All videos
Antvia Data · Episode 1 of 8

Data in one tour

A tour of Antvia Data for a data or technology leader. It starts on the home page, where every problem on the platform is ranked, then walks the product in the order data moves: sources and files coming in, workflows and runs keeping it moving, the catalog and lineage describing what exists, and access deciding who can read it. It ends by fixing a broken connector live and watching the repair reach a dashboard in Antvia Intelligence.

9:438 chapters1,173 wordsSilent, with captions

Also on YouTube, with the rest of the Antvia Data series. The whole episode is written out below, with the timing, if you would rather read it.

Chapters

What happens, in order.

0:00
Introduction
Title card.
0:07
What needs somebody
The home page: the business figures and whether they are current, the machinery behind them, and the ranked list of what needs attention.
1:54
Bring data in
Sources, with the broken channel stock feed opened; the upload nobody maintains; and the column register with its unprotected personal data.
4:08
Keep it running
Workflows that succeed and are still wrong, the channel stock refresh that carries on after failing, and the run log in Flow.
5:21
Know what you have
The catalog, with the gold-certified table that is nine days old; and lineage, tracing each connector to the dashboards and people it reaches.
6:32
Who can see what
Access: the account with no second factor and the dormant account; the redaction policy that is switched off; and the audit trail filtered to consequences.
7:59
Fix it, live
Re-authenticating the channel stock feed, the six-stage run, the source and lineage turning healthy, and the Business pulse dashboard in Antvia Intelligence becoming current.
9:20
Close
Closing card.
Transcript

The whole episode, in words.

The videos are silent, with the captions burned in. This is the same text, with the timing, so you can find the part you need without scrubbing for it.

Introduction

0:00
Antvia Data. Data in one tour. Where a company's data comes in, how it is kept honest, and what happens when something breaks.

What needs somebody

0:07
What needs somebody. The home page ranks every problem on the platform, worst first.
0:12
Antvia DataAntvia Data is where a company's data comes in, gets governed and is kept honest. This is the home page for the data team at Acme Home Appliances, a demo company.
0:25
The numbers firstThe top row is the figures the business runs on, read from the dashboard its leadership opens. Two of the four are not current, and the tiles say so.
0:38
Nine days oldChannel stock is nine days old. Nothing on the tile is failing. The feed behind it stopped arriving, and the number stayed on screen.
0:48
UnverifiedCustomer satisfaction is unverified: three quality checks on the feedback table are failing.
0:55
ArrivingArriving is the machinery. Twelve connectors with two broken, two failed runs this cycle, two workflows racing an input, and one spreadsheet nobody maintains.
1:05
What it becameWhat it became is the result: 22 governed tables, seven failing checks, and five columns of personal or sensitive data with nothing protecting them.
1:16
Needs somebodyNeeds somebody ranks every problem on the platform, worst first. Each row names the page where it gets fixed.
1:26
At the topAt the top: the channel stock feed has not landed for nine days, and a dashboard is still reading it.
1:35
SecondSecond: five columns of personal data are open to anyone who can query them. We will see which ones.
1:46
Recently changedOn the right, the latest governance changes and refusals from the audit trail, newest first.

Bring data in

1:54
Bring data in. Connectors, files people upload by hand, and the columns they carry.
2:01
SourcesSources lists every system connected: the ERP, the service CRM, distributor management, the plants and the marketplaces. Two are not landing, and the longest outage is nine days.
2:13
Broken firstBroken sources sort to the top. The channel stock feed is a critical source, and it has failed every morning since 30 July.
2:24
In plain wordsEach source says what broke in plain words. The distributor stock portal's API key expired on 30 July, and nobody was told.
2:34
Where it landsIt names the table it lands in, and offers the fix. We will come back and use it at the end.
2:44
A file dropThe survey agency's daily file has not matched for four days, because the agency renamed its files.
2:58
Files people uploadNot everything arrives through a connector. Import covers the spreadsheets people upload by hand: targets, mappings, price lists.
3:06
Nobody maintains itThe regional targets file was uploaded by an analyst who left in June. Three dashboards still read it, and no alert will ever fire, because an upload has no schedule to miss.
3:19
Nobody reads itAt the other end, a file collected for one analysis in June that nothing reads. That is how a workspace fills up.
3:35
Every columnDesign a source lists every governed column, what it holds, and whether a policy protects it. Five personal or sensitive columns have no protection.
3:45
Open to anyoneDistributor owners' email and mobile numbers are classified as personal, and anyone with query access can read them.
3:55
DriftAnd a column that changed under us: warranty claim amounts now arrive with a rupee sign, so they land as text and every sum on them returns zero.

