Roadmap

We have a in-app surveys product that allows our customers to show a survey to their users based on specific triggers, such as properties, events, or page views.

A product tours product would work in a similar way, and enable our customers to create no-code or low-code onboarding tours, announce new features, or have notification banners in-app.

We want some "just use this" way of displaying cookie banners that would let our users just be compliant and to get back to building stuff.

Question for community

• What requirements would you have / do you want us to just tell you what they are?
• If we were cheapest in the market, would you pay something for this or not?
• Would this mean you'd use our JS snippet on our website when you aren't already?

The basics would be:

• A way of users generating tickets/bug reports
• Assigning tickets to your team
• Messaging users when a ticket is created / responded to / resolved

We are unclear on (feedback hugely appreciated):

• If early stage companies would need it to work with discord/slack or if that's unusual
• If you are excited about this because you would happily replace an existing product - which one? Or if you'd just use us as you're not using anything else at the moment

Beyond this:

• Some way of prioritizing tickets based on user data
• Integrating session replay and errors into the ticket automatically

Google Tag Manager (GTM) is used to place tracking pixels on your website for ad platforms.

Requirements:

• Ability for people to place 3rd party pixels on specific URLs / groups of URLs
• Ability for people to show 3rd party pixels for specific users
• It must work without the need for code so non technical people can use it easily

Community Questions

• Any other requirements?

Very much a rough idea...

We have a really wide variety of features we could work on, coming from all over the place, and it's getting a bit messy internally at the moment:

support tickets recurring themes (ie "X" area of product is complicated) we could spot, perhaps with AI specific issues (ie X is broken) users asking for things on calls / social media / slack / email / meetings our grand ideas of what we should build next (ie posthog.com/roadmap!) and upvoting as a signal here should we build something that pulls all this together?

I could see it being our own version of a tool like Jira:

pulling stuff from all of the above, and creating issues connecting with github like jira would for example

Basics:

• Log Collection: Capture logs from web apps or other digital tools in real-time, supporting structured (JSON, XML) and unstructured (plain text) log formats.
• Centralized Dashboard: View and filter logs by source, severity, timestamp, or keywords, all from a unified interface. Search & Filtering: Easily search and filter logs based on customizable parameters (e.g., user ID, request ID) to find specific entries quickly.
• Error Tracking & Alerting: Automatically flag errors and alert relevant team members based on severity levels or specific keywords.

Questions for the Community:

• Would you use this? What would your use case be?
• Would this replace an existing tool in your current workflow? If so, which one?
• What additional features would be critical for your needs?

If you have any feedback, you can comment on the GitHub issue linked below.

Self-building products seems like the future, albeit with a sprinkle of human intervention.

There are a ton of these tools appearing and they're frankly awesome.

We'd want to ship something like:

• IDE based on a fork of VSCode or similar
• Chat UX for features you want it to build with approve/reject approach
• Multi-model support

We think we can do a better job of this long term because getting the MVP built is about 5% of the battle, the rest is making it good! And PostHog has all the tools to do the latter bit.

We could integrate our products into the IDE / suggestions really nicely – putting things behind flags, creating experiments etc. completely automatically in the further flung future.

If we have access to codebases we can then use this to enhance our product analytics AI features, for example, as we'll have more context on what your product does so will be able to answer vaguer usage questions way better.

This is a stretch for us as it feels like it needs a desktop app – frankly we've been looking for an excuse to build for desktop.

What do you like about these tools? What do you dislike?

As painful as it feels, a lot of issues is a good problem to have - it means you're in demand!

It didn't take long for us to need some kind of user-raised issue tracking system. For example, users reporting bugs / confusion or feature gaps.

There are third-party tools for this (like Zendesk), but we think they're underwhelming and the range of tools you need to integrate with them to make them feel like spaghetti – the bad kind, not the kind your mom makes.

We think a product here would have something like:

• Issue capture and categorization from users
• Issue triage across your team
• Alerting (email/Slack) so you realize that issues are happening
• Email messages back/forth to your users each time you respond
• Reporting on issues types so you can see what it causing most of your problems

Long term, we think we can win here because:

• We capture errors and replays for faster debugging
• We capture usage (product analytics), revenue and CRM data (via our warehouse), so have way more context on who users are and how to prioritize them
• We have loads of developers using us and can probably do some cool developer-specific features that our users invariably will ask for. Please tell us these!

