ArcGIS organization management, whether in ArcGIS Online or ArcGIS Enterprise, means juggling many moving parts. Users, licenses, content, groups, and credits change constantly, and keeping everything aligned requires more than one-time setup or occasional cleanup. Many organizations invest time defining Web GIS governance policies, but they often struggle to put those policies into practice. Without regular visibility into how people actually use the organization, even a well‑intentioned ArcGIS environment can quickly become unstructured and difficult to manage.
To help close that gap, we created the Admin Insights Template designed to support Web GIS governance in ArcGIS Online through regular monitoring and visibility. Built using ArcGIS Notebooks and ArcGIS Dashboards, the template surfaces actionable insights about an organization’s users, content, and groups. These insights help administrators understand whether governance policies are being followed and where intervention may be needed. By providing ongoing visibility into organizational health and usage, the template supports the Observability pillar from Esri’s Architecture Center.
What the Template Provides
The admin insights template leverages the ArcGIS API for Python to power a set of dashboards that surface insights across three key areas of ArcGIS organization management: users and licenses, content, and groups. Together, these dashboards help administrators understand how their organization is being used day to day, not just how it was configured.
The template complements the new Organization Status Dashboard (beta) available in ArcGIS Online, giving administrators a more complete picture of organizational activity and resource usage. Some of you may also recognize this pattern from the earlier blog Managing ArcGIS Online content with ArcGIS Dashboards and ArcGIS Notebooks. The admin insights template builds on that foundation by expanding the approach beyond content to support broader organizational monitoring and ongoing observability.
Users & Licenses: The Users and Licenses dashboard provides visibility into how an ArcGIS organization’s users are structured and how they are being used over time. You can use this dashboard to identify inactive or underutilized accounts, validate that licenses are assigned efficiently, and spot trends that may signal the need to adjust onboarding processes, licensing strategy, or user management policies.
Content: The Content dashboard helps you understand what content exists across the organization and how it is being used. It highlights stale or unused items, supports validation of sharing and metadata standards, and tracks the growth of non‑authoritative content. You can also monitor high‑impact items (publicly shared, highly viewed, or large content), and explore upstream and downstream content dependencies.
Groups: The Groups dashboard provides insight into how groups are being used to organize content and manage access across the organization. It helps identify groups that support critical sharing workflows, surface unused or low‑value groups (such as empty or single‑member groups), and validate that group sharing settings align with organizational access and collaboration policies.
How It Works
The admin insights template brings these dashboards together using an ArcGIS Experience Builder app, which serves as a single storefront for exploring users, content, and groups in one place.
Each dashboard is powered by a shared hosted feature service that maintains an inventory of users, content items, and groups. This feature service is updated by a notebook that uses the ArcGIS API for Python to collect details about each user, content item, and group in your organization. You can schedule this notebook to run regularly to keep the dashboards up to date.
The template was built using ArcGIS Online, but the same approach can be adapted for deployment in ArcGIS Enterprise.
Getting Started
Here’s the basic workflow to set up the admin insights template in your organization:
- Access the initialization notebook at Admin Insights Template – Initialization Notebook. If not already, log in to your organization in order to open the notebook.
- With the initialization notebook now open, choose the Save As option to store it as an item in your organization.
- Run all cells in your copy of the initialization notebook
- Once complete, take note of the
CLONED ITEMS SUMMARYthat is returned after the last cell. This lists the five new items created in your content. You will input the item ID of the data table in step 9 below. - You should now see five new items in your content:
- If you access the dashboards at this point, they will be empty! This is expected behavior. You need to create a copy of the inventorying notebook to populate the underlying data table.
- Access the inventorying notebook at Admin Insights Template – Organization Inventory Notebook. If not already, login to your organization in order to open the notebook.
- With the inventorying notebook now open, choose the Save As option to store it as an item in your organization.
- Open your copy of the inventorying notebook and update the
fs_idvariable to the new data table item id generated when you ran the initialization notebook in step 4 above. - Run all cells in your copy of the inventorying notebook.
- The underlying data table should now be updated, and the admin insights app loads as expected! You can consider scheduling the inventorying notebook to run weekly to observe your organization closely.
