Dashboard integrity review for BI leaders

Your dashboard output is scaling faster than your review process.

77 Rules helps Heads of BI, VP Analytics, BI CoE leaders, and Heads of Data catch visible integrity risks before dashboards and board-deck charts reach leadership. The review produces ranked findings, evidence tied to the chart, remediation priorities, and a re-audit trail.

Scope, stated up front: 77 Rules evaluates what is visible in the rendered dashboard: presentation, context, comparisons, scale, labeling, disclosure, and related integrity signals. It does not verify underlying data, metric definitions, calculations, joins, or pipeline logic. Inferential concerns are flagged for verification with your data team.

Request a 20-Minute Review Read a Full Audit Output

Bring one real dashboard or board-deck chart. No BI credentials or data connection required.

77 rules plus 15 bonus integrity checks Evidence tied to the chart Ranked remediation queue Re-audit trail

Fit

Built for the BI leader who owns the review standard.

The strongest fit is a Head of BI, VP Analytics, BI CoE leader, or Head of Data responsible for what reaches executives.

You are the right buyer if

You own BI, analytics, or a BI center of excellence, and dashboard authorship is decentralized across more than a handful of people.
AI-assisted chart and dashboard generation has increased output faster than your review capacity.
A chart has reached leadership with missing context, an ambiguous comparison, or another visible issue that changed how it was interpreted.
You need a written standard for what is safe to publish, not a one-off critique.
You support board, QBR, lender, or executive reporting where a misleading presentation can create immediate decision risk.

This is the wrong purchase if

You suspect your numbers are wrong. That is a data quality and lineage problem. 77 Rules reads pixels and will not find it.
You want dashboards redesigned. This produces findings and priorities, not a rebuild.
You have only a few low-stakes dashboards and no recurring review problem. The sample audit may be all you need.
You need automated validation of underlying workbook logic, semantic models, or data pipelines. 77 Rules does not inspect those layers from a screenshot.
Your governance model prohibits third-party processing and none of the available customer-controlled deployment options can meet your security requirements.

Scope

Dashboard integrity is the review layer between data quality and executive interpretation.

77 Rules does not certify that the underlying number is correct. It tests whether the rendered chart gives a competent reader enough visible context to interpret that number safely.

What the review tests

Whether a value carries the context needed to judge it: baseline, target, prior period, trend.
Whether the reporting window, filter state, and comparison basis are disclosed on the chart.
Whether encodings are honest: axis treatment, scale, proportion, dual axes, color meaning.
Whether units, magnitude conventions, and category labels are internally consistent.
Whether visual emphasis matches analytical importance.
Whether structure and layout support a fast read without inviting a wrong one.

What it cannot test

Whether the underlying data is correct, complete, or current.
Whether the metric definition matches the one finance uses.
Whether the calculation, join, or filter logic behind the chart is right.
Whether the right people have access to the right rows.
Anything invisible in the rendered image.

How the review runs: AI vision models plus deterministic rule logic are applied to the screenshot. AI systems can miss issues or raise false positives. Visually confirmable findings are tied to evidence on the chart; inferential concerns are explicitly marked for verification with your data team. Findings are advisory input to a human review, not a certification.

Sample output

A dashboard scored 87% on design and 66% on integrity. Both scores are correct.

This is a real audit of a Marketing Channel Dashboard. Design quality and interpretive safety are separate tests, and the second one is the one that reaches the boardroom.

Integrity66% D

Design87% B

Combined73% C

8 fail7 warning35 pass42 not applicable

77 rules plus 15 bonus integrity checks are applied according to chart context. Checks that do not apply are reported separately rather than counted as passes.

What the numbers mean: the dashboard was well built. Six KPI cards carried no baseline, the reporting window was never stated, and one monetary value used a different scale convention than its neighbors. None of that is a design defect. All of it changes what a reader concludes in eight seconds.

Rule 77 · Integrity fail

Six KPIs with no basis for judgment.

Gross revenue, orders, new customers, margin, ad spend, and acquisition cost are shown as bare absolute values with no target, prior period, or trend. What a reader concludes: that the number is good or bad, based on nothing on the screen.

Fix: add versus-prior, versus-target, or a compact sparkline to each prominent KPI.

Rule 78 · Integrity fail

No stated reporting window.

