Data quality · Data contracts · Data observability

Your data quality rules,

unified in one place.

Data quality tooling for enterprise platform teams. Every rule in one contract the whole team can point to.

Business and Engineers can find the rule, and who owns it.

When a check fails you get the rows, not an alert.

Your data stays in your network.

dataset: test_source/unity_catalog/arthur/retail_orders
verifications:
  - schéma: {}
  - fraîcheur:
      colonne: order_datetime
      seuil:
        unité: heure
        doit_être_inférieur_ou_égal_à: 24
colonnes:
  - nom: order_id
    type_de_donnée: string
  - nom: last_name
    type_de_donnée: string
  - nom: email
    type_de_donnée: string
  - nom: payment_method
    type_de_donnée: string
  - nom: order_value
    type_de_donnée: double
  - nom: order_quantity
    type_de_donnée: string
  - nom: order_datetime
    type_de_donnée: timestamp
  - nom: country_code
    type_de_donnée: string
  - nom: dim_idx
    type_de_donnée: int

“Most data quality tooling will tell you something broke after it broke.”

Sid Srivastava, 2K Games

Where do your data quality rules live today?

“There is often a loss in translation when business needs are converted into data quality checks.”

“Out of 1000s of tables, a critical one failed and no one looks into it.”

“We have around 40 teams operating in silos, bespoke builds everywhere, no consistency.”

“We track quality requirements in Excel. 800 critical data elements.”

1
2
3
4
5
6
7
8
9
10
11
12
13
dataset: .../retail_orders
checks:
  - schema: {}
  - freshness:
      column: created_at
      threshold:
        unit: hour
        must_be_less_than_or_equal: 24
columns:
  - name: discount_currency
    checks:
      - invalid:
          name: Allowed currency is USD

The rules exist. Nobody can point to them.

When every team can point to the rule.

Three jobs that stop being tickets.

Every dataset covered. Starting from zero.

✓

AI drafts the first version from your data. You approve every change.

✓

Declare the columns. A renamed field or a changed type fails the check.

✓

Reconcile source against target and prove nothing was lost.

The owner writes it. The engineer signs it off.

✓

The owner writes the rule in plain language. Nobody translates it.

✓

Every change lands as a diff in code. You can see who changed what.

dataset: test_source/unity_catalog/arthur/retail_orders
verifications:
  - schéma: {}
  - fraîcheur:
      colonne: order_datetime
      seuil:
        unité: heure
        doit_être_inférieur_ou_égal_à: 24
colonnes:
  - nom: order_id
    type_de_donnée: string
  - nom: last_name
    type_de_donnée: string
  - nom: email
    type_de_donnée: string
  - nom: payment_method
    type_de_donnée: string
  - nom: order_value
    type_de_donnée: double
  - nom: order_quantity
    type_de_donnée: string
  - nom: order_datetime
    type_de_donnée: timestamp
  - nom: country_code
    type_de_donnée: string
  - nom: dim_idx
    type_de_donnée: int

When a check fails, the rows come with it.

✓

Every failed record lands in your own warehouse. Not a sample.

✓

Anyone with warehouse access can query them. No row reaches Soda Cloud.

Missing

Duplicated

Invalid

Customer records

Sales records

Transactional records

Missing

Duplicated

Invalid

Customer records

Sales records

Transactional records

“We just want to go and observe what’s happening.”

Sid Srivastava, 2K Games. 1,945 checks across 984 datasets in under a year, against a 95% quality target they are hitting.

At the end of the day, we don’t want to be in there managing the checks, updating the checks, adding the checks. We just want to go and observe what’s happening, and that’s what Soda is enabling right now.

Sid SrivastavaDirector of Data Governance, Quality and MLOps at 2K Games

Before we used to wonder: who should we even contact, and who owns this? Data contracts brought transparency and also bigger cooperation between different teams.

Renata HlavováData Engineer at Make

Soda has integrated seamlessly into our technology stack and given us the confidence to find, analyze, implement, and resolve data issues through a simple self-serve capability.