# Build vs buy a game analytics pipeline

> What a game analytics pipeline needs, a cost table from public cloud prices with the arithmetic, the cost in time, and when a small studio should buy a tool.

By Yos Riady. Updated 2026-11-19. Published 2026-11-18. Canonical URL: https://manalens.dev/blog/build-vs-buy-game-analytics-pipeline

Build your own game analytics pipeline only if you need full control and have engineering time to spare. Cloud fees are small: we estimate about $2.71 in the first month for 30 million events on Google Cloud, at the public prices of 19 November 2026. The SDK, the collector, identity, retention math and dashboards cost far more in time.

## Key points

- Our estimate of cloud fees for 30 million events a month on Google Cloud is $2.71 in month 1 and $10.30 in month 12.
- Engineering time is the real cost of a pipeline, since the SDK, collector, retention SQL and dashboards need work every month after version one.
- Build your own pipeline only if you must join game events with server data or keep data in systems that no vendor supports.

## What does a game analytics pipeline need?

Five parts: an SDK in the game, a collector that accepts events, storage, models that turn events into metrics, and dashboards. A warehouse covers only storage and queries. The table lists what each part must handle for a PC game.

- [Offline event queue](https://manalens.dev/glossary/offline-event-queue)
- [Identity stitching](https://manalens.dev/glossary/identity-stitching)

*The five parts of a game analytics pipeline.*

| Part | What it must handle |
| --- | --- |
| SDK | Offline queue, retries, sessions, crash detection on the next launch, consent, editor runs |
| Collector | Keys, validation, duplicate events, bots, rate limits |
| Storage | Events by project, date and build, with deletes for privacy requests |
| Models | Retention by day, sessions, identity across devices, revenue in one currency |
| Dashboards | Funnels, cohorts, filters by build and storefront, Steam data next to game data |

Sources: Manalens Godot SDK, Unity SDK and app events documentation.

## What does building one cost in cloud fees?

A few dollars a month at indie scale. The table is our estimate, not a quote. It uses the public list prices that we read on 19 November 2026 and one scenario: 5,000 daily players, 200 events for each player each day, about 1 KB for each event, and batches of 20 events. That is 30 million events, 1.5 million requests and about 30 GiB a month. Replace these numbers with yours.

The total is about $2.71 in month 1 and about $10.30 in month 12, when 360 GiB is in storage. The month 12 number is a high estimate, because BigQuery bills data that did not change for 90 days at a lower long-term rate. Taxes, support plans, network transfer and a dashboard tool are not in the table.

*Estimated monthly cloud fees for 30 million events on Google Cloud. Public list prices in US dollars, read 19 November 2026. Our arithmetic.*

| Part | Service and public price | Arithmetic | Estimate for a month |
| --- | --- | --- | --- |
| Collector | Cloud Run, request-based billing, us-central1: $0.40 per million requests, $0.000024 per vCPU-second, $0.0000025 per GiB-second. Free each month: 2 million requests, 180,000 vCPU-seconds, 360,000 GiB-seconds. | 1.5 million requests × 0.1 second at 1 vCPU and 0.5 GiB = 150,000 vCPU-seconds and 75,000 GiB-seconds. All of it is inside the free tier. At list price: $0.60 + $3.60 + $0.19 = $4.39. | $0 ($4.39 without the free tier) |
| Queue | Pub/Sub: $40 per TiB of throughput after the first 10 GiB. A BigQuery subscription: $50 per TiB, with no free amount. | Publish: (30 - 10) GiB × $40 ÷ 1,024 = $0.781. Write to BigQuery: 30 GiB × $50 ÷ 1,024 = $1.465. Sum: $2.246. | $2.25 |
| Storage | BigQuery active logical storage: the first 10 GiB is free. Google's example is $23.552 for 1 TiB for a full month in us-central1, or $0.023 for 1 GiB. | Month 1: (30 - 10) GiB × $0.023 = $0.46. Month 12: (360 - 10) GiB × $0.023 = $8.05. | $0.46 in month 1, $8.05 in month 12 |
| Queries | BigQuery on-demand: $6.25 per TiB, and the first 1 TiB each month is free. | A full scan of 12 months of data is 360 GiB. Two full scans a month are 720 GiB, inside the free 1 TiB. | $0 |
| Total | Cloud Run, Pub/Sub and BigQuery | Month 1: $0 + $2.25 + $0.46 + $0. Month 12: $0 + $2.25 + $8.05 + $0. | $2.71 in month 1, $10.30 in month 12 |

