Design the records around real questions
Before fieldwork, ask the eventual users what they need to find, compare, and update. A facilities team might search for equipment by building and asset identifier. A drainage team might need connections between structures. These uses determine whether the dataset needs points, lines, polygons, related tables, attachments, or a combination.
Create a data dictionary that explains every field, permitted value, unit, and treatment of missing information. Distinguish unknown from not applicable. Leaving both as an empty cell makes later analysis ambiguous. Stable asset identifiers should survive a change in name, condition, or location so that observations can be connected across future surveys.
Test the collection form before scaling up
Pilot the form on a small, varied sample of assets. Include awkward cases: an unreadable label, multiple components at one location, inaccessible equipment, and an asset that does not fit the initial categories. Review the pilot with the receiving team. This is the right time to change the schema, before inconsistent records multiply.
Esri's documentation describes domains as lists or ranges of permitted field values. Such controls can reduce avoidable entry variation, but they should reflect the real assets. Include a clear exception process for unusual observations. Forcing a fieldworker to select an incorrect category creates cleaner-looking data at the cost of factual accuracy.
Treat the position as a documented observation
Define where each asset is represented. Is the point at the visible cover, the equipment centre, or a label mounted beside it? A consistent capture convention prevents false disagreement between later surveys. Record the coordinate reference system and the positioning method, together with any quality information needed for the intended use.
The map should also distinguish direct observations from positions copied from records or estimated from other features. If a location cannot be obtained under suitable conditions, mark it for review. A visually plausible position moved onto a background image without a recorded reason can lose its connection to the original evidence.
Validate meaning as well as file structure
Automated checks can find duplicate identifiers, missing required values, invalid categories, and broken attachment links. Geometry checks can examine relationships appropriate to the data model. Apply rules selectively: a gap may be an error in one network and a real condition in another. Every check needs a defined interpretation and a route to correction.
Manual review remains useful for questions that rules cannot settle. Does the photograph show the recorded asset? Does the condition description agree with the evidence? Are all expected areas covered? Compare collection progress with an agreed area or asset inventory so that an attractive map does not hide an omitted courtyard or building level.
Deliver a dataset the owner can maintain
Agree the receiving format early and test a sample import. Check field types, date handling, text length, attachments, and identifier preservation. A successful export does not prove that the asset system will interpret the information correctly. Include a concise handover session with someone who will actually manage updates.
The delivery should contain the data dictionary, coverage statement, collection dates, source descriptions, and a register of unresolved records. Name the authoritative copy and define how later corrections will be made. Without an ownership process, repeated spreadsheets and emailed edits can fragment an otherwise well-built dataset soon after handover.
Common questions
Is a spreadsheet enough for GIS delivery?
It may be adequate for a simple point inventory if coordinates, reference information, and field definitions are included. Networks, complex geometry, and related records may need a more suitable spatial format.
Should existing records be replaced during a survey?
Review them against the agreed update policy. Preserve useful identifiers and history, record the new evidence, and flag conflicting information for resolution instead of silently overwriting uncertain values.
Further reading
This article was rewritten in October 2026 around the subject of the original Geobiro post. The original publication date is retained; the update date identifies the current text.