What you actually get when you pay for a session recording

Contents

In PostHog your first 5,000 web recordings and 2,500 mobile recordings are free every month. After that, you pay per recording and the price per recording drops as you record more – this helps ensure bills don't get too massive if you need a lot of recordings.

The price per recording covers a wide range of responsibilities, from recording the session without slowing your product down to masking data, storing the recording, playing it back, and linking it to everything else you know about that user.

Recording a session is the easy part. rrweb, the open-source library our web recorder is built on, is free, and an AI agent can wire up a basic version in an afternoon – though it's much easier and also free to get started with PostHog.

This page explains the rest of the work beyond the recording itself, so you know what the price includes and can compare it with building your own system from scratch.

Every section starts in plain English and ends with the technical detail, so you don't need to be an engineer to follow it.

No bill yet? Every PostHog product has a free tier, and they're big enough that 97% of users never pay us a penny. If you record fewer than 5,000 web sessions a month it is always easier and cheaper to start with PostHog.

The price

A recording is one session: everything a person does in your product until they go quiet for 30 minutes, up to 24 hours. Session Replay bills per recording, so one person who visits three times in a day can make three recordings. Below are the live rates for web recordings.

5,000
recordings/month

$0

5k25k120k500k

Mobile recordings bill separately, with their own rates and a free tier of 2,500 a month. Keeping recordings for longer than 90 days needs a platform package.

Replay Vision, which uses AI to review recordings for you, bills in its own credits, and asking PostHog AI to find a recording in plain English spends AI credits.

Those are the only line items. We don't charge to watch a recording, to share one, or per seat, so your whole team can watch as many as they like.

What happens to one recording

A session replay tool has the same jobs whether you buy it or build it. It has to record what people do without getting in their way, keep what they type private, and get every piece of the recording to you. Then it has to play it back.

A frequent challenge for teams using session replays is finding which recordings are actually worth watching. PostHog solves this with Replay Vision, which can watch and summarize recordings for you with AI, but it is its own tool. AI summarization is not typically part of most session replay stacks.

In practice that's six steps.

Step 1: Recording sessions without slowing your product down

Your customer is halfway through checkout on an old laptop with 40 tabs open. The recorder runs in each tab and, every time the page changes, it has to notice, write the change, and send it off. The page can't feel any slower for it and, on a phone, it also has to go easy on the battery and the data plan.

How it works at PostHog: On the web, the recorder loads separately from the main JavaScript SDK. It takes one full snapshot of the page, then records only what changes: DOM mutations, clicks, scrolls, and input. It batches those changes together, compresses them, and sends them to a capture endpoint that handles recordings only.

Mobile apps don't have a DOM, so the mobile SDKs work differently. The native iOS and Android SDKs read the screen's view hierarchy and send a wireframe by default. React Native and Flutter send screenshots.

Sampling, minimum duration, and trigger rules run on the device. While a trigger waits to match, the recorder keeps a short buffer in memory, so a session that never qualifies is never uploaded and never billed.

A recorder is code that runs inside your product, so a bug in the recorder is a bug in your product. This is a major reason why it can make more sense to use PostHog than to build your own session recorder – because anything you build yourself is something you then have to maintain and manage. This includes being responsible for the performance impact.

Step 2: Keeping private data private

If people type passwords and card numbers into your product (or any other PII) then a recording that captures them is a liability you have to be accountable for.

How it works at PostHog: Privacy controls run in the browser or app, before anything is sent, so masked data never reaches PostHog. Every input field is masked by default, on mobile and web, and you can add the ph-no-capture class to any element to leave it out entirely. Network capture also scrubs common sensitive headers automatically, and you can add your own function to scrub request and response bodies.

Additionally, each recording is encrypted with its own key. When you delete a person and their recordings, we destroy the key, so nobody can ever play those recordings again.

Without these data protections in place, every new form your team ships is a new way to record something you shouldn't – and a new risk you have to manage in your product stack.

Step 3: Not losing it

A recording doesn't arrive in one piece. It arrives in many small batches, from every tab of your product the person has open, over as long as 24 hours. Some batches arrive late too, due to interference like a laptop closing mid-session.

