Q3 Stack Review: Tools Yang Dipakai vs Yang Cuma Dibayar
Q3 stack audit for small Indonesian dev teams — an honest look at which tools get used daily and which ones just charge your card.
Every Q3, the same realization hits: your team is paying for eight tools, actively using four, and tolerating three others because canceling feels like admin work.
This is a Q3 stack audit from the ground level — not “best tools of 2026” listicle energy, but a practical look at what small Indonesian dev teams (3-8 people, usually somewhere between Tangerang and BSD Bintaro Jaya in terms of budget expectations) are actually running versus what they signed up for six months ago.
The Audit Frame
The question is not “is this tool good?” It is “does anyone on your team open this tab before 10 AM voluntarily?” That is a rougher filter, but it is more honest.
Tools in a small team stack fall into three buckets after a quarter:
- Daily-use: opened without being prompted, integrated into workflow, someone notices immediately if it goes down
- Reactive-use: opened when there is a problem or a specific task, but not part of daily rhythm
- Zombie tools: auto-renewing, occasionally justified in planning meetings, rarely touched
The zombie category is where Q3 budget bleeds.
Observability: Sentry vs Datadog
Sentry is daily-use for teams that have someone owning error triage. The alert-to-Slack pipeline, grouped issues by stack trace, release tracking — it does what it promises. The free tier (5k errors/month) covers most MVPs. Team plan at USD 26/month is reasonable if three or more people are actively working from it.
The honest problem: most teams install Sentry, see errors flood in for the first week, get overwhelmed, and quietly mute the Slack channel. At that point it becomes a zombie tool with a USD 26 monthly subscription. If no one has “Sentry triage” as part of their weekly rhythm, you are not buying observability — you are buying guilt.
Datadog is in a different category. It is a genuine enterprise tool that small teams often adopt because a senior engineer used it at their previous job. The UX is excellent. The billing is not.
Datadog charges per host, per APM service, and per GB of log ingestion — separately. A team of four running three services can easily land at USD 250-400/month before realizing it. For context, that is a full VPS setup at Rp150,000/month running Grafana + Prometheus + Loki for essentially the same visibility into your own infra.
Q3 verdict on observability: Keep Sentry if someone owns triage. Replace Datadog with Grafana + Prometheus on your VPS unless your team is already past 10 engineers.
Project Management: Linear vs the Field
Linear became the default for product-focused small teams for a reason: it is fast, opinionated, and does not let you configure yourself into paralysis. Keyboard shortcuts work. Cycle management is clean. The free tier covers 250 issues, which is enough for a team of four for about six months of active sprints.
The caveat is that Linear’s value compounds only when the team actually closes cycles and reviews them. Teams that use Linear as a fancy Trello — drag cards around, never run retrospectives — are leaving half the tool on the table.
Notion is a separate case. Almost every small Indonesian dev team has a Notion workspace. Almost none of them agree on what Notion is for. Engineering specs? Meeting notes? Company wiki? HR docs? When a tool has to be everything, it usually becomes the place where things go to be forgotten. The Plus plan at USD 10/user/month for eight people is USD 80/month for a shared inbox nobody curates.
If your Notion has pages last edited in April, you are paying for an archive. Either nominate someone to own it (actual ownership, not “everyone should update it”) or cut to free tier and treat it as public-facing docs only.
Q3 verdict on project management: Linear is worth the upgrade to USD 8/user only if you run real cycles. Notion needs an owner or it needs a tier downgrade.
Hosting and Infra: Vercel vs Railway vs VPS
Vercel is still the fastest path from git push to live URL. For Next.js projects, the integration is frictionless. The free tier is generous enough that many teams do not think about it until they scale.
That thinking-about-it moment tends to arrive as a surprise. Bandwidth on Vercel scales with traffic, and the jump from Hobby to Pro (USD 20/month per team, plus bandwidth) catches teams off guard. Edge function invocations are metered separately. If you are running a marketing site or a content blog, Vercel free is fine. If you are running an API backend through Vercel functions, watch the meter.
Railway at USD 5/month base (pay-as-you-go) has become the sane default for backend services in small Indonesian teams. Predictable billing, simple deployment, good enough DX for Node/Go/Python services. It is not Heroku 2012 (simpler) or AWS (more complex) — it sits in the right spot for teams that want to ship without a dedicated DevOps person.
The real-world comparison: a three-service backend on Railway runs around USD 15-25/month depending on traffic. The same setup on Vercel Functions would be unpredictable. The same on EC2 would require someone to manage nginx, SSL certs, and deployment scripts — which is a valid path, but not if your team is three developers and two of them are full-time on feature work.
VPS (DigitalOcean, Vultr, or increasingly IDCloudHost for billing in IDR) makes sense when you self-host things like Posthog, Grafana, or Coolify. A Rp200,000/month droplet running Coolify can replace USD 80/month in Vercel Pro + Railway + partial Datadog costs if your team is willing to maintain it. Some teams are; most are not. Be honest about which kind of team you are before defaulting to “we should self-host everything.”
Analytics: Posthog as the Q3 Standout
If there is one tool that moved from “reactive-use” to “daily-use” for small teams this quarter, it is Posthog. The combination of product analytics, session recordings, feature flags, and A/B testing in a single self-hostable package changed the calculation.
Previously, a team would have Mixpanel for events, Hotjar for session recording, and maybe LaunchDarkly for feature flags — three bills, three integrations, three admin logins. Posthog self-hosted on a VPS covers all three for the cost of the VPS.
The gotcha: Posthog self-hosted requires maintenance. Schema migrations on upgrade, occasional Clickhouse memory tuning, and someone who is comfortable reading Docker logs when something breaks. For a team in Serpong Utara running at lean margin, this is a reasonable trade. For a team that has never self-hosted anything, the Posthog Cloud free tier (1M events/month) is the better starting point.
The Honest Q3 Summary
| Tool | Typical Usage | Q3 Action |
|---|---|---|
| Sentry | Reactive-use (often zombied) | Keep if you triage weekly, else cut |
| Datadog | Zombie for teams < 10 people | Replace with Grafana + Prometheus |
| Linear | Daily-use when cycles run properly | Worth USD 8/user if you use cycles |
| Notion | Zombie wiki for most teams | Cut to free or assign a real owner |
| Vercel | Daily-use for frontend deploys | Audit function billing every 30 days |
| Railway | Daily-use for backend services | Strong default for small teams |
| Posthog | Daily-use if self-hosted right | Best Q3 addition for product teams |
The pattern across all of these: tools do not become zombie tools because they are bad. They become zombie tools because adoption was not finished. Installing a tool is not using a tool. If your team cannot name one person who “owns” a given tool — who checks it, acts on it, and would notice if it went down — that tool is probably in the zombie bucket.
Cancel it, or actually adopt it. Tolerating it is just paying for the option to use it someday.
Ditulis oleh Asti Larasati