The filter reads "All Years." The KPI cards and the Revenue by Channel chart never state the period they cover. What a reader concludes: whichever window they assumed when they opened it, which is usually year to date.

Fix: put the explicit window in the title or subtitle.

Rule 86 · Integrity fail

Two magnitude conventions on one screen.

Cost of Acquisition renders as $128.418 while Gross Revenue and Ad Spend render as $163.6M and $29.3M. What a reader concludes: possibly that acquisition cost is $128 million, in the two seconds before someone corrects them out loud.

Fix: one scale convention across every monetary value, or none.

Read the Complete Audit Watch a Full Review (12 min)

The audit engine

Every finding is tied to a pixel you can point at.

Findings that cannot be shown are arguments. Findings that can be shown are decisions. Follow one dashboard from annotated evidence through ranked findings, severity, and category drilldown.

Use the speed control, pause anytime, or move one screen at a time.

Step 1 of 6

Mark the risk on the chart.

Failures and warnings are numbered where they occur, with the ten most consequential ranked alongside the image.

Audit evidenceFailures, warnings, rule markers, and ranked findings are tied directly to the audited dashboard.

Step 2 of 6

Separate design from integrity.

Two scores, deliberately. A dashboard can be beautiful and unsafe, and the combined score hides which one you have.

Audit evidenceSeparate scores expose whether the material weakness is visual quality, analytical integrity, or both.

Step 3 of 6

Find the pattern across a portfolio.

The heatmap is where one dashboard becomes a team problem. Recurring failures in the same rule group tell you what to teach, not just what to fix.

Audit evidenceGrouped checks make recurring weaknesses visible across structure, encoding, labels, layout, context, and filters.

Step 4 of 6

Drill into context and meaning.

The checks that decide whether leadership has enough on screen to interpret the chart correctly, each with its own pass, fail, or warning state.

Audit evidenceContext and meaning checks show individual pass, fail, and warning status with supporting evidence.

Step 5 of 6

Check disclosure and filter integrity.

Unit scaling, category consistency, disclosure, and filter state. The 15 bonus checks live here, and they catch the quiet ones.

Audit evidenceBonus integrity checks isolate unit scaling, category consistency, disclosure, and filter-integrity concerns.

Step 6 of 6

Keep the record, then re-audit.

Original image, annotated evidence, findings, exported report, and every prior run stay attached to the review, so improvement is measurable rather than asserted.

Audit evidenceOriginal and annotated views, prior results, report output, and re-audit history remain connected to the review.

Speed

1 / 6

How the engagement runs

One screenshot decides whether there is anything here worth buying.

The first question is not what to purchase. It is whether the risk is isolated to one chart or systemic across your team.

Send one screenshot

A dashboard already in front of an executive, or a page from the last board deck. No data connection, no tool access.

Twenty-minute review

We walk the ranked findings live, with the evidence marked on your image. You leave with the findings whether or not you buy anything.

Isolated or systemic

Two findings on one chart is a fix you make this afternoon. The same failure across six authors is a standards problem and a different conversation.

Standard, then trail

Where it is systemic, we install a written publishing standard, a prioritized remediation queue, and a re-audit record that proves movement.

Data handling

Where your screenshot goes, in plain language.

Your security reviewer will ask these questions. Here are the answers before they do, including the one you will not like.

The uncomfortable one first.

In the standard hosted configuration, uploaded images are transmitted to commercial production AI APIs for analysis. If that is disqualifying under your governance model, say so in the first email. Enterprise deployments can run on-premise, in your private cloud, or with privately hosted AI models inside your firewall and scope of control. If none of those architectures meet your requirements, we stop before any dashboard data is shared.

What we receive: Rendered images and the context fields you type. No connection to your warehouse, BI server, or credentials.
Processing: AI vision models plus deterministic rule logic. Third-party AI API processing applies in the standard hosted configuration. Enterprise deployments can instead use customer-controlled infrastructure and privately hosted AI models within your firewall or private network, as defined in the written deployment agreement.
Retention: Images, annotations, findings, reports, and re-audit history are retained during the engagement so corrections can be measured against the original.
Deletion: Requested at any time. Active working copies are removed after verification. Rotating backups may retain data up to 90 days. Deletion process.
Your content: You keep ownership. Identifiable dashboards are never used in public examples or marketing without written permission.
Redaction: Redact values before sending if you prefer. The review reads structure, labels, context, and encoding, and works on a redacted image.

