Brainforge Knowledge

training

10 min read

Eden Omni Stakeholder Training

Audience: mixed (stakeholders + a few builders)
Format: ~half overview / ~half hands-on
Duration: 45 minutes
Goal: enable Eden stakeholders to confidently consume dashboards, ask AI questions safely, and turn “daily snapshots” into an operational workflow; enable builders to create a basic chart and understand Topics.

Run of Show

Pre-reqs (do these before the session)

  • Access:
    • Ensure attendees can log into Omni.
    • Ensure at least 2 builders have Standard (or equivalent “can create charts/dashboards”) access.
    • Ensure the facilitator has the ability to create a chart, save to dashboard, and create a delivery (or at least show where it lives).
  • Content:
    • Confirm at least 1–2 “daily driver” dashboards exist (examples from Tableau era: Unit Economics/Profitability, Order Journey, Retention/Marketing).
    • Confirm at least 1 Topic exists and is usable for charts and AI.
  • Artifacts to have open:
    • notes.md (invite list + links)
    • omni_pricing_internal_bi.pdf (seat types + what viewers can do + tokens)

Key messages to reinforce (Eden-specific)

  • Topics ≈ curated datasets (joins + definitions + governance). A dashboard can combine charts from multiple Topics; each chart points to one Topic.
  • Definitions should stay consistent across Topics (avoid “revenue means X here, Y there”).
  • AI is an entitlement + a budget (tokens); not everyone needs AI-enabled seats.
  • “Snapshots → workflow”: scheduled deliveries/alerts replace ad-hoc screenshots and reduce seat pressure.

Minute-by-minute agenda

0:00–0:03 — Welcome + outcomes

  • What we’re doing today:
    • Understand users/roles (and why it affects cost).
    • Navigate familiar dashboards and export data.
    • Create one simple chart (builders) and understand Topics.
    • Use AI safely for questions/summaries.
    • Turn daily “snapshots” into a workflow (deliveries/alerts).

0:03–0:10 — Users + licensing (conceptual, no dollar amounts)

Use the “three seats” story from omni_pricing_internal_bi.pdf:

  • Developer: model + connections + permissions (small group).
  • Standard: create dashboards/charts + AI context + AI usage (builders).
  • Viewer: consume dashboards; can filter/drill/download and schedule deliveries + set alerts (most stakeholders).

Talking points:

  • Eden does not need “everyone” as a builder.
  • Start lean on Standard seats; expand only if there’s a real self-serve need.
  • AI: give to the people who will actually use it, and who understand metric intent.

0:10–0:14 — Pricing (how to think about it)

Keep it high-level:

  • Seat mix is the main lever: fewer builders, more viewers.
  • Use “deliveries” for broad distribution (daily Slack/email) instead of “seat everyone.”
  • AI usage uses tokens; treat tokens like a shared monthly budget.

0:14–0:22 — Familiar dashboards: find, filter, export

Live demo (facilitator):

  • Find a “daily driver” dashboard (examples to look for):
    • Unit Economics / Profitability
    • Order Journey / SLA / Pharmacy turnaround
    • Retention / Marketing cohorts
  • Demo:
    • Apply filters (date/product/pharmacy/plan/etc.)
    • Show drill-down behavior
    • Download/export a table to CSV (or show export path)

What to emphasize:

  • The fastest way to get value is not building new dashboards—it’s using filters + exports on existing ones.

0:22–0:30 — Topics (semantic layer) and why stakeholders should care

Explain without jargon:

  • Topic = the “approved dataset” for a set of questions.
  • It’s where:
    • joins live (what tables connect)
    • definitions live (what a field means, grain, assumptions)
    • governance lives (who can see what)

Eden-specific guidance:

  • Topics may be organized by function (marketing/finance/ops) or by use-case, but:
    • core definitions should stay consistent across Topics
    • don’t fork definitions by department

0:30–0:35 — AI features: what it’s for and how to use it safely

Demo (facilitator):

  • Ask 1–2 natural language questions against the relevant Topic/dashboard.
  • Use AI summary (if enabled) and sanity-check results.

Guardrails:

  • AI is great for exploration and first drafts.
  • Always confirm:
    • time windows (e.g., does “today” include partial data?)
    • grain (order vs transaction vs customer)
    • filters (paid vs pending vs cancelled)

0:35–0:45 — Hands-on block (split path: Viewer vs Builder)

Goal: everyone completes a “confidence loop.”

