When Google forced Universal Analytics into retirement in July 2023 and replaced it with GA4, it did something unintentional: it gave hundreds of thousands of developers a reason to look at alternatives. The migration was technically mandatory, the new interface was disorienting, and the data model had changed enough that historical comparisons were essentially impossible. Many teams spent weeks rebuilding dashboards, only to realize they were now tracking 200 metrics to answer the same three questions they had always cared about.

A meaningful number of those developers did not come back. Instead they found privacy-first analytics tools — Plausible, Umami, Fathom, PostHog — that offered genuinely simpler answers to the metrics that actually matter, with no cookies, no consent banners, and no GDPR compliance headaches. Having run Plausible on three production sites and self-hosted Umami for a fourth, my view is that the privacy-first category has crossed from niche to practical default for most developer projects and small products. Here is an honest accounting of what each tool offers and what it costs.

Why GA4 Burned So Much Developer Goodwill

Universal Analytics was not beloved, but it was understood. Developers had spent years building reports, setting up goals, and writing custom integrations around a stable data model. GA4 deprecated that entirely. Sessions, goals, and bounce rate were replaced with events, conversions, and engagement rate. The two models do not map cleanly onto each other, so historical data from Universal Analytics is not meaningfully comparable with GA4 data. Teams running year-over-year traffic analysis hit a hard discontinuity in their reporting.

The interface compounded the problem. GA4’s exploration reports require constructing queries from scratch — choosing dimensions, metrics, and filters through a multi-step UI that feels designed for analysts with dedicated training time, not developers who want a quick answer on a Tuesday morning. The standard reports are less useful than UA’s equivalent views, and the custom reports require enough familiarity with the GA4 data model that most teams just stopped using them.

There is also the privacy architecture. Universal Analytics set cookies that persisted for up to two years and used cross-site tracking methods that regulators have increasingly scrutinized. Under GDPR and the ePrivacy Directive, sites using these cookies need explicit consent before setting them. That requirement spawned the consent banner industry — those pop-up overlays that interrupt the user experience on nearly every website in the EU. Research from the Norwegian Consumer Council and others has documented how dark patterns in consent banners systematically steer users toward accepting tracking. Regulators in France, Italy, and Germany have all issued significant fines for non-compliant consent flows. The compliance burden falls on the site operator, not on Google.

Google Analytics 4 uses a different tracking mechanism than UA, but it still relies on user consent under GDPR in most implementations, particularly when using the Google Signals feature or when data is combined with advertising data. The legal position is not clean, and the documentation is not forthcoming about the edge cases.

The irony of GA4’s launch is that it created enough friction that developers started asking whether they needed Google Analytics at all. The answer, for most of them, turned out to be no.

What Privacy-First Analytics Actually Means

The term gets used loosely, so precision is worth establishing. A genuinely privacy-first analytics tool has three defining characteristics.

First, it does not set cookies or use persistent client-side storage to identify individual users across sessions. Instead of a cookie with a long-lived user ID, tools like Plausible and Umami generate a daily rotating hash of the IP address and user agent. This hash cannot identify a specific person, changes every 24 hours, and is not stored anywhere that a third party could access. It is sufficient to deduplicate page views within a day without enabling individual-level tracking.

Second, the data is not shared with third parties for advertising purposes. GA4 data feeds into Google’s advertising ecosystem by default, with opt-out possible but not the default configuration. Privacy-first tools have no advertising product and no financial incentive to pass data onward.

Third, because no personally identifiable information is collected and no consent-requiring cookies are set, these tools are exempt from the consent requirements that make cookie banners necessary. A site using Plausible or Umami can remove its consent popup entirely — legally, under GDPR — because there is nothing that requires consent. This is not a workaround; it reflects how the law defines the consent requirement. Consent is triggered by processing personal data. If the analytics tool does not process personal data, consent is not required.

The script size difference is also material in practice. Google’s analytics script loads approximately 45KB of JavaScript — a meaningful page weight penalty, particularly on mobile connections. Plausible’s tracking script is around 1KB. Umami’s is similarly lightweight. On a site with 100,000 monthly page views, the cumulative bandwidth difference across all those requests is non-trivial, and faster page loads directly affect search ranking through Core Web Vitals.