How it works at PostHog: Recordings go through the same kind of pipeline as events. The capture service writes each batch to Kafka before anything else touches it, so a traffic spike or a problem further down the line waits in the queue instead of losing data. A session that sends an unusual amount of data moves to an overflow lane, so it can't slow down anyone else's recordings.

Every batch carries a session ID and a window ID, so ingestion can put the pieces from different tabs back together in order. Without these, recordings have gaps where batches went missing, and two tabs play back as one confusing mess.

Step 4: Storing it

Your support team gets a bug report on Monday about something that happened on Friday. The recording needs to still be there, and it needs to load in seconds despite being much larger than just the individual event in question.

How it works at PostHog: Ingestion buffers each session for up to 10 seconds, compresses it with Snappy, and writes it to object storage (S3) in blocks. A ClickHouse table keeps the metadata for each recording, such as who it was, which pages they visited, how many times they clicked, and any console errors, plus where each block lives. That's what makes recordings searchable without opening each one.

Recordings are stored by retention period and expire on their own. Free plans keep recordings for 30 days, and pay-as-you-go keeps them for 90, while platform packages enable you to go up to one year, or five years on Enterprise. You can also export a recording to JSON to keep it for longer.

To be fair, this is the cheap part. Object storage costs very little per GB, and recordings expire. If you build your own, storage is unlikely to be your biggest bill.

Without the metadata, finding one recording means opening thousands.

Step 5: Watching it

A recording is essentially a list of page changes, so something has to rebuild them into a page, in order, at the right speed, with the right fonts and styles. And whoever is watching wants to skip the 20 minutes when the user went to make a coffee.

How it works at PostHog: The player fetches compressed blocks straight from storage through short-lived signed links, decompresses them in your browser, and rebuilds the page with rrweb. Next to it, a DevTools-style panel shows console logs, network requests, and errors in sync with the recording. You can skip inactive time, change the speed, leave comments, and share or embed a recording for someone outside PostHog.

Watching is in the price. Your whole team can watch the same recording a hundred times, and the bill doesn't change.

Without it, you have a bucket of JSON and nothing to watch it with.

Step 6: Finding the one that matters

Nobody watches all their recordings. Despite capturing all that data what you really want isn't the five hours of them successfully using your product – it's the five seconds where they rage-clicked a button because something wasn't working as they expected. So, how do you find just that segment?

How it works at PostHog: Every recording is linked to the person, their events, their feature flags, and their errors. You can filter recordings by any event or property, watch the users who dropped off a funnel, and go from an error straight to the session behind it.

Additionally, if you bring in the full power of what PostHog offers, then PostHog AI can find recordings from a plain-English description, and Replay Vision reviews recordings for you and surfaces the moments that match what you're looking for.

Without these systems, you have thousands of recordings and no way to find the one that explains the bug.

If you build it yourself

Every one of those jobs still exists when you run replay yourself. This table shows what each one looks like on both sides. The second column is what we run for you, the third is what you'd run instead, and the last is when the job needs time from your team.

JobWhat PostHog runsWhat you run insteadWhen it needs your team
Recording itA web recorder, SDKs for iOS, Android, React Native, and Flutter, and sampling and triggers on the devicerrweb on the web, and a separate approach for each mobile platformWhen a browser or framework update changes how your pages render
Keeping it privateMasking on the device, header scrubbing, and an encryption key for each recordingMasking rules for every sensitive screen, and a way to delete one person's recordingsWhen someone ships a new form
Not losing itA capture endpoint for recordings, Kafka, and an overflow laneAn endpoint, a queue, and logic that puts batches from many tabs back togetherWhen traffic spikes or batches arrive late
Storing itCompressed blocks in object storage, metadata in ClickHouse, and expiry by retention periodA bucket, a metadata database, and a job that deletes expired recordingsWhen storage grows faster than you planned
Watching itA player with console, network, and error panels, inactivity skipping, comments, and sharingrrweb's player, plus the UI around itWhen the player can't rebuild a page your team just changed
Finding itFilters on events, people, flags, and errors, plus AI search and Replay VisionA link from each recording to your analytics, and a search UIWhen someone asks for "the users who hit this bug"

Costs that arrive later

A first build covers recording and playback. The costs below show up once it's in production.