Who runs the review

The software creates consistency. The judgment remains human.

77 Rules is founder-led. Serious findings are reviewed in the context of the executive decision they could affect.

Stephen McDaniel

Founder, 77 Rules

Stephen has led analytics, data science, and analytic product work at Tableau, Netflix, SAS, Microsoft, Oracle, and Yahoo. He was the first data scientist at Netflix, later served as a Tableau director, and has taught analytics through INFORMS, TDWI, and the American Marketing Association.

Former Tableau Director Netflix first data scientist 5 years leading a $100M team at SAS Author and analytics faculty

Engagements

Scale the intervention to the size of the problem.

One bad board page is a fix. The same failure across six authors is a standard. Do not buy the second when you need the first.

Most common

Dashboard Integrity Sprint

For the Head of BI, VP Analytics, or CoE lead when the same failures keep appearing across authors.

5 to 10 dashboards reviewed against 77 rules plus 15 bonus integrity checks
Cross-dashboard failure pattern analysis
Prioritized remediation queue with owners
Written publishing standard for your team
Leadership readout and re-audit of corrections

Start With One Screenshot

Team Dashboard Tune-Up

For BI leads who need the team to internalize the standard rather than outsource the review.

3 to 8 high-traffic dashboards audited
Half-day working session with the authors
Review checklist your team runs without us
Before and after examples from your own dashboards

Start With One Screenshot

Board Deck Integrity Audit

For FP&A and board-reporting owners with a dated deadline and a deck that cannot be wrong.

Up to 3 board or CFO-facing chart pages
Findings split into cosmetic and interpretive risk
Written findings memo you can forward
60-minute readout before the meeting

Start With One Screenshot

Enterprise / CoE Engagement

For enterprise analytics leaders who need portfolio coverage, governance controls, and deployment inside their own security boundary.

15 to 40 dashboards across business units
Integrity-risk taxonomy mapped to your rule set
Governance roadmap and remediation standard
Internal audit workflow and team enablement
On-premise or private-cloud deployment
Private hosted AI models inside your firewall

Discuss Deployment

Every engagement starts the same way: one screenshot and twenty minutes. If the finding is isolated, we will tell you to fix it yourself, and that will be the end of it.

FAQ

The questions BI leaders actually ask.

Is this a design review with a governance vocabulary bolted on?

Partly, and the distinction matters. Design review asks whether a chart is well made. This asks whether a competent reader can reach a wrong conclusion from a well-made chart in eight seconds. The rules overlap with design practice because presentation is the mechanism. The findings are framed by the conclusion at risk, not by aesthetic preference.

What does the system add if I already have strong senior analysts?

Consistency and throughput. A strong analyst can catch many individual issues. 77 Rules applies the same review standard repeatedly across a growing dashboard portfolio, ties findings to visible evidence, and gives your team a prioritized review agenda that people can confirm, dismiss, or remediate.

Which BI tools does this work with?

All of them, because it never touches them. The input is a rendered image, so Tableau, Power BI, Looker, Qlik, Sigma, Excel, and a PDF page from last quarter's board deck are all equivalent inputs. That is a genuine limitation and a genuine advantage. It cannot inspect your workbook logic, and it does not need a single integration to review anything you can screenshot.

How many false positives should I expect?

Enough that you will dismiss some findings, which is why every finding carries the visible evidence and a one-click dismissal. Not-applicable results are reported separately from passes so the score is not inflated by rules that never applied. Treat the output as a prioritized review agenda, not a verdict.

What if the review finds nothing material?

Then we say so and there is no engagement. Manufacturing scope on a clean dashboard is a fast way to lose the next four referrals.

Our security team will not approve uploading dashboards to a third party.

Reasonable. In the standard hosted configuration, images are processed through commercial AI APIs. Enterprise deployments can instead run on-premise, inside your private cloud, or with privately hosted AI models inside your firewall and scope of control. Raise the requirement in the first email so we design the data path and model boundary correctly before any dashboard data is shared.

What does the rule set include?

77 rules plus 15 bonus integrity checks. Checks are applied according to chart context, and not-applicable results are reported separately rather than counted as passes.

Your dashboard output is growing. Your review standard should scale with it.

Bring one real dashboard or board-deck chart. We will show you what the review finds and whether the risk looks isolated or systemic.

Request a 20-Minute Review