Program dashboard · The slip you see too late

Build a program dashboard that catches drop-off mid-cycle.

Sopact reads every attendance record, survey, and mid-cycle reflection the moment it arrives — and joins it to the participant behind the tile, so the dashboard shows who is slipping while the cohort is still running. A dashboard that refreshes nightly surfaces Tuesday's problem on Wednesday — after a participant has missed two more sessions. This page is the step-by-step method, for the program managers and directors who run the program from the dashboard, not from a quarterly report.

By Unmesh Sheth · Founder & CEO, Sopact · Updated May 25, 2026

Bring one dashboard you report on See: Program evaluation

7 dashboards Worked examples, source to report

Qual + quant On one participant record

Read on arrival Not a nightly batch

2014 Building for impact data since

Definition

What is a program dashboard?

Plain definition

A program dashboard is the live view of whether one program is producing the change it was designed to produce. It carries three layers — operational health on top, outcomes underneath, the AI-themed reason from open-ended reflections on the side. The test that matters: every tile drills to the participant behind it, and the screen reflects current state — not a snapshot from last quarter.

Level 1 · A tile

"71% passed"

A number with no participant behind it. You cannot open it to see who the other 29% are, or follow up.

Level 2 · A dashboard

"Tiles, refreshed nightly"

Numbers in motion. But Tuesday's drop-off surfaces Wednesday, and the reason sits unread in a free-text field.

Level 3 · A program dashboard you run from

"Operational health, outcomes, and the AI-themed reason — every tile drilling to the participant, live."

The team runs the program from this one.

The gap

Why most program dashboards fail

Most program dashboards fail the same way: they become a reporting artifact nobody opens, because the action happens somewhere else — in a spreadsheet, in Slack, in the program manager's head. Four failure modes account for nearly all of it.

Failure 1

Tiles with no participant behind them

A tile shows 71%. You cannot open it to see who the other 29% are, which session they last attended, or who to call this afternoon.

Failure 2

Nightly batch, mid-cycle blindness

The dashboard refreshes overnight. Tuesday's barrier surfaces Wednesday morning — after a participant has missed two more sessions.

Failure 3

The reason sits unread

Mid-cycle reflections hold the why behind every number. With no AI layer they are a manual quote-pull — so the dashboard shows the number and never the reason.

Failure 4

A separate BI build every cycle

The dashboard is rebuilt from exports each reporting cycle. It lags the program by weeks, so the team falls back to the spreadsheet and the dashboard goes unopened.

Bottom line

A program dashboard fails when a tile cannot be opened to the participant behind it, and when it reports on a schedule instead of reading data as it arrives. Both are upstream of the chart — no visualization tool fixes either.

The approach

Immediate, continuous, and learning — not a nightly batch

The fix is not a prettier tile. It is a change in when the dashboard reads its data, and what it does with it once it has. Sopact builds program dashboards on three principles.

Principle 1 · Immediate

Read on arrival

An attendance dip and the mid-cycle reflection that explains it are themed, scored, and joined to the participant record the morning they happen — not held for a nightly batch. The signal and its reason land together.

Principle 2 · Continuous

One participant, one record

Every participant keeps one Persistent Contact ID across intake, mid-cycle, and exit. Every tile drills to the same record — so an outcome is a trajectory, and a number is always one click from the person behind it.

Principle 3 · Learning

The dashboard you run from

Operational health on top, outcomes underneath, the AI-themed reason on the side — with alerts that fire on attendance and dosage thresholds. The dashboard becomes the operating system for delivery, not a reporting artifact.

Why it matters

A nightly batch tells the program manager what happened yesterday. An immediate, continuous, learning dashboard flags the three participants who missed last session before the next one starts — while there is still time to reschedule.

The data layer

A dashboard is only as reliable as the data underneath it

Before any tile, two questions decide whether a program dashboard can be run from: where the data comes from, and whether the system knows what each field means. This is the layer Sopact owns — sources on the left, a live dashboard on the right.

Step 01 · Sources

Where the data comes from

Primary — you collect it

Intake form Attendance / session log Pre / mid / post survey Mid-cycle reflection

Secondary — systems you run

LMS / training system Attendance system Prior-cohort data

→

Step 02 · The join

Persistent Contact ID + data dictionary

One record per participant Qual + quant on one row Read on arrival

The data dictionary defines what counts as at risk and sets the alert thresholds. Every tile drills back to the participant record — governed, not guessed.

→

Step 03 · Output

A live program dashboard

Operational health Outcomes AI insight Funder view

Every tile opens back to the participant record it came from — traceable to source.

Primary data — collected directly in Sopact Sense Secondary data — integrated from systems you already run

Approach A

The primary-data approach

A program dashboard is mostly a live view of the program's own delivery and outcomes. Sopact Sense collects intake, attendance, surveys, and mid-cycle reflections clean at source — one record per participant, qualitative and quantitative on the same row. Lead with primary data: it is what makes every tile drill to a participant, and every alert fire on time.

