Data integrity — can I trust this number?

Data integrity answers one question: Can I trust this number? Every day it checks the records Beacon reads from your tools, lists every problem it finds with a link to the record itself, and says plainly which checks it could not run. It belongs to the Shared foundation, the family every other module reads — see Modules.

​

What it does

Once a day, after your tools have synced, Data integrity runs a set of checks over the records Beacon holds: customers, subscriptions, invoices and deals. A record fails if any one check finds a problem with it. The module reports two plain figures, each with the counts behind it:

• Data integrity — the records that pass every check, divided by the records checked.
• Data completeness — the records that arrived, divided by the records Beacon expected from each tool.

Nothing is weighted, and no customer is given a score. Beacon never fixes a record in your tools. It lists the problem, you fix it where the record lives, and the problem closes on the next daily check.

​

The checks

A check that is not run is never counted as passed. It shows Not measured, with the reason.

​

What you see on its page

In Beacon, open The Brain and choose Data integrity. The page has six tabs:

• Overview — the two figures with their counts, data integrity check by check since the first run, open issues by area, the tools it reads, and when it last ran.
• Dashboard — the module’s main charts. Most read Not measured or Not yet calculated today: integrity by tool and by module, CRM against billing revenue, and customers grouped by how far their data can be trusted.
• Signals & alerts — what the module watches, its alert rules, and whether each one is sending.
• Scorecard — every check with the number of records it found, or Not measured and why. Below it, the key fields and the line each must reach, and the rules.
• Issues — every open problem, the ones that need you first. Each record links to itself in HubSpot or Stripe, where you are already signed in, and the whole list can be downloaded.
• Setup — the questions the module uses, and the answers in use.

The module’s state reads Live only when every check and completeness are measured. Because completeness is not measured yet, every company sees Running with gaps today.

​

What it needs

See Connect your tools for how each connection is made.

​

Reads from and feeds

Reads from: Data model, for every record it checks, and Customer matching, whose possible matches appear here as issues so that one list holds every problem. Feeds: later, Forecast confidence, Customer fit, Signals, Churn risk and the revenue forecast will use it to discount a figure built on weak data. That needs the score for each record, which is not calculated yet — so today no other module adjusts its figures by Data integrity.

​

Alerts

Four alert rules are built: A fifth, for a customer whose data is too thin to score, waits for the score for each record. No alert is sent yet. The setting that switches alerts on for your company is still being built. Until then, the page shows each rule as off and nothing goes out. When alerts start, email goes to the person named as your data owner.

​

Setup questions

• Which tool is right about recurring revenue? Recommended: billing.
• Who looks after the data in each area? Each area’s owner gets the issues to fix in that area.
• Where do your contracts live? Recommended: each Stripe subscription is the contract.
• Should small customers have a lower bar? Recommended: no — one bar for every customer.

Today the Setup tab shows each question with the answer in use, and the module runs on those answers. Saving your own answers is not available yet.

​

What is not calculated yet

• Data completeness. Beacon does not yet count the records it expected from each delivery, so this figure reads Not measured.
• The four checks marked Not run yet in the table above.
• Any figure per customer — the score for each record described in How a record’s data is scored, and the Dashboard’s customer bands.
• One overall score from 0 to 100. Its five weighted parts are shown on the Scorecard as not yet calculated.
• Integrity by tool and by module on the Dashboard.
• The three automatic stops — every revenue forecast pausing when CRM and billing are over 15% apart for the company, a customer with very weak data treated as having none, and every alert stopping when the data overall is too weak. They depend on the figures above and are not in force.

​

Related