Plausible Analytics: The Hosted Simplicity Option

Plausible Analytics is a bootstrapped, EU-based company with a transparent business model and a product that is genuinely well-designed for developers who want traffic metrics without complexity. The dashboard shows pageviews, unique visitors, bounce rate, session duration, top pages, referral sources, UTM campaigns, countries, browsers, and operating systems — everything useful on a single screen, no configuration required.

The hosted service starts at $9 per month for up to 10,000 monthly page views, billed annually at $90. Pricing scales to $19/month for 100,000 views, $29/month for 200,000 views, and $69/month for 1 million views. These figures are flat per-site, with unlimited additional sites on any plan from the $19 tier upward. For a developer with several projects that collectively generate under 100,000 monthly page views, the cost is competitive with a single managed analytics product.

The codebase is also open source, which means Plausible can be self-hosted. The self-hosted version requires running a PostgreSQL database and a ClickHouse instance — the latter being a columnar database that handles the high write volumes analytics workloads generate. ClickHouse is not trivial to operate; the hosted version eliminates that concern entirely.

Plausible’s limitations are real and worth naming. There is no user-level tracking, by design. You cannot follow an individual user through a session, cannot segment cohorts by behavior over time, and cannot build the kind of funnel analysis that shows you where specific users dropped out of a conversion flow. The goal tracking is event-based and requires some configuration, but it cannot replicate the conversion path analysis that Mixpanel or a product analytics tool provides. If your job is conversion rate optimization on a complex funnel with multiple steps and variations, Plausible will tell you there is a problem but not enough about why.

For the majority of developer projects — blogs, documentation sites, SaaS marketing pages, small products — Plausible provides everything actually needed. Most sites have three or four questions: how many visitors, where are they coming from, what pages are they reading, is traffic growing or declining. Plausible answers all of those without configuration, customization, or any significant learning curve.

Umami: The Self-Hosting Path

Umami is an open-source privacy-respecting analytics tool with a feature set roughly comparable to Plausible. It provides pageviews, unique visitors, session data, referrers, device categories, and custom event tracking, all in a clean interface that takes about fifteen minutes to understand completely. The codebase is Node.js with a PostgreSQL backend, and deployment on a standard Linux VPS is well-documented.

The hosting economics look compelling on paper. If you are already running a Hetzner CX22 ($7/month) for other services, Umami adds negligible overhead — it is not resource-intensive at moderate traffic volumes. The data lives on your own infrastructure, under your control, with no subscription cost and no per-pageview pricing ceiling. For a developer running four or five sites on shared infrastructure, the incremental cost of adding Umami is effectively zero.

The reality of self-hosting Umami is more nuanced. Initial setup takes roughly two to three hours for someone comfortable with Docker Compose and Nginx. The ongoing maintenance burden is lower than Plausible’s self-hosted stack because Umami uses standard PostgreSQL rather than ClickHouse — but it is not zero. Database backups need to be configured and verified. Package updates need to be applied. When Umami releases a new version, the upgrade path requires testing. Over a year, the realistic maintenance overhead is four to eight hours for a small deployment.

At $75/hour opportunity cost, those eight hours cost $600. Plausible’s $90/year hosted plan costs $90. The math only favors Umami self-hosting when you are already running infrastructure for other reasons and the marginal maintenance time is genuinely low — or when your pageview volumes push Plausible’s pricing above $19/month.

Umami also offers a cloud-hosted tier, now branded as Umami Cloud, starting at free for up to 10,000 monthly events. The paid plans are comparable to Plausible’s pricing. For teams that want Umami’s feature set without operational overhead, the cloud option removes the self-hosting trade-off.

Fathom Analytics and PostHog: The Other End of the Spectrum

Fathom Analytics positions itself as the premium end of the privacy-first market. At $14 per month for up to 100,000 monthly page views, it is priced above Plausible but offers a specific differentiator: EU-isolated infrastructure. Fathom routes all EU visitor data through servers based in the EU, and offers what it calls “EU isolation” as a built-in compliance feature. This addresses a specific anxiety some European customers have about data leaving EU jurisdiction under any circumstances. For companies with explicit EU data residency requirements in customer contracts, Fathom’s architecture removes a conversation that Plausible — despite being EU-based — cannot resolve in the same way because its CDN has global nodes.

Fathom’s feature set is comparable to Plausible’s: clean dashboard, referrers, UTM tracking, uptime pages, email reports. The interface is arguably more polished. The trade-off is purely economic — you pay more for the infrastructure guarantee.