Road Ahead
Good governance is an ongoing practice. We hope this template gives your team a practical foundation for observing and managing your ArcGIS organization with confidence. Work on the admin insights template is ongoing, so keep an eye out for updates. And feel free to drop a comment or question below. We’d love to hear your feedback!
In step 10,
upstream_enriched, downstream_enriched, final_content_df = run_dependency_pipeline(
gis=gis,
items_dicts=items_dicts,
content_df=content_df,
dependency_mapping=dependency_mapping
)
returns
—————————————————————————
ValueError Traceback (most recent call last)
Cell In[16], line 1
—-> 1 upstream_enriched, downstream_enriched, final_content_df = run_dependency_pipeline(
2 gis=gis,
3 items_dicts=items_dicts,
4 content_df=content_df,
5 dependency_mapping=dependency_mapping
6 )
ValueError: too many values to unpack (expected 3)
Hi Kevin! We’ve resolved the dependency unpacking issue and updated the ArcGIS Admin Insights – Organization Inventory Notebook. I recommend reopening the notebook from the original link and rerunning it to ensure you’re using the latest version.
Being able to track upstream and downstream content is a very valuable ability (and a long time requested feature) not just for administrators but also for content owners, as they are the ones usually doing the creating, editing, and deleting. How can non-admins see this kind of information?
Looks good now, thanks.
You would need to be an admin to run the two notebooks and populate the table. But once that is done, I think you can just share the table, experience and 3 dashboards to a group.
Thanks for the article. Can you deliver this as a standard ArcGIS solution?
Is there a minimum supported version of ArcGIS Enterprise to use this tool?
This toolkit is currently configured for ArcGIS Online usage. A few planned updates include easier deployment in ArcGIS Enterprise organizations, support for monitoring multiple organizations, and additional flagging. This is a living toolkit, so keep an eye out for updates!
Tried to run against 11.5 and no joy, moved the table using the offline content manager and then ran teh inventory notebook. That fails at the getSubscritpionIno as not valid for an Enterprise Portal.
Hi & thanks for providing all of this. I tried running it in ArcGIS Online this morning for the first time and got to:
Publishing to table: Content
Total records to publish : 11,048
Chunk size : 500
Publishing rows 1–500…
Exception Traceback (most recent call last)
Cell In[16], line 30
24 tables_to_publish.extend([
25 (upstream_enriched, tables[“upstream”][“table”]),
26 (downstream_enriched, tables[“downstream”][“table”]),
27 ])
29 for df, table in tables_to_publish:
—> 30 publish_dataframe_to_table(df, table)
32 print(“\n✓ Content snapshot complete”)
Exception:
Field description has invalid html content.
(Error Code: 400)
Super exciting. For us, the real use would be in portal, we’re still at 11.3. When implementing, thought it would be worth noting.
Hello
This is awesome—I just ran it, and like @Chris Taylor, I’m getting the same error.
*-*-*-*-*-*-*-*
Exception Traceback (most recent call last)
Cell In[24], line 30
24 tables_to_publish.extend([
25 (upstream_enriched, tables[“upstream”][“table”]),
26 (downstream_enriched, tables[“downstream”][“table”]),
27 ])
29 for df, table in tables_to_publish:
—> 30 publish_dataframe_to_table(df, table)
32 print(“\n✓ Content snapshot complete”)
Cell In[11], line 118, in publish_dataframe_to_table(df, table_layer, chunk_size)
115 print(f”\nPublishing rows {start + 1}–{end}…”)
117 features = [Feature(attributes=row.to_dict()) for _, row in chunk.iterrows()]
–> 118 result = table_layer.edit_features(adds=features)
119 add_results = result.get(“addResults”, []) or []
121 attempted = len(features)
File /opt/conda/lib/python3.13/site-packages/arcgis/features/layer.py:3510, in FeatureLayer.edit_features(self, adds, updates, deletes, gdb_version, use_global_ids, rollback_on_failure, return_edit_moment, attachments, true_curve_client, session_id, use_previous_moment, datum_transformation, future)
3508 return EditFeatureJob(future, self._con)
3509 # return future
-> 3510 return self._con.post_multipart(path=edit_url, postdata=params)
3511 except Exception as e:
