Features on Feasible

The one-page dashboard

Mon, 01 Jan 0001 00:00:00 +0000

Six numbers. One graph. Five reports. Click any row to filter the page.

The six numbers

Unique visitors, total visits, total pageviews, views per visit, bounce rate, visit duration. In that order, always.

Click a tile to graph it. Press I to switch between hourly, daily, weekly, and monthly views. A period with no data shows a gap, not a false zero.

Compare against the previous period or last year. Change markers appear on all six tiles.

No cookies, and no banner for most sites

Mon, 01 Jan 0001 00:00:00 +0000

Feasible sets no cookies on the sites it measures.

The tracker writes two things into localStorage, and neither identifies anybody: feasible_ignore, which you set yourself to stop counting your own visits, and feasible_outbox, the retry queue that holds an event for a moment if the network drops it.

What never touches disk

The visitor’s IP address.

That isn’t a policy, it’s the shape of the pipeline. Geolocation and the visitor hash both happen in the ingest tier - the only place the address exists - and the address is discarded before anything is written. The durable event type has nowhere to put one.

Goals, funnels and conversions

Mon, 01 Jan 0001 00:00:00 +0000

A goal is your definition of a win - a page reached, an event fired, a page scrolled far enough. A funnel is two to eight of those goals in order, with the drop-off between each step.

Both live in the Behavior card at the bottom of the dashboard, and both are configured in site settings or through the API.

Three kinds of goal

Page goals match a path, with wildcards - /thanks, or /docs/*. Up to 2,000 characters of pattern, which is more than anyone has ever needed.

Custom events and properties

Mon, 01 Jan 0001 00:00:00 +0000

A custom event is one line of JavaScript:

feasible('Signup', { props: { plan: 'scale', trial: 'yes' } })

The properties ride along with the event, and every report in Feasible can then be broken down by them. What your application knows - plan, experiment variant, article category, whether the visitor was logged in - becomes a dimension next to browser and country.

Pageviews take properties too. Same idea, p instead of props.

Revenue tracking

Mon, 01 Jan 0001 00:00:00 +0000

Any goal can carry money. Mark it as a revenue goal, pick a currency, and send the amount with the event:

feasible('Purchase', { revenue: { amount: 49.99, currency: 'USD' } })

You get total revenue, average revenue, and revenue per visitor. Break each one down by any dimension.

First-touch attribution

This is the decision that makes the report worth reading.

Picture the sequence. Somebody clicks a link on Hacker News, reads for a while, comes back the next morning, clicks buy, and gets redirected to a hosted checkout page. Payment succeeds, they land back on your thank-you page, and the referrer on that final pageview is your payment provider.

Teams, roles and shared dashboards

Mon, 01 Jan 0001 00:00:00 +0000

Every plan has unlimited teammates. Five roles cover the team. Two more cover guests on one site.

Who can do what

Owner - everything, including deleting the team and transferring ownership. There’s one, always.

Admin - sites, settings, members and billing. Everything except ending the team or handing it on.

Editor - sites and site settings, no access to members or money.

Billing - the dashboard and the billing screens. For the person who pays the invoice and shouldn’t be able to delete a site.

Reports and alerts

Mon, 01 Jan 0001 00:00:00 +0000

Schedule a report. Get an alert when traffic changes. Send either one by email, Slack, or both.

Reports on your schedule

A weekly report goes out at the site’s own local Monday 00:00. A monthly one at local midnight on the 1st.

That sounds like a detail and it’s the hard part. The job runs every hour and asks each site whether its own clock has just crossed the boundary, which is the only approach that works when one install serves sites in forty time zones - including the offsets that aren’t whole hours, and including the Sunday in March when one of those hours doesn’t exist.

The data health panel

Mon, 01 Jan 0001 00:00:00 +0000

The health panel shows accepted, dropped, bot, and shortened events from the last 24 hours. Every drop has a reason, such as bot, datacenter_ip, unknown_site, hostname_not_allowed, shield_page, or invalid_payload. There’s no miscellaneous bucket.

The last request we received

The panel prints the most recent request in full: the client IP we resolved, which header we believed it from, whether trusted proxies are configured, the hostname, the path, the tracker version, the user agent, and what we did with it.

Filtering, shields and path cleaning

Mon, 01 Jan 0001 00:00:00 +0000

Three different tools that get confused with each other. Filters change the report you’re looking at. Shields stop traffic being counted at all. Path cleaning changes how URLs are grouped before either of those runs.

Filters

Six operators, and that’s the complete list: is, is not, contains, does not contain, matches regex, does not match regex.

Repeated filters AND together. Several values inside one filter OR. You get up to 32 filters at once, 1,000 values in any one of them, and five dimensions in a single query - limits nobody has ever hit by clicking, and which exist so the API can’t be handed something that takes a minute to answer.

Import your history, export everything

Mon, 01 Jan 0001 00:00:00 +0000

Import loose CSVs or a ZIP. Export one ZIP with ten reports and every raw event.

Bringing history in

Export your GA4 reports as CSV and upload the folder - pages, sources, locations, devices and browsers land on the same charts as your live traffic, dated where they belong. It’s the same path any other tool’s export takes.

Import takes a ZIP or loose files. Recognized dimension columns cover page, hostname, page title, event name, entry and exit page, referrer, source, channel, the three UTM fields, country, region, city, device, screen size, browser and version, operating system and version, and language. Metric columns are visitors, visits, pageviews, events, bounces, exits, visit duration and time on page. If your previous tool exports daily roll-up CSVs with columns named like that, they come straight in.

The API, webhooks and MCP

Mon, 01 Jan 0001 00:00:00 +0000

Create one key for the stats API, site management, and webhooks. Add scopes when you want less access. Keys belong to a team and stop working when their creator leaves.

What you can do with it

Read stats. POST /api/v2/query is the modern endpoint and it’s the same query engine the dashboard runs on - fifteen metrics, every dimension, the same six filter operators, up to 32 filters and five dimensions at a time. There are also v1 endpoints for aggregate, timeseries, breakdown and realtime visitors if you want a URL you can curl without a JSON body.

Why it's fast and small

Mon, 01 Jan 0001 00:00:00 +0000

3,377 bytes gzipped. That’s the whole tracking script - 7,099 bytes before compression, one request, cached for an hour.

Measured against Google’s gtag.js in September 2026, that’s about 43 times smaller. The difference is a few hundred milliseconds on a slow phone. We’ll take it.

A hard size limit

There’s a byte budget of 3,584 for the base script. Go over it and the build exits 1 - enforced twice, once in the bundler and again in a Go test that reads the shipped file.