The recorder is code in your product

You own its performance budget, its bundle size, and its bugs. Every browser release and frontend framework upgrade can change what it sees. Our January incident shows how that can go wrong, even for a team that works on replay full time.

Privacy is never finished

Every new screen your team ships needs someone to check what the recorder can see. A missed mask on a payment form is a security incident you can spend weeks dealing with.

The player

rrweb ships a basic player, but your team will want speed controls, inactivity skipping, the console log and network info next to the recording, comments, and share links. Every request is reasonable, and every one is more code to maintain.

A recording on its own tells you what one person did. Most of its value comes from finding it at the right moment: from the error, the funnel step, or the support ticket. That needs your recordings and your analytics to share user and session IDs, and someone to build the search. In PostHog, the events you already send do this job. Without this, it's another complete set of systems you need to build.

Time and trust

The rest is the same as for events. There are the hours your team doesn't spend on your product, and the controls your own build would need to match: SOC 2, HIPAA, and a public status page. Recordings show exactly what your users see, so they need at least as much care as your events.

Your own ledger

This is where you do the math. You'll want an engineer for the technical rows, but anyone can fill in the rest. If an agent is helping with your estimate, ask it to fill this in too and say where each number came from.

InputYour numberNotes
Recordings per monthYour billing page shows it
Compressed MB per recordingMeasure it on your real recordings. Canvas, network bodies, and long sessions change it a lot
Months retainedHow far back you need to watch. Pay-as-you-go keeps 90 days
Storage price per GB-monthYour cloud's object storage list price
Requests and egressEvery batch you write and every recording you play back
Ingestion compute and queueCapture, a queue, and the workers that assemble sessions
Metadata databaseWhat you search recordings with
Engineer hours per monthRecorder upgrades, masking reviews, and player fixes
On-callWho, and what it costs them

In plain English, work out how much you'll store by a given month and what that costs. Add the servers and the people, then divide by the number of recordings you make. As a formula:

GB stored at month N = recordings × MB × min(N, months retained) ÷ 1,000
storage at month N = GB stored × price per GB-month
monthly total = storage + requests + compute + queue + database + (engineer hours × loaded rate) + on-call
your cost per recording = monthly total ÷ recordings per month

Compare the last line with a PostHog bill at the same volume. Take the Session Replay total from the pricing calculator and divide it by your monthly recordings. The free 5,000 and the falling tiers mean that average comes out lower than the rates at the top of this page. If an agent is doing the math, point it at our pricing guide for agents.

Don't be surprised if storage is the smallest line. The ledger also leaves out the costs you can't put a price on, like a privacy miss, a slower checkout, and recordings nobody can find.

When building is the right call

Building your own replay product is the better choice for some teams. These are the cases where we'd build it too.

You only need a few recordings of one internal tool. rrweb and a storage bucket will do, and you can skip most of this page because you're neither handling personal data, requiring long-term storage, nor needing a UI to surface and poll it all.

Recordings can never leave your infrastructure. If you need a self-hosted product, you need a self-hosted product. The important thing to clarify however is whether a specific policy is in place for this – often teams may think they need a self-hosted product without being aware of what masking can do for them.

If you're not sure which side you're on, fill in the ledger.

Common questions

Does PostHog charge to watch or share a recording?

No. You pay once per recording. Watching, sharing, comments, and filters are included for everyone on your team. Replay Vision and AI search bill in their own credits.

Do mobile recordings cost more?

Yes. Mobile recordings have their own rates, which are higher than web rates, and their own free tier of 2,500 a month. The pricing page has both tables.

How long does PostHog keep recordings?

30 days on the free plan and 90 days on pay-as-you-go. Platform packages go up to one year, and Enterprise up to five. See recording retention.

Is it cheaper to self-host rrweb?

The recording library is free, and storing recordings is cheap. Most of the cost is the work around them: a recorder that's safe to ship, masking, a player, and search. The ledger works it out with your numbers.

How to lower your PostHog bill

If you want to keep using PostHog but pay less, these changes help most.

  • Record only the sessions you want. Sampling, a minimum duration, and triggers on URLs, events, or feature flags cut the count without losing the recordings that matter.

  • Set a billing limit so you know the most you'll pay in a month.

  • Pick the retention you need, not the longest one available.

If you'd like help with your setup, talk to us.

Still have questions?

Was this page useful?