PostHog occupies a different category entirely. It is not simply privacy-first analytics; it is a product analytics platform that happens to have a strong privacy stance. The open-source edition includes pageview analytics, event tracking, funnels, cohort analysis, session recording, feature flags, and A/B testing — a feature surface that competes with the combination of Mixpanel, LaunchDarkly, and Hotjar. PostHog Cloud offers a generous free tier of one million events per month before billing starts.

The privacy architecture is sound: PostHog can be fully self-hosted, and even the cloud version offers data deletion, data export, and compliance tooling. Session recording can be configured to automatically redact input fields. The company is transparent about its data practices in a way that Google Analytics is not.

PostHog’s complexity is real, though. The onboarding experience requires decisions — which products to enable, how to instrument events, how to configure person properties — that Plausible never asks of you. For a marketing site or a blog, that complexity is overhead without commensurate benefit. For a SaaS product where you want to understand activation rates, feature adoption, and conversion funnels, PostHog’s feature depth is genuinely useful and its cost structure is competitive with commercial alternatives at comparable functionality.

Feature Comparison

The Metrics That Actually Matter

One under-discussed reason the privacy-first category resonates with experienced developers is that it forces clarity about what you actually need to measure. GA4, used fully, can track hundreds of event types, dozens of dimensions, and deeply segmented audiences. Most of that capability goes unused by 95% of sites — but its existence creates a gravitational pull toward complexity. Teams spend engineering time on tagging implementations, data layer schemas, and custom event hierarchies for data that no one ever queries.

For the overwhelming majority of developer projects, the analytically meaningful questions are: how many people visited, where did they come from, what did they read, and is the trend moving in the right direction? Plausible and Umami answer all of those directly, without configuration, without an analytics engineering function, and without a 60-slide onboarding deck. The constraint of what these tools can measure is not primarily a limitation — it is a forcing function toward intentionality.

The cases where GA4 or PostHog’s depth genuinely earns its complexity are specific. Ecommerce with multiple conversion paths, where funnel analysis across channels identifies real revenue leakage. SaaS products where activation analysis and cohort retention reveal which features drive long-term retention. Any context where user-level behavior over time is the analytical object of interest. Below that threshold of complexity, you are paying in cognitive overhead and compliance burden for capability you will never use.

The Real Cost of Self-Hosting Umami

The self-hosting decision deserves specific numbers rather than hand-waving about “operational overhead.” Here is an honest breakdown for a solo developer running Umami on existing VPS infrastructure.

• Initial setup (Docker Compose + Nginx + SSL): 2–3 hours
• Configuring automated PostgreSQL backups: 1 hour
• Monthly maintenance (package updates, backup verification): 30–45 minutes
• Version upgrades (3–4 per year, minor): 30 minutes each
• Incident response (estimate 0–1 per year for a stable setup): 1–3 hours
• Annual total: approximately 12–20 hours

At a conservative $75/hour for developer time, that is $900–$1,500 per year in time cost. Against Plausible’s $90/year hosted plan for sites under 100,000 monthly page views, self-hosting Umami is financially rational only when the marginal time cost is genuinely near zero — which it is when you are already maintaining a VPS for other purposes and Umami is one of several services running on the same box.

The crossover point where self-hosting clearly wins on economics is when pageview volumes push Plausible’s pricing into the $69/month tier (1 million monthly views). At that level, the $828/year for Plausible versus the server costs of a dedicated ClickHouse instance starts to favor self-hosting even with realistic time accounting. Below that threshold, the honest answer for most developers is that the hosted Plausible subscription is cheaper when time is valued appropriately.

The non-economic case for self-hosting is data sovereignty: the data lives on your hardware, under your administrative control, accessible by no third party without your consent. For some applications — particularly those handling sensitive user contexts even without PII — that argument is compelling independent of the economics.

Migrating from GA4 Without Losing Historical Context

The transition from GA4 to a privacy-first tool has one non-negotiable reality: you cannot migrate historical data. GA4’s data is proprietary, stored in Google’s infrastructure, and not exportable in a format that Plausible, Umami, or Fathom can ingest. The historical record of your traffic before the migration date will only exist in GA4, and GA4 will eventually retire that data access as well.

The practical migration approach has four steps.