3512 if str(e).lower().find(“Invalid Token”.lower()) > -1:
File /opt/conda/lib/python3.13/site-packages/arcgis/gis/_impl/_con/_connection.py:1274, in Connection.post_multipart(self, path, params, files, **kwargs)
1272 if return_raw_response:
1273 return resp
-> 1274 return self._handle_response(
1275 resp=resp,
1276 out_path=out_path,
1277 file_name=file_name,
1278 try_json=try_json,
1279 force_bytes=kwargs.pop(“force_bytes”, False),
1280 )
File /opt/conda/lib/python3.13/site-packages/arcgis/gis/_impl/_con/_connection.py:1005, in Connection._handle_response(self, resp, file_name, out_path, try_json, force_bytes, ignore_error_key)
1003 return data
1004 errorcode = data[“error”][“code”] if “code” in data[“error”] else 0
-> 1005 self._handle_json_error(data[“error”], errorcode)
1006 return data
1007 else:
File /opt/conda/lib/python3.13/site-packages/arcgis/gis/_impl/_con/_connection.py:1028, in Connection._handle_json_error(self, error, errorcode)
1025 # _log.error(errordetail)
1027 errormessage = errormessage + “\n(Error Code: ” + str(errorcode) + “)”
-> 1028 raise Exception(errormessage)
Exception:
Field description has invalid html content.
(Error Code: 400)
*-*-*-*-*-*-*-*-*-*-*-*-*-
I’ve tried running the inventory notebook 3 times, and so far, it fails at exactly the same place. My org has 12.5K users, over 160k+ items, and 1700 groups. The inventory script gets to step 3c.1 and errors out while building the Governance table for all 160K+ items, specifically at processing 40100/164275 items. Is your script limited in how many records it can process and list for the organizations items? Please feel free to email me directly.
Hi Alex! Are you running the Notebook manually, or scheduling it? If you are doing it manually, I would suggest scheduling the Notebook using the “Tasks” tab because this ensures that the kernel won’t die before the script has finished its inventory. There should be no limit to the number of items it can process. Let us know if this resolves the issue or if you are receiving a more specific error!
For those who want to try and get this up and running in your ArcGIS Enterprise…godspeed and hope this helps. Esri was a little optomistic on this: “…can be adapted for deployment in ArcGIS Enterprise with minimal changes.” Also, after all the work to populate everything, I went to open the Dashboards/Experience Builder app and since we are on 11.5, it doesn’t work! @Esri any way you can share a downgraded version or a way to downgrade it?
Initialization Notebook
1. Dual GIS connections required
The source items live on AGOL. A second anonymous (or authenticated) AGOL connection is needed to fetch them:
pythongis = GIS(“pro”) # Enterprise target
gis_agol = GIS() # AGOL source
All gis.content.get() calls for source items must use gis_agol, while all clone_items() calls use gis.
2. copy_data=False on hosted table clone
Cross-platform cloning fails with a KeyError: ‘objectid’ case mismatch when copying data. The table is empty by design anyway:
pythongis.content.clone_items(…, copy_data=False)
3. force=True on remap_data()
The original AGOL item IDs don’t exist in Enterprise, so validation fails without this:
pythoncloned_item.remap_data({hosted_table: cloned_table_id}, force=True)
4. from_dash=True on dashboard clone_items()
Cross-platform dependency resolution fails on dashboards without this:
pythongis.content.clone_items(…, from_dash=True)
5. Native Enterprise hosted table required
clone_items() creates an item in Enterprise that still points to the AGOL-hosted service (services.arcgis.com). The table must be recreated natively using create_service() + add_to_definition() with the schema extracted from the AGOL source:
pythonpublished_item = gis.content.create_service(
name=”ArcGIS_Admin_Insights_Table”,
service_type=”featureService”,
folder=target_folder
)
flc = FeatureLayerCollection.fromitem(published_item)
flc.manager.add_to_definition({“tables”: table_definitions})
Organization Inventory Notebook
1. subscriptionInfo not available on Enterprise
build_license_summary() calls gis.properties.subscriptionInfo which is AGOL-only. Add an Enterprise fallback that derives license counts from user attributes instead.
2. Pass fs_item not agk_fs to get_table_url()
The function expects an Item object, not a FeatureLayerCollection.