Sources: Google Cloud pricing pages for Cloud Run, Pub/Sub and BigQuery, read 19 November 2026.

## Do the fees change on another cloud?

Not by much. The second table prices the collector, the queue and a raw event archive on Cloudflare for the same scenario. The estimate is $6.70 in month 1, and most of it is the $5 minimum charge of the Workers Paid plan. R2 stores files, so you still need a query engine and dashboards, which this table does not price.

So on both clouds the bill for 30 million events is less than $11 a month. The cloud bill is rarely the reason to build or to buy.

*Estimated monthly fees for the same 30 million events on Cloudflare. Public list prices in US dollars, read 19 November 2026. Our arithmetic.*

| Part | Service and public price | Arithmetic | Estimate for month 1 |
| --- | --- | --- | --- |
| Collector | Workers Paid: a minimum charge of $5 a month, with 10 million requests included, then $0.30 per million. | 1.5 million requests are inside the included 10 million. | $5.00 |
| Queue | Queues: 1 million operations a month included, then $0.40 per million. A message takes 3 operations: write, read and delete. | One message for each batch: 1.5 million × 3 = 4.5 million operations. (4.5 - 1) million × $0.40 = $1.40. | $1.40 |
| Raw archive | R2 standard storage: $0.015 per GB-month, with 10 GB-month and 1 million Class A operations free each month. | (30 - 10) GB × $0.015 = $0.30. One file a minute is 43,200 writes a month, inside the free operations. | $0.30 |
| Total | Workers, Queues and R2 | $5.00 + $1.40 + $0.30 | $6.70 |

Sources: Cloudflare pricing documentation for Workers, Queues and R2, read 19 November 2026.

## What does building one cost in time?

Much more than the cloud fees, and no price list gives the number. Count the hours to write and test an SDK for each engine, a collector, the SQL for retention and funnels, and a dashboard. Then count the hours each month to fix and update them.

Use this formula for the first year: (build hours + 12 × monthly hours) × your hourly cost + 12 months of cloud fees. Compare it with a subscription. With 5,000 daily players, the scenario has more than 10,000 MAU if fewer than half of its monthly players play on a given day. For example, the Manalens Scale 20K plan is $199 a month for up to 20,000 MAU, so 12 × $199 = $2,388 a year. Twelve months at the month 12 estimate of $10.30 are $123.60 of cloud fees. So building is cheaper in year one only if your hours cost less than $2,388 - $123.60 = $2,264.40 in total. That is an estimate with our scenario, not a rule.

- [What is MAU?](https://manalens.dev/glossary/mau)

Sources: Manalens pricing; our arithmetic from the tables above.

## When should a studio build its own pipeline?

Build when the data must live in your own systems or join data that no tool sees. The typical cases are listed below. If none of these apply, the time you spend on a pipeline is time you do not spend on the game or its marketing.

- You already run a data warehouse and a data engineer for a backend or a live game.
- You must join game events with server data, such as matchmaking or a custom store.
- You have rules about where data lives that no vendor meets.
- You need metrics or models that no tool offers, and you will maintain them.

Sources: Manalens documentation.

## When should a studio buy a tool?

Buy when the team is small and the questions are standard: installs, retention, funnels, crashes and sales. Most indie studios are in this group. The table shows the public prices of three game analytics tools. General product analytics tools, such as PostHog, Mixpanel and Amplitude, are a fourth option.

