What the video shows
The demo walks through the canonical Banana Analytics workflow. It starts with an explicit, identifiable hospital (AdventHealth Orlando) so the math behind the service area is defensible. We resolve the five-county Primary Service Area from CMS fee-for-service claims rather than guessing from ZIP codes. From there: a scored cohort, a visual pin to the Canvas workspace, an AI-drafted CHNA narrative anchored on the cohort's actual data, an assistant interaction to generate concrete strategies for a finding the narrative surfaces, and a final export.
Every step matters for a CHNA that survives IRS review under 26 CFR §1.501(r)-3: the data is sourced and dated, the methodology is consistent across counties, the cohort can be re-anchored when service-area data shifts, and the narrative names the data it uses.
Chapter notes
Each section is short. You can deep-link to any of them on YouTube if you want to skip around. Time codes link to the video on YouTube.
- 010:00
Define a hospital service area
Search for a hospital by name (AdventHealth Orlando in this demo). The platform identifies the primary service area from CMS fee-for-service data. Five counties, no manual mapping.
- 020:55
Save the cohort and open the Geographic Explorer
Weighted scores across disease burden, environmental risk, and SDOH render instantly. A cohort-filtered choropleth shows Community Health Risk across all five counties.
- 031:30
Pin the map to a new Canvas workspace
One click captures the active map view as a chart block in a new Canvas. The visual becomes evidence that anchors the report's narrative claims.
- 041:45
Generate the AI summary
Pick the Community Health Needs Assessment preset; the platform synthesizes millions of data points into a localized, IRS §501(r)(3)-aware narrative. The PDF shows up in the summary inbox when ready.
- 053:00
Tour the Canvas workspace
Document editor in the center, pinned sources on the left, AI assistant on the right with retrieval-augmented access to your cohort data and the full platform.
- 063:30
Ask the AI assistant for actionable strategies
Highlight a key finding (mental distress across the cohort), ask for three concrete strategies. Telehealth expansion, behavioral health integration, community health worker programs. All copy-paste-ready.
- 074:50
Export the finished report
PDF (branded print), HTML (digital-first publishing), or Markdown (for CMS / Notion / re-use). One click; brand colors and logo applied automatically.
What didn't make it into the video
The cohort is portable. Once you've defined a service area in the platform, you can also build it from a ZIP attribution file, drive-time isochrone around a facility address, or a state boundary. The downstream workflow (scored cohort → Canvas → narrative → export) is identical regardless of how the cohort got assembled.
Every numeric claim is verified. Before the AI summary lands in your inbox, the generated narrative is run through a separate verifier model with structured-output support that checks every quantitative claim against the source payload. Each claim gets labeled VERIFIED, DERIVED, UNSUPPORTED, or INTERPRETIVE; the UI surfaces the count of unsupported claims on the PDF so you know what to scrutinize. The generator and verifier are independently selected from the current frontier model lineup based on empirical testing, and the specific models used on each report are listed on the PDF's methodology page.
The Canvas survives revisions. Every change to a Canvas document is auto-saved; you can also save named revisions (think Google Docs version history) and roll back. That matters when a CHNA is reviewed by multiple stakeholders and the document needs to track changes through board approval.
Brand-aware exports. The PDF, HTML, and Markdown exports all pull from a single account-branding configuration: logo, primary color, footer text. Consultants running parallel CHNAs for multiple hospitals can swap brand settings between exports without re-styling the output.
Full transcript
Cleaned of filler beats. Useful for skimming the workflow, quoting in proposals, or feeding to a screen reader.
Welcome to Banana Analytics. Today, we'll walk through how to go from raw data to a comprehensive, actionable Community Health Needs Assessment in just a few minutes.
Let's start by defining our service area. While you can upload custom zip codes, our platform also allows you to simply search for a hospital by name or ID. We'll select AdventHealth Orlando, the first health system in our list. The platform automatically identifies the surrounding counties that make up its primary service area based on CMS fee-for-service data. With a single click, we build our cohort. Instantly, Banana Analytics calculates weighted scores covering disease burden, environmental risk, and social determinants of health across the cohort region.
We'll save this cohort to unlock deeper geographic analysis in our platform and then we'll head over to the Geographic Explorer where we'll generate our summary and explore the canvas editor.
By having our saved cohort automatically filtered, we get a highly detailed visual of the Community Health Risk across all five counties. This is exactly the kind of visual evidence we want in our final report. We'll capture this map and pin it directly to a new Canvas workspace.
Now for the heavy lifting. Our counties are already selected as part of this cohort so now we need to simply click Generate AI Summary. We'll choose the Community Health Needs Assessment preset, or the appropriate style template depending on your role, and hit generate. In the background, the platform is synthesizing millions of data points into a cohesive, localized narrative. Once our summary is ready in the inbox, we can review the generated PDF to see our baseline findings.
Now that our summary is complete, we have a great starting point. Let's take a quick look at the PDF version. From here, it's easy to download high-quality, professional reports complete with beautiful custom charts. Next, we'll learn how to import this summary directly into the Canvas to customize it even further.
Welcome to the Canvas. This is your dynamic editing workspace. On the center of the screen, you have your document editor. On the left, you'll see your saved sources, including charts and the map we pinned earlier. And on the right of the screen you'll find your powerful AI assistant which has access to all the canvas context plus all the Banana Analytics platform data in a retrieval-augmented generation patterned system. With a simple click, we can insert that map right into the executive summary to provide immediate visual context for our readers.
As we review the generated text, we might see a key finding about elevated mental distress across the cohort. Let's ask our built-in AI Assistant to help us draft actionable strategies to address this specific need. We'll ask the assistant to draft three concrete ways to tackle this cohort-wide mental health challenge, tailored to fit right into this section of our document.
The assistant understands the context of our report and provides three targeted strategies: telehealth expansion, behavioral health integration in primary care, and community health worker programs. We can easily copy this generated text, paste it into our document, and format it with headers to match our exact style.
Once everything looks perfect, we simply export the Canvas. And just like that, we have a polished, professional, data-driven health assessment ready for stakeholders. Thanks for watching, and welcome to the future of community and environmental health data intelligence.
Try it on your own service area
Free single-county profiles are available without a credit card. Hospital service-area builds with multi-county scoring, AI summary generation, and Canvas access are part of the Consultant Studio tier; pro-bono licenses are available for community health centers that can't afford the subscription. Tell us what you're working on and we'll figure something out.