GIS Data Quality Checklist for Utilities: What to Test Before You Accept a Deliverable

What Is GIS Data QA/QC and Why Does It Matter for Utilities?

GIS QA/QC is the set of checks that proves a dataset meets its spec. First, quality assurance stops errors while the data is made. Then quality control finds the errors that slipped through.

For a utility, the test at handover is the last gate. After it, the data goes live. It also feeds outage tools, asset lists, network studies and field crews. So a broken link or a wrong voltage code is no longer just a data fault. Instead, it becomes a fault in daily work.

Most utilities also buy GIS data from outside firms. For example, they pay for data digitisation of paper maps, moves between platforms and field surveys. Each of these needs a test before payment. If there is no test, the utility takes on the defects. It then pays a second time to fix them.

Our piece on data quality in GE Vernova Smallworld GNM shows how bad data spreads. This article, however, looks at handover. In short, it covers what to test, how much to test and when to say yes.

What Pass Marks Should a Utility Set Before Data Capture Starts?

Set the pass marks in the scope of work before capture begins. Each one needs a test, a limit and a clear result if the limit is missed.

Because vague words such as “data shall be accurate” resist testing, state what you measure, what you compare it with and what counts as a pass. For instance, “every service line must link to a customer point and a source” is clear. In addition, a script can check it every time.

Next, freeze the data model and spec in one signed document. It should list feature classes, required fields, allowed values, link rules, the map grid and metadata fields. If the model changes mid-project, log it as a formal change. Otherwise, you test against a moving target, and that leads to disputes.

In Malaysia, public sector GIS has a national guide. The MyGDI standards work covers feature and attribute codes (MS1759), a metadata standard and dataset quality. So utilities that share data with agencies can match those codes early. As a result, later data exchange needs less rework.

Which Quality Elements Should a GIS Data Quality Checklist Cover?

A good checklist covers six elements. These are completeness, logical consistency, positional accuracy, thematic accuracy, temporal quality and usability. They come from ISO 19157-1:2023, the global standard for geographic data quality.

Using these terms gives both sides one shared language. So a defect report can say “logical consistency” and not “the lines look wrong”. The table below links each element to sample tests.

You do not need every test on every job. Still, write down which elements apply and which do not. That choice is part of the spec too.

How Do You Run Topology Checks and Value Tests on a Utility Network?

First, topology checks show that features link up the way the model says. Next, value tests show that each field holds an allowed value. Both should run by script on every feature, not just a sample.

Topology rules worth testing

Most GIS tools also ship with rule sets. For example, ArcGIS Pro topology has line rules such as Must Not Have Dangles and Must Not Overlap. It also has Must Not Self-Intersect and Endpoint Must Be Covered By. Similarly, the QGIS Topology Checker can flag dangles, duplicates and pseudo nodes.

• Dangles: cable ends that stop short of a device. So traces break.
• Overlaps: two lines drawn on top of each other. As a result, lengths are double counted.
• Pseudo nodes: breaks in a line with no real joint. They add clutter but no meaning.
• Loose ends: each cable end should land on a node, switch or transformer. If not, flag it.

Generic rules are only a start. Utilities also need network rules. For instance, phases should carry through, voltage should match across a joint and radial feeders should have no loops. In GE Vernova Smallworld, these checks often run on the network model itself.

Attribute domains and required fields

Attribute domains limit a field to valid values. A coded value domain uses a fixed pick list, while a range domain sets a low and a high limit. So the test is simple. Count each value outside its domain and each empty required field.

Also check how fields relate. For example, a transformer rating should match its type. Likewise, a build date should come before the capture date.

Then check the shapes themselves. The FME GeometryValidator, for instance, finds self-intersections, duplicate points and broken shapes. It can also try to repair them. But count the faults first, because silent repairs hide the vendor’s true error rate.

How Should You Test Location and Choose a Sample Size?

Test location by comparing a sample with a separate source of higher accuracy. Then use random samples for anything a rule cannot check. Also, set the sample size and pass mark in the contract.

The US National Standard for Spatial Data Accuracy (NSSDA) is a common method. It compares your data with a source of higher accuracy. Then it reports the result at the 95% confidence level. It asks for at least 20 check points. In addition, the check data should be collected apart from the data used to build the dataset.

Be clear about what location accuracy means for each source. For example, data traced from a schematic can be right in its links but rough in its place. In that case, set a tight limit against the drawing. Then set a looser limit against the ground. Otherwise, the vendor fails a test the source could never pass.

Sampling for manual checks

Manual checks take time, so most teams sample. For example, ISO 2859-1 sets sampling methods for inspection by attributes. It gives a structured way to pick a sample size. It also sets how many faults a batch may have before it fails.

Whichever method you choose, agree it first. Then pick samples at random across areas, crews and asset types.

Which QA/QC Checks Should Run by Script and Which Need a Person?

Script every check that can be written as a rule. Then run it on the full dataset. Keep people for judgement, such as reading drawings, spotting odd designs and testing real use.

Scripts are also cheap to repeat. So once a rule set exists, it can run on every batch and every resubmission. Tools such as FME suit this well. One workspace can read the vendor’s format and run shape, link and value tests. It then writes a defect report. After that, the native tools in ArcGIS, QGIS or Smallworld confirm the data works in the target system.

In addition, keep script and manual results in one defect log. Each defect should have a feature ID, the rule broken, the quality element, a severity and a status. As a result, you can report error rates per batch. So you no longer argue over screenshots.

What Does a Practical GIS Data Quality Checklist Look Like?

A practical checklist runs in four phases. First comes the work before delivery. Next comes receipt, then testing and finally sign-off.

Use the list below as a start. Then adapt the limits to your own spec.

Metadata needs its own line, because teams often skip it. ISO 19115-1 sets a schema to describe geographic data. It covers extent, quality, reference systems and distribution. At the very least, record lineage, capture method and date. Then future users know how far to trust the data.

What Are the Most Common Mistakes in GIS Data Acceptance?

The most common mistakes are testing too late and testing against the wrong source. Another is letting repairs hide the real error rate. In each case, data passes on paper but fails in the field.

There is also a softer mistake. Some teams treat QA/QC as the vendor’s job alone. The vendor should run its own checks. But the client still needs a separate test, because otherwise nobody tests the tester.

How Should Utilities Write QA/QC Into Vendor Contracts?

Write QA/QC into the contract as clear deliverables with a set process. That way, the vendor knows the tests, limits and rework terms before pricing the job.

Also, tie payments to accepted batches, not delivered ones. In addition, ask the vendor to send its own defect log with each batch. If its numbers differ a lot from yours, its own checks may be weak.

Finally, think about a third party for large jobs, such as a data migration. A third party has no stake in the result. It can also reuse rule sets across vendors. For example, our GIS QA/QC services cover topology checks, value checks and error detection.

Frequently Asked Questions

What Should Utilities Do Next to Improve Data Acceptance?

First, start with the spec. If your contracts lack clear pass marks per element, fix that before the next tender. Then build a rule set once and reuse it on every batch.

A sound GIS data quality checklist protects every system that relies on the GIS. So the effort at handover is small next to the cost of fixing a live network later. If you want a second view on vendor data, or help writing clear tests, get in touch with our team.