ChatGPT prompts for your PM work
These examples are sourced from Lenny's Newsletter - https://www.lennysnewsletter.com/p/how-to-use-chatgpt-in-your-pm-work Go here for my best prompt hack.
Prompt engineering plays a crucial role in enhancing the performance of AI language models and ensuring more accurate, relevant, and reliable outputs.
Duplicate this document and continue adding your own!
Collect and summarize user feedback and usage data.
Synthesize survey results
I'm going to send you a list of X survey responses to the question "...."
Can you group these into buckets of insights, with their associated weight (in %), and list some examples of associated responses for each insight?
===
And the list of responses in the following messageSource
Find feature ideas and bugs from app store reviews
I want to scrape App Store reviews of our app using Javascript. Walk me through it step by step.
_________
Then, I pasted all the reviews into the chat and asked GPT-4 to find the top 3 most requested features and complaints.Source
Extract insights from raw usage metrics
Converting our product usage metrics data to text and then just asking GPT questions about the data.Source
Analysing lead funnel to identify conversion issues
You can paste in the data in a table and ask for the conversion percentages. And then ask scenario questions. "What is the volume of raw leads that would be needed to generate x conversions?"
Source
Come up with product name suggestions
Strengthen your argument
Come up with critical questions your audience may ask
PRD Review: "Assume you are the CTO, review this PRD and give me critical, but fair feedback)
Before meetings : "I'm meeting the VP of data science for a 30 minute call to discuss X. What should I ask her?"
After meetings : "Summarize my notes into minutes, action items"Source
Identify gaps and hidden assumptions in your thinking
“What am I missing here?”
“What am I being overly optimistic about?”
“What is a macro event that can totally reverse the outcomes of this test”Source
I am building collaboration software for users to create their mind maps and share them with their teams. The following are the expected steps by a user
1. User sees a sample mind map on our website
2. User clicks on a button titled 'Use this template'
3. User signs up for an account and then makes their own mind map based on the template
4. The user shares the mind map with their team members for review
5. After completing the trial period, the user upgrades to a paid version
Considering the above steps as an ideal behaviour. Can you list out some assumptions that we are making about the user, their context and their motivations which might prevent them from upgrading to a paid plan?
===
That's a good start. Can you dig deeper and come up with 20 more assumptions. Think from a perspective of feasibility of building the software, usability of the software, desirability and viability at each step
===
Thank you. Out of these, what are the top 3 assumptions that we should be focusing on.
Also suggest some experiments or action items to mitigate their risk.Source
Highlight edge cases and counterarguments:
1- ‘Tell me 5 reasons this feature won’t work as intended’
2- ‘Tell me 5 unintended consequences of this feature’Source
Steelmanning the other side of an argument when drafting a proposal (“why shouldn’t we do this?”)
Source
Inspire roadmap ideas
Assist with roadmap ideation
Brainstorming use cases. example: "I'm considering implementing discounts in my ecom business. We use Shopify connected to SAP as our ERP. What are the use cases I should document and review before writing up requirements?"
Source
Develop frameworks
“Come up with a framework for…” staring at a blank doc is sometimes the worst part.
Source
Come up with customer interview questions
You are a {industry} product manager interviewing customers about {topic}. Objectives: {objectives}. Please give me {number} questions to ask customers. Consider {supplemental information about industry/customers}.Source
Inspire PRDs and user stories
Create a v1 PRD or Jira ticket
PRD Writing: Write a product requirement document for an A/B test project for ("Short A/B test project description")
Write Jira Tickets: Write a Jira ticket with user story and acceptance criteria for - ("Short project description")Source
Identify drag metrics
Besides the basics like grammar. I've found it to be very helpful in identifying counter metrics (or guardrail metrics) for the PRD's success criteria.
Prompt includes an overview of the feature, customer pain points, user stories, and success criteriaSource
Write user stories
write user stories for onboarding process of a shipping app
===
create a detailed prd for a fixed audit asset management product called XYZ
===
I need a cost benefit analysis on what it'll take to build [product name] over a competittion product Source
Improve your writing
P1: What would be a better way to say this : add your version
P2: What data points should i add to make my argument
P3: in the above email, add another point related to this :
P4: say the same in a different way
P5: Polish this text:"insert text"
Fun fact , you can also generate specific tones.Do market research
how much did TV publisher X make from ad revenue vs cable subscription fees over the last five years.Source
Ask technical questions
SQL queries
-
-
Source
Programming specifics
I'm a little rusty in R, how do I import this set of .csv files and make a plot of this data over time?Source
Explain a broad concept
Explain <topic> and cite your sourcesSource
Write the code for you
write a react function which is a button that when you click it, it downloads SVGSource
Create a v1 landing page
I am designing the landing page for a collaborative knowledge workspace software called [insert product name].
It helps individuals and teams [insert benefits].
Can you help me draft the copy for the landing page?
Separate sections such as [headings], [features], [CTAs].Source
Create a v1 pitch deck
Can you provide an example of a bottom-up market analysis for a tech productSource
And, maybe most importantly, help you say no
Source
My best hack for prompting ChatGPT is to talk like it's "baby AI."
When you talk with a toddler, you start by explaining the concept and then make your point at the end if you think they grasped it.
So this is how you can improve your prompts:
- **Ask the model what it knows about a subject -> **what do you know about B2B SaaS Pricing best practices?
- Double down on one of the items listed. Reiterate the key points you want the answer for. -> let's focus on Add-on pricing.lease give me some best practices and strategies to implement an add-on to a PLG B2B SaaS company with an average ACV of $YY
- Continue on more items if necessary -> let's focus on add-on item number X
- Final prompt ->
Act like a pricing expert for B2B SaaS with 10 years of experience helping companies with pricing. Recommend the pricing strategies with the highest chance of improving the conversion rate from free trial user to customer. Consider that the add-on pricing model is preferred. List all the actions a pricing expert would take to improve the conversion rate from user to customer.
Generative Models Explained
This is the best video I found explaining how generative models work:
Prompt Engineering Best Practices
Prompt engineering is the process of designing, refining, and optimizing prompts to effectively interact with and elicit desired responses from AI language models, such as GPT-4. The main goal of prompt engineering is to improve the quality of AI-generated outputs by crafting well-structured, unambiguous, and context-rich prompts.
Since AI language models are trained on large datasets and learn to generate human-like text, the quality of their responses can be influenced by the way users phrase their queries or statements. Prompt engineering involves various techniques, such as:
- Being explicit: Clearly specifying the desired format or type of response.
- Adding context: Providing relevant background information or context to help the model understand the query better.
- Repeating important information: Restating key points or concepts to emphasize their importance.
- Requesting step-by-step explanations: Asking the model to provide detailed explanations or reasoning behind its response.
- Experimenting with different phrasings: Trying different ways of asking the same question to find the most effective prompt.
Remember:
- LLM are good at generating the next syntactically correct word
- they can give a false impression that the actually understand the meaning
- be carefull for false naratives.