Skip to content

data-intelligence · agent

Driftnote

When a KPI breaks, the note says which cut moved — and whether it is a tracking bug

popular

11 h/mo

Saved

50 min

Setup

8

Systems touched

55

Credits per run

$3,900

Perpetual single-workspace license · Band 3 · Operator

Or included in Growth at $990/mo, along with everything else that tier carries.

Book a demo
Deployment
connected
Setup
50 min
Hours saved / month
11
Included from
Growth
Credits per run
55

03 — Proof

What is measured, and what is not.

The hours and the setup time are catalog math at a stated volume. Production telemetry is not published yet, and nothing here pretends otherwise.

Catalog math

12 material KPI breaks/month × 40 min to notice, slice, and write up (8h), plus ~4h of Monday dashboard hunting the alerts replace, minus ~1h remaining true-incident work

Run telemetry

04 — What it does / 5 outputs

Watches named KPIs, flags breaks vs the baseline you set, and writes the first-cut explanation.

  • Break note with cut
  • tracking-suspect flag
  • owner ping

05 — One run

What one run takes in, and what it puts back.

No sandbox on this page yet. This is the real input and output contract from the record, not a demo recording.

In

  • KPI definitions
  • grain
  • thresholds
  • dimension whitelist
  • owner list
  • volume/null floors
  • Slack channel

Out

  • Break note with cut
  • tracking-suspect flag
  • owner ping
  • ack log
  • reverse notice

06 — How it works / 50 min setup

Three steps, then it is furniture.

  1. 01

    Connect Snowflake and BigQuery

    Scoped authorisation, revocable from your IdP or the dashboard.

  2. 02

    Point it at the work

    KPI definitions, grain.

  3. 03

    It runs

    Break note with cut — on a trigger or a schedule, 50 minutes after you start.

07 — Features / 7 listed

What is actually in the box.

  • Baseline you choose: trailing 28, same weekday, or a fixed band — not a black-box “anomaly score”
  • First-cut slice on dimensions you whitelist; it will not invent a dimension that is not in the table
  • Tracking-suspect flag when event volume or null-rate moves with the KPI
  • Named owner per metric; Slack once per break, not a stream of sibling cuts
  • Quiet unless the break persists or reverses; no daily “still down 0.3%” drip
  • Links the query or Look behind the figure
  • Human ack required before it marks the break closed — it does not close itself because the tile went green overnight

08 — Data

Does it fit your stack and your data?

Inputs this product consumes and outputs it produces
InputOutput
KPI definitionsBreak note with cut
graintracking-suspect flag
thresholdsowner ping
dimension whitelistack log
owner listreverse notice
volume/null floors
Slack channel

09 — Integrations / 8 systems

It touches these, and nothing else.

10 — The money

Your hours, your hourly cost.

Two numbers: the hours this product removes at the catalog's stated volume, and the cost of an hour, which only you know.

11

You type this one. We do not know what an hour costs you.

Recovered per month

$550

11 h × $50 an hour, before the credits a run spends.

Where the hours come from: 12 material KPI breaks/month × 40 min to notice, slice, and write up (8h), plus ~4h of Monday dashboard hunting the alerts replace, minus ~1h remaining true-incident work.

11 — Pricing

Buy it outright, or get it with a plan.

One-time

$3,900

  • Perpetual single-workspace license to this product’s logic, UI and connectors. Runs still spend credits.
  • Returns $1,560 as non-expiring credits at your tier’s rate.
  • Twelve months of updates and connector maintenance included.

Per run

55 credits

$1.10 a run at the Starter rate of $0.0200 per credit, less on every tier above it.

Credit class: standard

Every run carries a published hard ceiling. A run that hits it stops and re-quotes instead of billing on.

See how plans and credits work

12 — Packs / 3 packs

What else is in the pack.

  • Part of the Closeout pack

    The month-end binder, assembled, before you start actually closing

    • Tiedown

      No close checklist item is real until the bank and card lines are matched to the GL, and Tiedown produces exactly the signed-off rec packet plus the named break list that Closeout's checklist wants attached.

    • Billtray

      The unbilled vendor is the classic close accrual, and Billtray is the product that gets vendor bills out of the AP mailbox and coded into the GL before cutoff instead of after it.

    • Outlay

      Expense cutoff fails on the receipts still in someone's phone; Outlay codes them into a report and posts the reimbursable spend, so the accrual stub has something to reconcile against.

    • Sheetfold

      Workpapers arrive as a dozen scattered workbooks, and Sheetfold consolidates them into one mapped table with a control total and a per-file error list — the footing check a binder needs.