It turns out people are creating rats' nests of prompts. And we don't like rat nests – they have rats in them.

Also, we have heard less technical folk than engineers are well placed to tweak prompts – like product managers.

We think a product that stored logic for this could keep things neat and tidy, and it should make it easy for PMs to manage prompts. Engineers can finally free themselves from convincing ChatGPT to bend to their will.

Are there tools you're using for this already? What do you like? Have you self built something to do this that'd be a helpful product for the world at large?

"How do I know my prompts are rad" you may be asking yourself.

We saw a lot of developers just peeking at a handful manually like it's still 2024.

Well, hello - welcome to 2025. This product would do these things:

• Use AI to evaluate if your prompts are good for you
• Capture a signal if users are happy with your prompts if possible (but not required)
• Create suggestions for how to improve them
• Integrate beautifully with our in product experimentation tool to offer experimentation across prompts/models/configurations
• Offer a range of models to test prompts with

Any feature requests?

This feature would enable to PostHog to automatically disable flags when...

• a metric has dropped below a certain threshold
• we are seeing an increase in errors, either through PostHog-native error tracking or
• third party integrations

We're considering building a link tracking product to allow you to organize all of your short links in a single platform, and be able to properly attribute conversions, revenue, etc. to them.

You should be able to:

• Create short links: PostHog will provision a short domain for this, but you should be able to choose your own short domain as well
• Monitor performance: Keep track of clicks, conversion rates, revenue, etc.
• Link customization: Set custom Open Graph images for social media sharing

What else? Let us know!

We have a ton of AI companies using us, we think they need roughly this list of things as a prompt engineering product:

• Playground: Somewhere to iterate on prompts in collaborative playground with all major proivders, variables, automatic versioning
• Data sets: Build datasets using CSV uploads
• Activity tracking: Usage, latency, logs
• Evaluations: Context recall, llm-rubric
• Prompt versioning: So you can rollback changes that break things

How you can help

• Email james at you can guess it with which of these matter, which don't, and if i have missed anything, if you use something else already, and if you'd pay something for the above!

tl;dr - mo' users = mo' problems, ai - pls fix

As we've grown, we've gone from a few to thousands of customer questions appearing. AI can handle quite a lot of these.

There are loads of third-party products to help here, but we'd rather just offer something along with the rest of our platform.

What features do you like? Anything you don't like about competitive products here? Do you even have docs this could use? Please let us know in the GitHub issue

We think the main features are roughly:

• Somewhere you upload or link to your docs
• An API where you send a question to us and we return a response
• A front end chat widget for your users to ask questions / get answers
• Contextual memory of each conversation with users
• A confidence score the response is correct
• Draft mode in the widget (where your support people can approve a response vs it always answering things automatically)
• Zendesk integration
• Slack integration

Longer term, we think we can do a better job of this than everyone else:

• We could generate docs for you based on session replays
• We have all customer data – i.e errors/usage stuff, so we've more context to answer questions
• We could import your codebase to understand your product more deeply. This takes trust and we think we're trusted given we've so many customers, and are open source

We offer transformations of event data using our legacy "plugins" system, which is very limited to specific use cases.

The next generation would be to build on our new "Hog" based destinations and offer the same power for transformations, allowing users to customize to write their own code to transform the data as it comes in.

We have custom webhook destinations for sending data out. What if we offered webhook "sources" for getting data in.

Essentially this would offer a way of setting up a PostHog webhook that would accept almost any payload shape that can then be transformed via Hog Function into the appropriate shape for our event ingestion.

A toolbar for mobile SDKs, bringing similar functionality to what is already available on the web.

Features may include:

• Creating actions directly from autocapture elements
• Visualizing heatmaps and scroll-depth
• Other features? please share your ideas!

What it could look like:

• In-App Toolbar: Floats on top of your app's UI
• Remote Toolbar: Accessible via a local browser interface
• Other?

Community questions

• Any specific ideas or requirements for your projects?
• Any existing solutions out there that do this well? we’d love to hear about them!

Lots of people build or want to build on our APIs.

We should find a way to let people access our AI product via API so they can offer in built AI capabilities to their in app analytics (or whatever else it can touch)

We'd need to charge for this separately so we don't rack up huge bills etc