Step A (Everyone — 4 min):

  • Open the chosen dashboard.
  • Change 2 filters (date + one business dimension).
  • Export a table (or download data).

Step B (Viewers — 3 min):

  • Find “Deliveries/Email/Slack” (or equivalent) from the dashboard.
  • Draft a daily delivery that sends a PDF/CSV to an email/Slack channel.
    • If you can’t create it (permissions), identify:
      • where it would live
      • what recipients + schedule should be

Step C (Builders — 3 min):

  • Create a basic chart off the Topic:
    • pick a measure (e.g., revenue, orders, SLA %)
    • group by one dimension (e.g., product, pharmacy)
    • save it to a “Training” dashboard

Wrap (final minute):

  • Ask: “What’s one report you currently screenshot daily?” → convert that into a delivery/alert workflow as the follow-up action.

Suggested Q&A prompts (to pull from stakeholders)

  • “What dashboard do you use every day (or wish you had)?”
  • “What decision do you make from it?”
  • “What’s the ‘snapshot’ that currently gets pasted into Slack?”
  • “If we could alert you when X crosses a threshold, what should X be?”
  • Create/confirm:
    • 1 “Exec Daily” dashboard delivery (PDF) to a shared channel/email
    • 1 operational alert (e.g., SLA % < target)
  • Identify the top 3 Topics needed for P0 dashboards and AI usefulness.

Hands-On Worksheet

This is designed for a mixed room. Everyone completes Part 1. Viewers do Part 2A. Builders do Part 2B.

What you need before starting

  • Omni login
  • A dashboard you can access (ask facilitator which one to use)
  • If you are a builder: “Standard” access (can create charts/dashboards)

Part 1 (Everyone) — Navigate, filter, export (8–10 min)

1) Open a “daily driver” dashboard

Pick one:

  • Unit Economics / Profitability
  • Order Journey / SLA / pharmacy turnaround
  • Retention / marketing cohorts

If you don’t see those, pick any dashboard that looks like a daily KPI view.

2) Change filters (do both)

  • Change the date range
  • Change one other filter:
    • product
    • pharmacy
    • membership plan
    • cohort / channel

3) Answer one question using the dashboard

Pick one:

  • “What changed vs last week?”
  • “Which product (or pharmacy) is driving the change?”
  • “Is this a data change or a filter/definition issue?”

Write down your answer (1–2 sentences).

4) Export data (CSV)

Find an export/download action and export the relevant table to CSV.

If you can’t export:

  • note what you clicked
  • ask the facilitator if export is permissioned or tile-specific

Part 2A (Viewers) — Turn a “snapshot” into a workflow (8–10 min)

Goal: replace a screenshot or manual copy/paste with a scheduled delivery.

5) Find Deliveries / Scheduled sends / Email / Slack (wording varies)

From the dashboard, look for sharing/sending options.

6) Draft a daily delivery

Set up (or “dry-run” the settings) for:

  • Destination: email or Slack channel
  • Format: PDF for exec snapshots; CSV when someone needs rows
  • Schedule: daily on weekdays (pick a time)
  • Filters: same ones you used in Part 1

Write down:

  • who should receive it
  • what time
  • what action it should trigger (e.g., “Ops reviews orders nearing SLA”)

7) Optional: draft one alert

Pick one threshold:

  • SLA % below target
  • orders nearing SLA above threshold
  • revenue dips below expected range

If you can’t create alerts (permissions), just write down the threshold and owner.

Part 2B (Builders) — Create a simple chart + save it (8–10 min)

Goal: build one chart tied to a Topic and save it to a dashboard.

5) Identify the Topic / dataset powering your analysis

Look for a label that indicates the dataset/Topic used for the chart/table.

6) Create a chart

Build something simple:

  • Metric: revenue / orders / SLA % / retention %
  • Group by: product OR pharmacy OR membership plan
  • Time: trend by day or week

7) Sanity checks (do at least two)

  • Does the timeframe include partial days?
  • Does the grain make sense (order vs transaction vs customer)?
  • Are cancelled/archived states excluded or included intentionally?