Step one: run both in parallel for 30 days. Install your new privacy-first tool alongside GA4 during a 30-day overlap period. This lets you validate that the new tool’s numbers are reasonable — privacy-first tools typically report 10–20% more unique visitors than GA4 because they are not filtered by consent refusals and do not use sampling on low-traffic sites. Understanding this difference before you remove GA4 prevents misinterpreting the new baseline as a traffic change.

Step two: export and archive GA4 data. Before cutting over, use GA4’s BigQuery export (available on free accounts) or the GA4 Data API to pull your most important historical reports and store them as CSV or Parquet files. Document the export date, the metrics definitions used, and the fact that post-migration data comes from a different tool with a different counting methodology. Future you will want this context.

Step three: configure your new tool’s goals before removing GA4. Map your GA4 conversion events to equivalent goal configurations in Plausible or Umami. Plausible’s custom events use a simple JavaScript API. Test the event firing against a staging environment before deploying to production.

Step four: remove the GA4 script and the consent banner. Once you have removed GA4 from your site, audit every page for consent management platform code and cookie banner scripts. If Google Analytics was the only tool requiring consent, these can all be removed. Removing consent infrastructure eliminates 30–150KB of JavaScript on most sites and meaningfully improves page load performance.

The loss of GA4’s historical data is real but often overstated as a migration barrier. Most teams, when pressed, need historical trend data for the last 12–18 months. Anything older is queried occasionally and at low analytical value. Archive it, document the methodology difference, and move on.

Cookie Banners: Why You Can Remove Them

This is the single most underappreciated operational benefit of switching to privacy-first analytics. Under GDPR and the ePrivacy Directive, the consent requirement is triggered by collecting personal data or by setting cookies that are not strictly necessary for the service to function. Analytics cookies fall into the latter category, which is why every site using GA4 needs a consent mechanism.

Plausible, Umami, and Fathom do not set cookies. They do not collect personal data — the daily-rotating hash derived from IP and user agent does not constitute personal data under GDPR because it cannot identify a natural person either directly or indirectly. The EU’s data protection authorities have confirmed this approach in multiple guidance documents. Plausible has received legal review from GDPR specialists and publishes the analysis publicly.

The implication is direct: sites using only these tools as their analytics layer can remove consent banners entirely. No popup, no overlay, no cookie preference center. This has a measurable impact on user experience. Studies on e-commerce sites have documented 1–3% improvements in conversion rate from removing consent friction at the entry page. For a site doing $500,000 in annual revenue, that range represents $5,000–$15,000 in additional converted revenue — far more than any analytics subscription costs.

One caveat: if your site also uses third-party advertising pixels, remarketing scripts, or marketing automation tools that set cookies or collect identifiable data, those tools independently trigger the consent requirement. Removing GA4 while keeping a Facebook Pixel does not eliminate the consent obligation. The analysis needs to cover all third-party scripts, not just analytics.

Choosing the Right Tool for Your Context

The decision among these tools is not particularly complicated when you match the tool’s capabilities to your actual analytical requirements.

Use Plausible if you want traffic metrics with zero configuration overhead, you value a bootstrapped privacy-respecting company, and your pageview volume keeps you in the $9–$19/month tier. It is the right default for most developer projects.

Use Umami (self-hosted) if you are already running your own infrastructure, data sovereignty is a priority, and you are comfortable absorbing the setup and maintenance time. The economics only favor it meaningfully when your hardware is already there.

Use Fathom if your customer contracts include explicit EU data residency requirements and you want hosted infrastructure with that guarantee built in. The $14/month premium over Plausible is reasonable for the specific compliance assurance it provides.

Use PostHog if you are building a SaaS product and need product analytics — funnels, cohort retention, session recording, A/B testing — not just traffic metrics. Its privacy architecture is solid, and the free tier for up to one million events monthly covers most early-stage products completely.

The common thread across all four is that the analytics complexity of Google Analytics 4 is not a feature most developer projects need. The privacy-first category exists because the requirements of most sites — know your traffic, know your sources, know whether you are growing — are simpler than GA4’s answer to them, and because privacy compliance is much easier to achieve when the tool’s architecture makes it the default rather than a configuration option.

If your site is still running GA4 because migrating feels like a project, the actual work is two hours: install Plausible, run it in parallel for 30 days, archive your GA4 data, remove the GA4 script and the consent banner. The compliance burden drops immediately. The page performance improves on the first deploy. The analytical picture you get from Plausible will answer the questions you actually ask about your traffic — and unlike your current GA4 dashboard, you will know how to read it without consulting documentation.