Loading…
Latest snapshot per provider account. Prepaid balances are what the ledger is reconciled against.
Last run per job. Stale = later than twice its cadence.
Who costs the most to run today — the CEO call on who to move to self-hosted starts here, sorted by projected cost/mo, highest first. voice path and channel-pool peak are not shown per user yet: there is no per-user self-hosted switch until QUIET-129 ships, and no concurrency tracking until QUIET-130 ships. Margin vs subscription price is left out until pricing lands — cost only, for now.
One row per offer (QUIET-348). granted is trials actually given — counted when Stripe first reports a subscription trialing, so an abandoned checkout is not counted. conv is settled trials only: a trial still running has decided nothing, so it is left out rather than counted as a loss. ended is any grant whose subscription is now cancelled, including one that converted, paid for months and then left — telling those apart needs invoice history we do not store. no fwd is trials that never proved forwarding, so they received no calls at all and cost only their number. cost is this month's spend by that offer's cohort, from the same per-user figures as the rest of the page. Visitors per offer are in the Signup funnel panel below, filtered by offer.
Web signup, from the first setup screen on (QUIET-356). Dates are your local days. A visitor is one browser (a first-party cookie); a step counts each visitor once however often they viewed it. of prev is visitors who reached the step over those who reached the one before — for the assistant screen, over those the number question did not send to the waitlist, because the waitlist is a branch rather than a drop. of start is over everyone who answered the number question. Neither is clamped: a visitor who resumes mid-funnel counts only on the steps they viewed in the range, so a step can exceed the one before it. to next is the median time from first view of the step to first view of the next. The highlighted row is the step the largest share of visitors did not reach; they stopped on the step above it. An offer is the one a visitor was last shown; no offer link is the baseline a campaign is compared with. Outcomes are what those visitors' accounts are now. Before launch every visitor is a tester.
attributed: the provider tied the charge to this user's call (Telnyx per-leg cost). allocated: this user's share of account-level spend, split by recorded units (seconds, characters, tokens, calls). Sum of all rows incl. “no user” = account total for the window.
est = part of the total priced from the provider's published rate because no invoice covers that period yet; entering the invoice replaces it. effective rate = actual spend ÷ our recorded units. units gap = provider-billed units vs ours (Telnyx bills 60 s minimums; a gap here is a counting difference, not a rate difference). Flagged beyond ±15 % or when no expected rate exists.
What the prepaid balance actually fell by, minus what the ledger books. A positive residual is money the vendor took that we do not record — the gap this dashboard exists to expose. Only days whose balance snapshots cover the full 24 hours are counted; days covered says how many of the window's days those are. Day-level figures oscillate (rental is debited in a lump but booked evenly), so read the window number.
Telnyx's own daily total per product vs the sum of the per-record ledger for that day. Residual ≠ 0 = records the API did not return, a classification gap, or a rate difference. partial day = the totals were last synced before that day closed, so its residual is an artefact of sync timing, not a finding.
For providers with no spend API (Fly, Vercel, Supabase, Cartesia plan). Pro-rated by day into every window and allocated by the line's basis.