Digital twin pilots: a site checklist for teams with GenAnywhere
A short checklist that keeps a pilot pointed at one decision instead of an impressive demonstration.
A digital twin pilot fails for ordinary reasons. The scope grows, the data never arrives, or the model looks impressive and answers nothing. A short checklist keeps the early months pointed at a decision instead of a demonstration. The list below is short on purpose, because a pilot with ten objectives usually satisfies none of them.

1 What a pilot should decide before it starts
The useful question is not which platform to buy. It is which decision the pilot should improve and who will make that decision differently as a result.
1.1 Pick one decision, not a platform
A pilot that tries to prove the value of a twin in general usually proves nothing. Pick a decision with a date, such as whether to keep a maintenance yard or move it. Then ask what information the decision needs and whether a 3D site model supplies it. A narrow decision gives the pilot a finish line and a way to measure whether the work helped. If nobody can name the decision, the pilot has already found an early problem.
1.2 Define the area and the edge
Draw the pilot boundary on a map and write down why it stops where it does. A boundary that follows a fence line is easier to explain than one that follows a data extent nobody can see. Include the access roads and the neighboring parcels that shape the decision, then state what sits outside the scope. A boundary that includes a road but excludes the junction is likely to fail in practice, so walk the edge before signing off.
2 Build a site layer inventory
A scene is a stack of layers, and each layer carries a source and a date. The inventory is the part of the pilot that survives long after the demo.
2.1 List what the scene must contain
Start with the layers the decision needs: buildings, terrain, water, vegetation, roads, and any site feature the team will point at in a meeting. With GenAnywhere, a team maps the pilot area and generates an editable 3D site model from satellite and map data, with buildings, terrain, and vegetation kept as separate layers. Volumes can be replaced or adjusted to test a scenario, and the scene can be previewed before export. The inventory should name each layer, its source, and its date. A layer without a source is a rumor with geometry.

2.2 Separate known from inferred
Some layers come from mapped imagery and some are inferred, such as building height or use. Mark the difference in the inventory. A pilot that treats an estimate as a measurement will make confident decisions on soft ground. The mark also tells a reviewer which parts of the scene need a field check before the next phase, which is the kind of finding a pilot should produce. A reviewer can then decide which estimates to replace before the next phase, rather than discovering the gap at handoff.
3 Set the handoff format early
Pilots often end with a scene that cannot leave the tool that made it. Choosing the handoff format at the start prevents that outcome and forces a useful conversation about the receiving workflow.
3.1 Choose the tool the team will use
Ask which environment the team will open next: a game engine, a BIM tool, a GIS platform, or a rendering package. Each has its own expectations about scale, orientation, and object grouping. Write those expectations into the pilot brief. A scene that opens cleanly in the intended tool is more valuable than a richer scene that requires a rebuild. If two tools are plausible, test both with a small export rather than choosing on a feature list.
3.2 Check the export against the brief
GenAnywhere exports a general GLB for 3D, BIM, GIS, and game engine workflows, including Unreal Engine, Unity, Blender, Revit, Rhino, SketchUp, ArcGIS Pro, GISBox, and Wings Engine. The export carries a permanent commercial license, with limits on resale of raw files and on AI training use. Before the pilot closes, open the export in the receiving tool and check scale, orientation, layer names, and object grouping against the brief. Receiving teams still confirm their own data rights and fit for purpose. Keep the test export with the brief so a later reviewer can see what was checked.

4 Plan the evaluation
A pilot needs a definition of useful before it starts. Otherwise the team debates impressions at the end.
4.1 Agree on what counts as useful
Write two or three questions the pilot should answer, such as whether the team can brief a site option without a site visit or whether a reviewer can spot a conflict in the scene. Score the answers at the end with the people who attended. A simple before-and-after comparison is enough. The point is to learn whether the scene changed a conversation, not whether it impressed a visitor. Write the answers down while the memory of the meeting is fresh, and keep the notes with the pilot brief.
4.2 Write the stop conditions
Decide in advance what would end the pilot early: a data gap that cannot be closed, a handoff that fails in the receiving tool, or a decision that moves outside the pilot area. Stop conditions protect the budget and the team's attention. They also make the pilot easier to approve, because the sponsor can see that the work has an edge. Review the conditions at each checkpoint; a stop condition that nobody checks is decoration. A checkpoint that nobody can attend is not a checkpoint, so schedule it with the people who make the decision.
5 Keep the pilot from drifting
Pilots drift when small additions arrive without a record. Two habits keep the scope visible.
5.1 Date the source and the model
Record the imagery date, the date the scene was built, and the date of any later edit. When a colleague asks whether the model shows a new building, the dates answer the question. They also make a later comparison possible, because the team can tell which version of the scene supported which meeting. A short version note is enough, such as the date and the layers that changed.
5.2 Record what the pilot did not include
List the layers, areas, and questions the pilot left out. Say where verified data is still needed and who will supply it. That list becomes the agenda for the next phase and prevents a pilot from being described later as a complete twin. Keep it with the brief so the next team knows what to verify.
6 Bringing the pilot checklist together
A one-page brief and a short trial will test more than a long slide deck.
6.1 Write the one-page brief
Put the decision, the boundary, the layer inventory, the handoff format, and the evaluation questions on one page. If the brief needs a second page, the pilot is probably trying to answer too much. The site workflow overview shows which steps belong in the trial.
6.2 Run a short trial on real ground
Use a site the team already knows. A familiar place exposes errors in scale, orientation, and layer naming that a showcase site would hide. Invite the people who will use the scene, not just the people who built it, and record where they hesitate. Hesitation usually points to a naming problem or a scale problem, and both are cheap to fix early.
6.3 Decide what carries forward
At the end, decide which layers, views, and habits continue into the next phase. The Khronos glTF overview is a useful reference for what a handoff file can and cannot carry. Write the decision down and assign an owner, so the pilot ends with a next step rather than a folder of files. Assign a review date as well, so the next phase has a checkpoint instead of an open-ended promise.
Open the GenAnywhere workspace to build a pilot scene on a site your team knows.
See how the site workflow fits together before the pilot brief is approved.
7 FAQ
FAQ-1 What should a pilot prove?
It should prove that a shared 3D site scene improves a specific decision for a specific team. A pilot is not a general demonstration of the technology and does not need to cover every use case. Write the decision down, then measure whether the people who make it found the scene useful. The strongest evidence is usually a changed conversation, not a larger model.
FAQ-2 Does GenAnywhere create a full replica?
No. The platform builds initial site context from satellite and map data: buildings, terrain, water, and vegetation in layers. It does not connect live sensors, asset registries, or operations data, and it does not maintain a constantly updated replica. A team can export the scene as GLB and combine it with other work, but the result is a site context, not a complete twin.
FAQ-3 Which export format should a pilot use?
The general GLB export is the practical choice for most pilots because it moves into 3D, BIM, GIS, and game engine workflows. Before committing, open a test export in the receiving tool and check scale, orientation, layer names, and grouping. Confirm the license terms as well, since raw files cannot be resold or used for AI training and third-party data rights need separate review.
FAQ-4 How should a team judge pilot success?
Judge it against the questions written before the pilot started. Compare how the team briefed and decided before and after the scene arrived, and note where people hesitated or asked for missing information. A short review with the people who attended is enough. If the scene did not change a decision, that is a finding worth recording rather than a failure to hide.
Content and technical notes
This article was prepared by the GenAnywhere team with AI-assisted drafting and human review.
Product features, formats, and availability change as the platform evolves; refer to the official GenAnywhere website for current details.
For technical questions or corrections, contact support@genanywhere.com.
Last reviewed: 2026-11-09.