Client loginStart the Lab
/ Lab

Ask your own datawho does not fit

For a game, a shop or an app that keeps a record of what people do. You build a rule step by step, and it names the ones past the line.

From $49 a month20 rules, 5 of them on a schedule. Stop any month.

Start the Lab

The rule

Flag every buyer whose number of refund.issued over the last week is above the 99th percentile of others who share spend_band.

1 of 55 measured is past the line

Which one
buyer_0117
Measured
21
Had to beat
14.9

In plain words54 buyers filed one or two refunds that week. One filed 21.

Illustration: a rule from the Lab's own examples, measured on its example shop data. buyer_0117 is not a real customer.

How a rule is made, in order

  1. Point it at your records

    Parquet or JSON files in your own storage bucket, read where they are with read-only access. Or begin on the example data and connect nothing.

  2. Start from a question

    Six of them, in plain words: who does far more of something than people like them, who suddenly changed, who stopped, what happens too evenly for a person to be doing it. Or take a worked example and change it.

  3. Say what the line is compared with

    players flagged against one flat number
    28
    flagged against players who have played as long
    3

    Measured on the Lab's example game week, with one question asked two ways. A flat number mostly finds whoever plays the most. You pick from four comparisons: a number you set, everyone else, people like them, or their own past.

  4. Read who it found, then leave it running

    A named list: which one, what was measured, and the line it had to beat. You mark each case confirmed or dismissed, with a note. Then it runs every hour or once a day, and a case that is already open is updated, not filed twice.

What it will not do

  • It does not take your files

    They are read where they are and not copied. Kept here: the rule, the counts of each run, the cases it found, and a profile of your columns with a few example values.

  • No model decides who is on the list

    A rule is arithmetic. It reads back as one sentence, and the same data gives the same list every time.

  • It never acts on anybody

    No rule bans, blocks or restricts anyone. It files a case, and a person decides.

  • It does not report from too little

    A peer group of fewer than five is held back, and comparing someone with their own past needs five earlier periods first.

Good to know

Records of things that happened
One row per event, with a time, a name for who or what it is about, and a column called event that says what kind of row it is.
Parquet or JSON lines
With flat columns. Nested fields have to be flattened before they reach the bucket.
Taking the list with you
Each run exports as a CSV file. A letter on Monday reaches you only when a finding has changed.
Read-only access
For Amazon S3 you create a role that can list and read, and nothing else. Other storage is connected together with us.
What a token is
Measuring your own data is metered: one token is 1,000 measured records. The example data is never metered.
When the tokens are spent
Rules on your own data wait until the month resets or you buy a pack. Nothing more is charged unless you turned automatic refills on.

What it costs

  • Start here

    Lab - Bench

    $49 a month

    • $102.90 for 3 months up front, save $44.10 (30%)
    • $176.40 for 6 months up front, save $117.60 (40%)

    20 rules, 5 of them on a schedule, 3 sources of your own data and 1,000 tokens a month.

    The example data and the worked examples are open on every plan.

    Start the Lab

  • Lab - Studio

    $149 a month

    • $312.90 for 3 months up front, save $134.10 (30%)
    • $536.40 for 6 months up front, save $357.60 (40%)
    rules
    100
    scheduled
    40
    sources
    10
    tokens a month
    5,000
    Start Lab - Studio

  • Lab - Works

    $499 a month

    • $1,047.90 for 3 months up front, save $449.10 (30%)
    • $1,796.40 for 6 months up front, save $1,197.60 (40%)
    rules
    500
    scheduled
    200
    sources
    50
    tokens a month
    25,000
    Start Lab - Works

Stop at any time. The plan runs to the end of the month you paid for.

You pay on the next step, with no account to make first. The account is made for you from the receipt, and the Lab opens on the example data.

Looking first costs nothing: an account without a plan keeps three rules, one of them on a schedule, and one source of your own.

Open a free account

Not today?

We email you this page and what it costs. No account, no list.

One mail with the page. Tick the box if you also want a follow-up. How we handle the data

Before you point it at your data

You would be letting us read your records. You should know who we are and what stays here.

The companyOakmont Technologies LLC

Oakmont Technologies LLC5511 Parkcrest Dr. Suite 103, Austin, TX 78731, United StatesFull company details

What is keptThe rule, the counts and the cases. Not your files.

Each rule, the counts of each run, and the cases a run found: the name it flagged, what was measured and the line. A profile of your columns is kept too, with a few example values, so the steps can offer them. Your files stay in your storage. How we handle the data

Stopping itAny month, from Billing in your account.

Cancel from Billing in your account. The plan runs to the end of the month you paid for, and nothing is charged after it.

A personA person reads every message and replies.

admin@oakmonttechnologies.com, or the contact page. A person reads every message and replies, usually within one business day.