Manalens fits premium PC games on Steam and itch.io. It has Godot and Unity SDKs, game KPIs, a Steam wishlist and sales sync (the automatic Steam sync is in early access) and install attribution in one project. It has no free plan and no Unreal SDK.

- [Game analytics pricing compared: MAU vs event based plans](https://manalens.dev/blog/game-analytics-pricing-mau-vs-events)
- [PostHog, Mixpanel or Amplitude for games?](https://manalens.dev/blog/posthog-mixpanel-amplitude-for-games)
- [Best game telemetry tools in 2026](https://manalens.dev/blog/best-game-telemetry-tools)

*Public prices of three game analytics tools, read 19 November 2026.*

| Tool | Public price | Free plan |
| --- | --- | --- |
| Manalens | Indie: $79 a month for up to 10,000 MAU. Scale: $199 to $999 a month for up to 100,000 MAU. No overage charges. | No. A 7-day trial. |
| GameAnalytics | Paid plans add features to the free plan | Yes, with no MAU cap |
| Unity Analytics | After 50,000 MAU: from $0.00360 per MAU | Yes, up to 50,000 active users a month |

Sources: Manalens pricing; GameAnalytics pricing; Unity Gaming Services pricing.

## Can I buy now and keep my data portable?

Yes, if the tool lets you query and export. Choose a tool with SQL access, CSV export and an API, and keep your event names in a tracking plan that you own. Then a later move is a mapping job, not a rewrite.

In Manalens, the Explorer runs SQL on your raw events, each dashboard chart exports to CSV, and the analytics API and MCP server read the same data. Events from your own server can come in through the server API. For an engine without an SDK, the HTTP envelope is plain JSON.

- [Data export](https://manalens.dev/glossary/data-export)

``` HTTP
curl -sS https://YOUR_COLLECTOR_ENDPOINT/v0/events \
  -H 'x-publishable-key: YOUR_PUBLISHABLE_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "events": [{
      "type": "track",
      "event": "level_complete",
      "channel": "app",
      "version": "0.1.0",
      "message_id": "6a0f8e9c-2c36-4c55-9a43-6a5d0f3c2b11",
      "anonymous_id": "3f0c5c1e-8a8e-4c1b-9d7e-2b1f4e6a9c10",
      "session_id": "1762689600000",
      "original_timestamp": "2026-11-09T10:00:00.000Z",
      "context": {
        "app": { "platform": "windows", "storefront": "steam", "app_version": "1.4.0", "build": "571", "engine": "godot" },
        "locale": "en",
        "library_name": "my-game",
        "library_version": "1.4.0"
      },
      "properties": { "level": 3, "score": 1200 }
    }]
  }'
```

Sources: Manalens Explorer, dashboards, MCP server and app events documentation.

## Which hidden costs do teams forget?

The work after version one. These items take time every month, whether you build or not, but a vendor carries them for you.

The SDK alone has many rules. For example, the Unity SDK keeps unsent events in a queue, reports events that it drops because the queue is full or the events are older than 7 days, and in consent mode deletes the ID, the session and the queue when a player says no. Each rule is work you would write and test yourself.

- New engine versions and platforms, such as a Steam Deck or web build.
- Duplicate events from retries, which need a message ID and a dedupe step.
- Clock errors on player machines, which put events on the wrong day.
- Privacy requests that need a delete across every table.
- Changes to Steam reports and APIs that break an import.
- On-call time when the collector stops during a launch.

Sources: Manalens SDK and privacy documentation.

## What does a build plan look like?

If you decide to build, scope the first version tightly and ship it before launch, not during it.

1. Write a tracking plan of 10 to 20 events with their properties.
2. Build the SDK for one engine first, with an offline queue, retries and a unique ID per event.
3. Build a collector that checks a key, validates events and drops duplicates.
4. Load events into a warehouse table partitioned by date. Batch loads into BigQuery are free on the shared slot pool.
5. Write and test the SQL for DAU, D1, D7 and a first session funnel.
6. Add a dashboard, then a delete job for privacy requests.