    Included in Enterprise · from $7,500/mo

    See the pack

  • Part of the Lanehold pack

    Milestone exceptions and missing PODs, named, before the customer asks where the freight is

    • Shipask

      "Where is my freight" is a where-is-my-order ticket answered from the fulfilment record, and Shipask is the product that answers it with tracking and ETA and parks the ones it cannot match.

    • Heatlist

      The exception list is a queue a coordinator works down before the customer calls, and Heatlist produces exactly that daily list with the misses that escalated off it logged.

    • Fieldpull

      A POD is a document whose absence blocks the invoice, and Fieldpull is what reads the delivery paperwork into the shipment record and queues the ones with nothing attached.

    • Ledgerline

      The invoice-hold flag only means something if the invoice actually goes the moment the POD lands, and Ledgerline is the finance-shelf product that drafts, sends and posts it.

    $14,900

    One-time

    Or included in Growth at $990/mo.

    See the pack

  • Part of the Topsheet pack

    Monday’s one-pager for people who will not open Looker

    • Pipeshade

      Topsheet's own key features name the pre-board appendix as "last month vs Pipeshade/Stripe actuals when connected", so the forecast with its visible haircut is a stated part of the packet.

    • Tableside

      The pack blocks the send when a figure has no citation, and Tableside is what supplies a checkable figure with the SQL and the assumed grain attached to it.

    • Cohortcut

      "What moved" is only meaningful in the cut the company runs on, and Cohortcut rebuilds retention, LTV and segment cuts from Stripe with a definition checksum and a diff against last run.

    • Actionline

      The decision list rolls forward week to week and each entry needs an owner, which is Actionline's rule — no task written without a named owner, everything else to the exception list.

    Included in Enterprise · from $7,500/mo

    See the pack

13 — Security

Runs on nox.markets, acts in your tools.

A scoped OAuth token pulls only the objects the inputs above require, writes the outputs back to your system of record, and keeps the run payload for your workspace retention window. Revoke the token in your IdP or in the dashboard; connector secrets are deleted within 24 hours of a disconnect.

Run payloads are kept for your workspace retention window — 30 days by default on Starter and Growth, 90 on Scale and Enterprise, with zero retention available on Enterprise.

Read the security controls

14 — Customers

Someone like you, running this.

15 — Detail

The long version, including what it will not do.

KPI “monitoring” in this size band is a person opening Looker on Monday and hoping they remember what last Tuesday looked like. The miss is rarely the 40% crash — that one shows up in Slack from sales. It is the refund rate that ticked from 2.1% to 3.4% over nine days, or checkout conversion that died in one browser, while the top-line tile still looks fine. By the time someone slices it, a week of paid traffic is gone.

Driftnote watches the KPIs you name in the warehouse, Looker, Metabase, GA, or Stripe. A break is a threshold you set (absolute, percent vs trailing 28, or vs the same weekday). The note is not a sparkline. It is: which cut moved (plan, region, channel, SKU — if that dimension exists), whether volume collapsed (suspect tracking) or mix shifted, and who owns the metric. It posts to Slack once, then stays quiet unless the break persists or reverses.

It will not page at 2am and it will not open a PagerDuty incident. It does not know your tracking pixel is broken unless row volume or a null-rate bound is in the definition. Amplitude, Anodot, and a real observability stack exist; do not stack this on top of them. Buy it when detection is still a human with a Monday habit.

16 — Questions / 5 answered

Before you buy.

What does it need from me before it can run?

KPI definitions; grain; thresholds; dimension whitelist; owner list; volume/null floors; Slack channel. Nothing else is collected for a run.

How long does setup take?

50 minutes. Runs on nox.markets, acts in your tools. You authorise Snowflake, BigQuery, Postgres, Looker, Metabase, Google Analytics, Stripe, Slack once.

What does it cost to own?

$3,900 once — band 3, Operator. Runs still spend credits. It is also included in the Growth plan at $990/mo — buying it outright is the way to keep it after a plan ends.

What does it hand back?

Break note with cut; tracking-suspect flag; owner ping; ack log; reverse notice.

Where does the hours-saved number come from?

12 material KPI breaks/month × 40 min to notice, slice, and write up (8h), plus ~4h of Monday dashboard hunting the alerts replace, minus ~1h remaining true-incident work

17 — Related / 4 products

More on the Data & Intelligence shelf.

18 — Collections / 3 shelves

Where it sits on a shelf.

Driftnote runs 11 hours of this a month.

Book a demo

Or keep looking: Data & Intelligence · Products for this team · The whole catalog

$3,900Included in Growth · $990/mo