Inform: what visibility actually means

Outcome

By the end of this lesson, you will be able to distinguish raw visibility from actionable visibility, and design a cost reporting cadence that fits each audience.

New to the cloud? The optional track Cloud, from zero covers the basics first.

1. Concept

FinOps, the practice of managing cloud spending jointly across engineering, finance and the business, runs in three phases defined by the FinOps Foundation (the industry body for this work). Inform is the first of them: making spending visible. It answers one question: who knows what about cloud spending, when, in what form, and with what they can actually do about it.

Most cost programmes stall right here, because seeing the data gets mistaken for being able to act on it. A dashboard, in this lesson, is a screen of charts that people open to see the numbers.

Raw visibility vs. actionable visibility

RAW VISIBILITY                     ACTIONABLE VISIBILITY
────────────────────────────────────────────────────────────────────
A dashboard exists                 The dashboard is opened weekly
The bill is exportable             The bill is parsed by someone who owns the cost
There is a cost report             The report carries a verdict ("we are over budget")
You can see your team's spend      You can compare your team's spend to peers
There is a chat alert              The alert names a specific action and an owner

The test is whether the data leads to a decision.

A dashboard nobody opens is not visibility. A weekly report that gets forwarded but not read is not visibility. A chat channel (in a tool such as Slack) where alerts pile up unread is not visibility either; it is a log.

The three audiences

This phase fails when one report tries to serve three different readers. An engineer, an engineering leader and a finance partner need different detail, at different intervals, in different shapes.

One dashboard for all three produces a tool nobody uses. The engineer needs more detail than a CFO can absorb. The CFO needs more business context than the engineer wants to maintain.

The four cost-data shapes

Cost data comes in four shapes, and each answers a different kind of question.

1. Snapshot. “What did we spend last month?” One number. Good for telling people where things stand, useless for finding out why.

2. Trend. “Are we going up or down?” The same number over time. Good for catching slow drifts, poor at catching a single bad day.

3. Breakdown. “Where is the money?” Stacked bar or Sankey (a flow chart where the width of each band is the amount of money). Useful for understanding composition. Misses outliers within a stack.

4. Anomaly. “What changed unexpectedly?” An alert when spend on one service, team or account jumps away from its normal level. Useful for catching sharp events. Useless unless someone can say what caused it.

A healthy Inform practice produces all four. Most practices produce one or two.

The two questions test

Throughout this curriculum, cadence means how often something happens on a regular rhythm: a weekly review, a daily digest, a check every six hours.

For each cost surface in the organisation, answer two questions:

• Who reads this?
• What decision do they make after reading it?

If the answer to either is unknown, the surface is not yet earning its keep. Either find the audience and the decision, or retire it.

What “accurate” means in the Inform principle

Principle 4 of the Foundation’s six emphasizes cost data that is accessible and timely (its own wording); accuracy is the practical third leg this lesson adds. Accuracy in cloud cost is a band, not a point. Calculated cost is what is running multiplied by the provider’s public price list, called the rack rate: the calculation itself tracks the published prices to within a few percent, but rack rate can sit 20-40 percent or more ABOVE billed cost, because the gap between them is exactly your discount stack. That stack is commitments (promising a year or more of use for a lower price), Spot (cheap spare capacity the provider can take back), Google’s automatic sustained-use discount, and EDP (a negotiated enterprise discount on AWS). Do not read rack rate as within a few percent of what you pay. Daily billing cost, what the provider says it charged each day, is accurate to within ~1 percent of monthly settled. Monthly settled is the closed invoice.

Each band serves a use. “Accurate” means knowing which band you are looking at and labelling it. (See M0.1 L3.)

2. Demo

The same monthly spend question, three audiences, three different surfaces:

Engineering (daily, self-serve):

Dashboard: live cost trend by service, grouped by team. Filter to “my team.” Anomaly markers on the chart. Click a marker → root-cause panel.

Engineering leadership (weekly, one-page):

Email-friendly report: total bill, week-over-week change, top three movers (up and down), open recommendations (suggested fixes) with savings totals.

Finance (monthly, one-page):

PDF or shareable URL: total spend, variance vs. budget, cost per monthly active user (MAU) trend, top three cost centers, forecast for next quarter with confidence band.

Same underlying data. Three rendered surfaces. Three usable decisions.

3. Hands-on (6 min)

For each of the three audiences in your organization, answer:

1. What cost surface does this audience use today?
   Engineering team:     ________________
   Engineering leader:   ________________
   Finance / leadership: ________________

2. How often is it read?
   Engineering team:     ________________
   Engineering leader:   ________________
   Finance / leadership: ________________

3. What decision does each one make from it?
   Engineering team:     ________________
   Engineering leader:   ________________
   Finance / leadership: ________________

Any row answering “rarely” or “no decision” is a surface that is not earning its place. Either change its shape, change how much detail it carries, or shut it down and build the one people would actually use.

Ask for it instead. The same task you just did by hand, asked in one sentence to an AI assistant connected to ZopNight, the cloud cost product taught in later tracks.

BEFORE  A ZopNight account with one cloud connected. Teams set up, or resources carrying an owner tag (a label naming who owns it).
ASK     "Show me last month's spend by team, and by tag."
CHECK   the untagged bucket. That number is the honest measure of how far Inform has got in your org.

Tools behind it (ZopNight’s MCP tools, MCP being the standard way AI assistants call a product’s functions): get_cost_by_team (read, Cost), get_cost_by_tag (read, Cost). Connection details are in the MCP integration docs.

4. Knowledge check

Q1

A monthly cost report is sent as a PDF to engineering teams. Most teams do not open it. Best diagnosis:

A. The shape and cadence are wrong for the audience B. The engineering teams do not care about cost C. The PDF is not visually appealing D. Cost is not the team’s job

Q2

A FinOps team launches a Slack channel for cost alerts. After two weeks the channel has 400 messages and is on mute. Most accurate diagnosis:

A. Slack is the wrong tool B. Alert volume is too high relative to actionability C. The engineers do not check their Slack at all D. The integration is broken

Q3

An engineering leader asks: “Which teams need help?” Which Inform shape best answers this?

A. Breakdown B. Snapshot C. Trend D. Anomaly

5. Apply

ZopNight’s Inform surface has three personas baked in:

• Engineering team → Costs → Reports, Cloud Estate → Inventory, MCP (an AI assistant) for self-serve queries
• Engineering leadership → the Engineering dashboard preset, Weekly summary email
• Finance / leadership → the Finance and Leadership dashboard presets, plus Costs → Unit Economics for cost per unit

The four shapes are all present: Trend (the cost chart on Costs → Reports → Summary), Breakdown (the Costs → Flow Sankey), Snapshot (Dashboards), Anomaly (the anomaly banner on that Summary chart, checked once a day).

Open ZopNight Dashboards (deep link)

Related lessons

• L3: Optimize: the four levers (next)
• T4.M4.7: Dashboards as a governance surface
• T2.M2.10: Cost anomaly detection

Glossary terms touched

FinOps lifecycle