Know which Google searchReddit threadChatGPT answernewsletterx.com postsource paid you, and how we know.

Referrers become sources you recognise: t.co is x.com, chatgpt.com is ChatGPT, and news.ycombinator.com is Hacker News. Each source belongs to a channel, and people sent by AI assistants get a channel of their own. Every breakdown row shows its visitors and the revenue behind them.

AI crawlers are counted apart, never as visitors; most never run a page's script, so a few lines on your server report them. Other bots and automated browsers aren't counted at all. And Direct, the visits with no referrer, is read for what it most likely was: typed or bookmarked, a shared link, a return from checkout, or an app's own browser. The hero's dashboard shows it.

Peak hours lays the period out by weekday and hour, for visits, sales or revenue, and names the busiest three hours only when the data supports it.

One row per visitor: a friendly name and an avatar drawn from their id, where they are and on what, where their latest visit came from and the pages it took, engaged time per visit, and what they've paid. Open a row and every visit is there; the journey page has each page, goal and payment.

Sort by Likely to buy and the score says what drives it: coming back, and time on pricing and checkout. In cookieless mode a row is a visit, and the page says so.

Connect Stripe, Paddle, Lemon Squeezy or Polar with a key you create, limited to what Tally reads and permission to add its own webhook. It brings in the last 365 days first, then each payment as it happens.

Stripe goes deepest: Checkout, Payment Links, subscriptions and invoices, and the snippet tags Payment Links by itself. The other three link a sale by the visitor's id passed into their checkout, or by the buyer's email.

New sales, renewals, upgrades, refunds and chargebacks are told apart. Amounts leave out tax and are converted to US dollars at the European Central Bank's rate for that day. A payment is counted once, however the provider reports it.

Coverage is the share of your revenue that's linked to a visitor. Under it is the fix-it list: each reason a payment couldn't be linked, how many payments and how much money it involves, and the code that fixes it.

A payment nothing links stays unlinked and counts against coverage. If evidence arrives later, such as the buyer's visit to your success page, it's linked then.

Path to purchase is the journey behind the period's new sales: how long buyers take from their first visit, how many visits and pages it takes, and how many buy on the first visit. The flow chart runs from the source that first brought each buyer in to the one credited with the sale.

Lifetime value counts each customer under the source of their first visit, with everything they've paid since: renewals, upgrades and refunds included.

The clock runs while the tab is in view and the visitor has scrolled, typed, tapped or moved the pointer in the last 30 seconds. A tab left open behind others for 8 hours adds nothing. The dashboard shows the median per visit, so one forgotten tab can't pull it up.

Pages adds how far down people scroll: each page's reading time and depth, how many reach the end, and how buyers read a page before paying against everyone who didn't.

Turn replay on for a site and its recorded visits play back with every page, goal, rage click and payment on the timeline. A buyer's journey and their receipt link straight to the recording. Every input is masked and passwords are never recorded.

Heatmaps draw clicks, rage clicks, dead clicks and scroll depth over the page itself, for desktop, tablet and phone.

The last five minutes on a globe or a flat map, with a list beside it: each visitor's flag, page, city and source. A new arrival rings in; a sale bursts in coral at the buyer's city. Click a visitor to find them, and replay the last eight hours in 15-minute steps.

Two experimental views, drawn only from signals Tally has. The ecosystem places every visitor of the last 90 days by what they did, gathers them into clusters named by what stands out, and colours them by health; play it and it moves week by week.

The river follows visitors from the source that brought them, through the pages they reached, to whether they bought, week by week, coloured by source, by experience or by outcome. Health is our reading of renewals, refunds, chargebacks and recency, never a prediction.

A message for every sale, with where the buyer came from and how the sale was linked; a spike or a drop against the usual for that hour; trouble with a payment provider; and a daily or weekly digest. Quiet hours hold them until morning.

Make a read-only key on a site's Setup page; only a hash of it is kept. The REST API answers with the dashboard's own numbers, the command line prints them as tables, and the MCP server lets an AI assistant ask its own questions.

Choose a mode for each site. Whichever you choose:

• The visitor's IP address is used to find their city, then dropped. It's never stored, logged or written to disk.
• Cookieless mode needs no cookie. It counts a visitor by a hash of the IP address and the browser with a salt that changes every day and is then deleted, so nobody can be followed from one day to the next. The hash is kept; the address never is.
• Visitor ids never appear in logs.

The privacy notice says what is kept, for how long and who processes it, and the data processing agreement covers your visitors' data.

Tally reads what DataFast's tag reads, so the code on your pages keeps working: window.datafast() goals, data-fast-goal attributes with their data-fast-goal-* properties, and data-website-id in place of data-site.

Returning visitors keep their datafast_visitor_id, so checkout code that already reads it still links sales, and so does the datafast_visitor_id in your Stripe metadata.

The tag is about 3 KB, loads with defer and has no dependencies, so it never holds up your page. Engaged time, goals and clicks on links to other sites come with it. A goal is tally('signup') or a data-tally-goal attribute on a button.