Keep it running

4:08
Keep it running. Scheduled workflows, and the log of every run.
4:16
Green, and wrongWorkflows are the scheduled jobs. Two ran green this morning and are still wrong, because a step reads a table without waiting for the step that makes it.
4:30
Carry onThe channel stock refresh is set to carry on when its first step fails. So everything below it has published stale stock for nine mornings, without a single red run.
4:43
What waits for whatThe graph shows what each step waits for, laid out left to right, and the step that failed at the start of it.
4:56
FlowFlow is the run log: every attempt, the rows it wrote, how long it took and when it runs next.
5:06
Wrote nothingThe channel stock run ended after 38 seconds with nothing written: the portal refused the expired key.
5:14
Slowing downSecondary sales took fifteen minutes against a thirty-minute schedule. Two more slowdowns and it overlaps itself.

Know what you have

5:21
Know what you have. The catalog of governed tables, and where each one leads.
5:29
The catalogThe catalog holds 22 governed tables and 278 million rows. Nine need attention: three are failing checks and six are stale.
5:39
Correct and oldfact_channel_stock is gold certified and passes every check. It is also nine days old, which is the harder problem to notice.
5:49
Fed byEvery table names the connector that feeds it, so a stale number can be followed straight back to the cause.
6:02
LineageLineage follows each connector downstream. The two broken ones reach sixteen things between them.
6:10
The blast radiusThe channel stock feed lands in three tables and feeds a model, three dashboards and two alerts. Hover any of them to trace the line.
6:21
Who saw itAnd it says who was affected: 78 people opened those dashboards while the feed was down, and every one of them saw nine-day-old stock.

Who can see what

6:32
Who can see what. People, the policies that mask columns, and the audit trail.
6:39
AccessAccess shows everyone who can reach the data. One live account has no second factor, one has not been used in a month, and two invitations are still open.
6:52
Password onlyAn area sales manager with live access and a password only. The exemption was granted at onboarding in January as a temporary measure, and it is still in force.
7:06
DormantA leadership account unused for thirty days, still holding everything it was granted. Nothing has gone wrong, which is why nobody notices.
7:16
PoliciesPolicies decide which columns are masked, and for whom. A policy covers columns; a group can hold an exemption from it.
7:26
Switched offOne policy is written and switched off. Anyone checking whether customer comments are protected finds a rule and stops looking.
7:38
ActivityActivity is the audit trail, opened on the events that changed who can see something, or ran into that boundary.
7:47
RefusedAn area sales account tried to query receivables and was refused. The third attempt this month, which usually means somebody needs a number nobody has given them.

Fix it, live

7:59
Fix it, live. Repair the broken feed and watch the fix travel.
8:08
Fix itBack to the broken feed. With a fresh key in place, re-authenticate and re-run.
8:17
Six stagesThe run goes through six stages: connect, extract, land in open format in the company's own cloud, rebuild the models, run the quality gates, and publish.
8:30
HealthyThe feed is healthy and ran just now. Every table it lands in is current again.
8:41
One leftLineage now counts only the survey drop as broken. The channel stock chain is clear.
8:51
Current againOn the home page the channel stock tile is current, and its item has left the list.
9:01
In IntelligenceAnd in Antvia Intelligence, the Business pulse dashboard reads the same table. The warning on channel stock has gone; only the satisfaction tile is still flagged.
9:13
One platformOne fix, in one place, and every surface that reads the table knows about it.

Close

9:20
Data that says when it is wrong. Every connector, file and column in one register. Problems ranked, each linked to where it is fixed. Lineage from a broken feed to the people it misled. A fix in Data reaches every dashboard that reads it.
Next step

See it on your own data.

Thirty minutes on the half you came for. Bring the report that breaks every month, or the decision you never settled.