3. Access tables directly via agk_fs for write operations
Token/permission errors occur when using table references returned by get_table_url(). Use direct table access instead:
pythonusers_table = next(t for t in agk_fs.tables if t.properties.name == “Users”)
4. Timestamp columns must be converted to millisecond epochs
Enterprise rejects string dates with Cannot convert ‘String’ to ‘TIMESTAMP’. Convert before publishing:
pythondf[col] = pd.to_datetime(df[col], errors=’coerce’).apply(
lambda x: int(x.timestamp() * 1000) if pd.notnull(x) else None
)
Affects: creation_date, last_updated, last_viewed in Content, Upstream, and Downstream tables.
5. Float size columns must be cast to Int64
Enterprise rejects float64 in INTEGER schema fields with Cannot convert ‘BigDecimal’ to ‘INTEGER’:
pythondf[‘size_mb’] = df[‘size_mb’].astype(‘Int64’)
6. Filter external hosts before fetching item sizes
The default add_item_sizes() call attempts to fetch sizes for all 1,790 items including externally referenced services, causing multi-hour runtimes. Filter to internal items only inside run_governance_pipeline():
pythoninternal_ids = set(
i[‘id’] for i in items
if i.get(‘url’) is None
or ‘your-enterprise-host.com’ in str(i.get(‘url’, ”))
or i.get(‘url’) == ”
)
internal_df = df[df[‘item_id’].isin(internal_ids)].copy()
Hi Jon, thank you for bringing this issue to our attention. This error typically occurs when a string value in the data frame being published to a feature layer contains invalid HTML formatting. I was able to reproduce the behavior on my end and have added an update to the notebook’s publishing function that resolved the issue in my testing. Hopefully, this addresses the problem for you!
Is there a timeline already set to get this to work within ArcGIS Enterprise 11.5 and above?
This is great! Two questions:
First, I’m getting an odd thing. I’m logged in to AGOL, and I go to the Experience Builder just fine, but then when I click on a tab to load the embedded dashboard, I am prompted to sign in again. I’ve tried a few experiments, and it does it every time. Once I’ve logged in, I can jump between dashboards. But if I close the Experience Builder and re-open it, I’m prompted to log in again. I’m not sure why it doesn’t inherit my credentials.
Second, any way to get credits usage loaded in here too? I like the new credit dashboard ok, but this is way more useful as a full picture of everything for governance.
Hi Ryan, this is expected behavior; however, there’s an easy fix to prevent the repeated login prompts.
The simplest solution is to update the embedded Dashboard URLs in your Experience Builder app to use the full organization-specific URL, and then republish the app.
Make sure each embedded dashboard uses the URL copied directly from the published dashboard with your organization’s name. It should look something like:
https://{YOUR_ORGANIZATION}.maps.arcgis.com/apps/dashboards/{ITEM_ID}
To apply the fix:
– Open your Experience Builder app in edit (draft) mode
– Go to each page that contains an embedded Dashboard
– In the left panel, select the Embed widget
– Replace the existing URL with the full organization-specific Dashboard URL (copied from the published Dashboard)
– Click Save, then Publish
Once updated, the login prompts should stop occurring when navigating between Dashboards.
I’m running it manually so I can evaluate it. I was planning on running it as a task after the initial run so I could see what it turns out. I’ll see if a task helps it.
Has anyone found a solution to this error I’m encountering in the last cell while running Step 3c.3 (Publish to Content Table)?
I’ve posted the error here before, but haven’t received a response from the blog author. I’m wondering if anyone else has experienced the same issue and, if so, how they resolved it.