8) Save it

  • Save chart to a “Training” dashboard (or a sandbox dashboard)
  • Name it clearly:
    • [Training] <metric> by <dimension> (last 30 days)

Troubleshooting quick hits

  • “I can’t see a dashboard” → you likely don’t have a Viewer seat or you’re not in the right group.
  • “I can’t create charts” → you likely need Standard access.
  • “AI answers seem wrong” → check time window, filters, and Topic definitions; escalate if the definition is unclear.
  • “Exports/deliveries missing” → permissioned feature or tile type; ask facilitator to confirm entitlements.

FAQ

Do we need to buy seats for everyone?

Not necessarily. Most stakeholders should be Viewers (consume dashboards, export, schedule deliveries, alerts). Keep Standard seats limited to the people who truly need to build charts/dashboards. Keep Developer seats very limited.

Why does role/seat choice matter?

Because capabilities differ:

  • Viewer is for consumption + operational workflows (export/schedule/alerts).
  • Standard is for building + AI.
  • Developer is for modeling, connections, and permissions.

Who should get AI features?

Start with:

  • builders/analysts who create charts and understand metric intent
  • a small number of power stakeholders who will actually use AI weekly

Then expand if usage proves valuable. AI usage consumes tokens from a monthly budget.

What are “tokens”?

Tokens are standardized units of AI consumption (e.g., natural-language questions, AI summaries). Tokens are budgeted monthly and do not roll over.

What is a “Topic”?

A Topic is a curated dataset (joins + definitions + governance) that powers charts and AI understanding. If a question fails or results look wrong, it often means the Topic needs clearer definitions or corrected joins.

Can different teams have different “revenue” definitions in different Topics?

Avoid this. Keep core definitions consistent across Topics. It’s okay to have different views or starting points by team, but not conflicting definitions.

How do we keep “familiar dashboards” while moving off Tableau?

Prioritize the “daily driver” dashboards first (P0), validate they match stakeholder expectations, then iterate. Don’t rebuild everything—start with what’s used daily.

How do we create charts without making a mess?

Use:

  • a dedicated “Training” or “Sandbox” dashboard for experimentation
  • consistent naming conventions
  • governance via Topics (approved datasets)

What does “turning snapshots into a workflow” mean?

Instead of screenshots pasted into Slack, you create:

  • scheduled deliveries (PDF/CSV) to email or Slack at set times
  • alerts that trigger when thresholds are crossed

This improves reliability and can reduce seat needs.

Can viewers schedule deliveries and alerts?

Yes—Viewer seats are intended to explore dashboards and can typically download/schedule/set alerts (subject to admin settings).

Cheat Sheet

Who is this for?

  • Viewers: consume dashboards, filter/drill, export, schedule, alerts
  • Builders (Standard): create charts/dashboards, use AI tools, edit context
  • Developers: model/semantic layer, connections, permissions

1) Find and use dashboards (the daily flow)

  • Start with the dashboard you use daily (Unit Economics, Order Journey/SLA, Retention).
  • Change filters first (date + one business dimension).
  • Trust, but verify: check time window, filters, and definitions.

Fast checks when numbers look “off”

  • Time window: are you including today (partial data)?
  • Grain: is this per-order, per-transaction, or per-customer?
  • Status filters: are cancelled/archived included?

2) Create charts (builders)

  • Pick a Topic (dataset), then:
    • select a metric
    • group by one dimension
    • sanity-check filters + grain
    • save to a dashboard

Naming pattern:

  • [Team] Metric by Dimension (time window)

3) Topics (semantic layer) — the mental model

  • Topic = approved dataset for a set of questions
  • Topics contain:
    • joins (what tables connect)
    • definitions and assumptions
    • governance (who can see what)

Important rule:

  • Core definitions must remain consistent across Topics (don’t fork “revenue”).

4) AI features (when enabled)

Good for:

  • exploring trends
  • summarizing “what changed”
  • drafting a chart/question

Always validate:

  • timeframe
  • grain
  • filters

AI “fuel” concept:

  • usage consumes tokens from a monthly budget (does not roll over).

5) Mobile view

Use mobile for:

  • quick KPI checks
  • on-the-go decisions

Expect:

  • a mobile layout may differ from desktop; ensure key tiles are readable.

6) Turn snapshots into a workflow (the big win)

Replace:

  • screenshots + manual Slack posts

With:

  • scheduled deliveries (PDF for exec snapshots, CSV for row-level ops)
  • alerts when thresholds are crossed

Recipe:

  • Decide owner → decide cadence → decide destination → decide action

Graph View