço visual durante longas sessões — o trabalho subjacente é conforto, não estética. A restrição de 'procurar o trabalho subjacente' empurra consistentemente os modelos para além de listas de funcionalidades superficiais.
O requisito de nota de confiança é igualmente importante. Se você cola 12 avaliações e pede uma análise de temas, a maioria dos modelos inventa padrões. Este prompt instrui o modelo a declarar quando a amostra é muito pequena para tirar conclusões confiáveis — o que acaba sendo surpreendentemente útil quando seus dados são escassos.
Como usar
Preencha quatro campos: o número de itens de feedback, de onde vieram, uma descrição do produto em uma frase e sua área de foco de desenvolvimento atual. Em seguida, cole o feedback bruto no final. O modelo cuida do resto — agrupamento, pontuação e escrita do parágrafo de brief interno que você pode jogar diretamente em uma mensagem do Slack ou documento de planejamento.
Funciona melhor com Claude 4 Sonnet, GPT-5 ou Gemini 2.5 Pro. Para grandes volumes de feedback (500+ itens), cole em lotes de 150 a 200 e peça ao modelo para mesclar os clusters de temas em uma passada final.
Como é a saída
O prompt produz seis seções estruturadas: clusters de temas com pontuações de frequência e impacto, Quick Wins com justificativa, Prioridades Estratégicas com justificativa, segmentos mal atendidos (quando presentes), um brief interno em linguagem simples e uma nota de confiança sobre a qualidade dos dados. O brief interno foi projetado para ser legível por um fundador ou executivo não técnico sem edição.
O exemplo de saída neste post é baseado em um produto SaaS fictício com 600 itens de feedback em cinco temas — desde atrito na exportação até crashes no Android e confusão de preços. É representativo da profundidade e especificidade que o prompt produz de forma confiável.