Thanks
Error:
Exception Traceback (most recent call last)
Cell In[16], line 30
24 tables_to_publish.extend([
25 (upstream_enriched, tables[“upstream”][“table”]),
26 (downstream_enriched, tables[“downstream”][“table”]),
27 ])
29 for df, table in tables_to_publish:
—> 30 publish_dataframe_to_table(df, table)
32 print(“\n✓ Content snapshot complete”)
Cell In[5], line 118, in publish_dataframe_to_table(df, table_layer, chunk_size)
115 print(f”\nPublishing rows {start + 1}–{end}…”)
117 features = [Feature(attributes=row.to_dict()) for _, row in chunk.iterrows()]
–> 118 result = table_layer.edit_features(adds=features)
119 add_results = result.get(“addResults”, []) or []
121 attempted = len(features)
File /opt/conda/lib/python3.13/site-packages/arcgis/features/layer.py:3510, in FeatureLayer.edit_features(self, adds, updates, deletes, gdb_version, use_global_ids, rollback_on_failure, return_edit_moment, attachments, true_curve_client, session_id, use_previous_moment, datum_transformation, future)
3508 return EditFeatureJob(future, self._con)
3509 # return future
-> 3510 return self._con.post_multipart(path=edit_url, postdata=params)
3511 except Exception as e:
3512 if str(e).lower().find(“Invalid Token”.lower()) > -1:
File /opt/conda/lib/python3.13/site-packages/arcgis/gis/_impl/_con/_connection.py:1274, in Connection.post_multipart(self, path, params, files, **kwargs)
1272 if return_raw_response:
1273 return resp
-> 1274 return self._handle_response(
1275 resp=resp,
1276 out_path=out_path,
1277 file_name=file_name,
1278 try_json=try_json,
1279 force_bytes=kwargs.pop(“force_bytes”, False),
1280 )
File /opt/conda/lib/python3.13/site-packages/arcgis/gis/_impl/_con/_connection.py:1005, in Connection._handle_response(self, resp, file_name, out_path, try_json, force_bytes, ignore_error_key)
1003 return data
1004 errorcode = data[“error”][“code”] if “code” in data[“error”] else 0
-> 1005 self._handle_json_error(data[“error”], errorcode)
1006 return data
1007 else:
File /opt/conda/lib/python3.13/site-packages/arcgis/gis/_impl/_con/_connection.py:1028, in Connection._handle_json_error(self, error, errorcode)
1025 # _log.error(errordetail)
1027 errormessage = errormessage + “\n(Error Code: ” + str(errorcode) + “)”
-> 1028 raise Exception(errormessage)
Exception:
Field description has invalid html content.
(Error Code: 400)
Hi Ahmad, thank you for bringing this issue to our attention. This error typically occurs when a string value in the data frame being published to a feature layer contains invalid HTML formatting. I was able to reproduce the behavior on my end and have added an update to the notebook’s publishing function that resolved the issue in my testing. Hopefully, this addresses the problem for you!
This is great and highly needed! Thank you so much for all the work you folks put into doing this. Unfortunately I’m running into a similar, but different, error as well on Step 3:
Cloned Hosted Table: Admin Insights Template – Data Table
Cloned Dashboard: Admin Insights Template – Content Dashboard (No Dependencies)
Cloned Dashboard: Admin Insights Template – Groups Dashboard
Cloned Dashboard: Admin Insights Template – Users and Licensing Dashboard
—————————————————————————
IndexError Traceback (most recent call last)
Cell In[15], line 10
2 cloned_dashboards_id, cloned_dashboards, cloned_table_id = clone_dashboards_with_shared_table(
3 gis,
4 original_dashboards,
5 dependency_mapping,
6 target_folder
7 )
9 # Clone Experience Builder
—> 10 cloned_exb = clone_remap_exb(
11 gis,
12 cloned_table_id,
13 cloned_dashboards_id,
14 admin_insights_exb,
15 target_folder
16 )
18 # Summarize Clones
19 summarize_cloned_items(
20 gis,
21 cloned_table_id,
22 cloned_dashboards,
23 cloned_exb
24 )
Cell In[14], line 127, in clone_remap_exb(gis, cloned_table_id, cloned_dashboards_id, admin_insights_exb, target_folder)
120 # Clone the EXB with table remapping
121 cloned_exb_list = gis.content.clone_items(
122 items=[exb_item],
123 folder=target_folder,
124 item_mapping={hosted_table: cloned_table_id},
125 preserve_item_id=False
126 )
–> 127 cloned_exb = cloned_exb_list[0]
129 data = cloned_exb.get_data()
131 updated_data = data
IndexError: list index out of range
Bonjour, C’est une solution essentielle pour les organisations ArcGIS Online dont les usages sont nombreux. Vivement une solution pour les environnements ArcGIS Enterprise! Merci beaucoup