- [Game tracking plan generator](https://manalens.dev/tools/game-tracking-plan-generator)

Sources: Manalens Godot SDK, Unity SDK and app events documentation; Google Cloud BigQuery pricing.

## What does buying cost for a small studio?

Manalens is the analytics and growth engine built for games. Plans are priced by monthly active players only, from $79 a month for up to 10,000 MAU, with a 7-day trial and no overage charges.

- [Pricing](https://manalens.dev/pricing)

Sources: Manalens product documentation.

## Frequently asked questions

### Is BigQuery enough for game analytics?

For storage and queries, often yes, and it is cheap at indie scale. It does not give you an SDK, a collector, retention models or game dashboards. You build and maintain those yourself, or connect other tools to it. Count that time before you choose.

### Can I move away from a tool later?

Yes, if you plan for it. Keep your own tracking plan, choose a tool with SQL access, CSV export and an API, and avoid names that only one vendor understands. Then you can rebuild the same charts elsewhere from your event names and properties.

### Can I send events from my own game server?

Yes. Manalens has a server API and a Node SDK for events from your backend, sent with the secret key. Never ship the secret key in a game build; builds use the publishable key, which can send events but cannot read data. Both kinds of events land in the same project.

### Is building cheaper than a subscription?

In cloud fees, yes at small scale: our estimate is $2.71 to $10.30 a month for 30 million events on Google Cloud. In total cost, usually no for a small team, because the engineering time to build and maintain the pipeline costs more than the fees you save.

### How exact is the cost table?

It is an estimate. The prices are public list prices in US dollars for one region, read on 19 November 2026, and cloud prices change. The scenario of 30 million events a month is ours. Put your own players, events and event size into the same arithmetic before you decide.

## Sources

- [Google Cloud: Cloud Run pricing (read 19 November 2026)](https://cloud.google.com/run/pricing)
- [Google Cloud: Pub/Sub pricing (read 19 November 2026)](https://cloud.google.com/pubsub/pricing)
- [Google Cloud: BigQuery pricing (read 19 November 2026)](https://cloud.google.com/bigquery/pricing)
- [Cloudflare documentation: Workers pricing (read 19 November 2026)](https://developers.cloudflare.com/workers/platform/pricing/)
- [Cloudflare documentation: Queues pricing (read 19 November 2026)](https://developers.cloudflare.com/queues/platform/pricing/)
- [Cloudflare documentation: R2 pricing (read 19 November 2026)](https://developers.cloudflare.com/r2/pricing/)
- [GameAnalytics: Pricing (read 19 November 2026)](https://www.gameanalytics.com/pricing)
- [Unity: Unity Gaming Services pricing (read 19 November 2026)](https://unity.com/products/gaming-services/pricing)

## Related guides

- [Game marketing measurement guide](https://manalens.dev/blog/game-marketing-measurement-guide)
- [Game analytics pricing compared: MAU vs event based plans](https://manalens.dev/blog/game-analytics-pricing-mau-vs-events)
- [PostHog, Mixpanel or Amplitude for games?](https://manalens.dev/blog/posthog-mixpanel-amplitude-for-games)
- [Game analytics with AI agents and MCP](https://manalens.dev/blog/game-analytics-ai-agents-mcp)
- [Game tracking plan generator](https://manalens.dev/tools/game-tracking-plan-generator)
- [Player analytics guides](https://manalens.dev/blog/topic/player-analytics)
- [How to move from GameAnalytics to Manalens](https://manalens.dev/blog/migrate-from-gameanalytics)
- [Best game telemetry tools in 2026](https://manalens.dev/blog/best-game-telemetry-tools)
- [Best analytics tools for three.js games in 2026](https://manalens.dev/blog/best-threejs-analytics-tools)
- [How to add analytics to a Godot game](https://manalens.dev/blog/godot-analytics)