Approach B

Integrating primary + secondary

A learning management system or attendance system may already hold session data, and a funder may require a set framework. Integrate secondary data when those systems hold what you would otherwise re-collect. The data dictionary maps each field to the participant record, so the live view reads one dataset.

The proprietary layer

Sopact's layer is the combination — qualitative data, quantitative data, and the data dictionary that defines at-risk thresholds and governs the join. It is what lets a tile open to the participant behind it — and an alert fire the morning a participant slips, not at cohort close.

The method

How to build a program dashboard, step by step

Here is the build, in the order Sopact runs it — six steps from the program manager's weekly decisions to a dashboard that refreshes itself and fires alerts on its own.

1

Name the decisions the program manager makes each week

Start from the standup: who to follow up with, which session to fix, which site is slipping. "Show me who missed last session" beats "build a program dashboard." The weekly decisions decide which tiles the dashboard needs and which alerts it fires.

2

Write the logic model and data dictionary first

Turn the program theory into a logic model, then define every field — what counts as at risk, the attendance and dosage thresholds, the survey scale. Both are signed before collection starts. The thresholds are what let the dashboard alert instead of just display.

3

Collect primary data clean at source

Run intake, attendance, surveys, and mid-cycle reflections through Sopact Sense. Each participant gets one Persistent Contact ID at intake; every later touchpoint links to it; duplicates and typos are caught in the form, not at cohort close.

4

Integrate the systems that already hold session data

Connect a learning management or attendance system, and prior-cohort data, through the data dictionary. Each field maps to the participant record, so the dashboard reads one dataset — no nightly export, no re-keying.

5

Read on arrival, then build the layered view

Sopact reads every response and reflection the moment it lands — theming the narratives, scoring outcomes, firing alerts on the thresholds. The view is then assembled in plain language — operational on top, outcomes underneath. This is the step an AI build tool finishes in minutes.

6

Set it live — one source, three roles

The dashboard updates as records update. Role-based views give program staff participant detail, leadership the cross-cohort trend, funders the reporting layer — all filtered from one source, no separate build per audience.

Time

The nightly-batch lag is gone — the dashboard is live, and the funder view is a filter, not a build.

Money

No separate BI build per reporting cycle — staff hours move from rebuilding the report to running the program.

Risk

The at-risk participant is flagged while there is still time to act — not named in the post-mortem.

The output

What a live program dashboard looks like

The method produces a dashboard the team runs the program from — operational health on top, outcomes underneath, the AI-themed reason on the side. Below is a live view for a sample workforce program. Sample data, illustrative.

Program dashboard · live

Live program dashboard

Sample workforce program · Cohort 03 · 109 participants · 3 sites

Operational health — today

109

Participants enrolled, Cohort 03 active

Source: intake record

87%

On the attendance plan, cohort-wide

Source: session log

7

At-risk flags open, 3 raised this week

Source: alert thresholds

Weekly attendance, Cohort 03 — the dip is mid-cycle

Week 2

96%

Week 4

91%

Week 6

84%

Week 8

79%

Outcomes by cohort — Cohort 03 still in progress

Cohort	Enrolled	Completion	Credential pass	90-day placement
Cohort 01	102	79%	64%	58%
Cohort 02	115	84%	71%	64%
Cohort 03	109	active	—	—

The AI-themed reason, and the alerts

What mid-cycle reflections say

Childcare: the top barrier in Cohort 03, up from transportation in Cohort 02 — cited 18 times in 14 days.
Shift change: "shift change at the warehouse" appeared 3 times in 48 hours, all from one employer cohort.
Peer support: named most often as the reason participants stay enrolled.

Alerts fired this week

3 participants, same employer cohort, missed last session — flagged Monday 8:30 a.m.
Site B attendance below the 80% dosage threshold at week 8.
1 participant has no mid-cycle reflection on file — follow-up due.

Sample data, illustrative · every tile drills to a participant record under one Persistent Contact ID

Read it together

Week 8 attendance at 79% and "childcare" rising in the reflections are not two findings. They are one finding — the number and its reason — on one screen. The alert fired Monday. The cohort is still running, so the team can still act.

The examples

Seven program dashboards, and the data behind each

Seven views cover most of what a program manager needs to run one program. Each is a filtered view of the same participant record — and each names its sources and the risk it is built to catch.

Primary — collected directly Secondary — integrated from a system you run

1

Enrollment & attendance dashboard

Intake form Session log

Surfaces

Enrollment by cohort, attendance and dosage by week and site.

Risk caught

A dosage dip that becomes a completion problem.

2

Participant progress & at-risk dashboard

Attendance Assessments Reflections

Surfaces

Each participant's trajectory, and who is slipping right now — with the alert that fired.

Risk caught

A participant who drops while the tile still shows green.

3

Outcome dashboard (pre/post)

Pre survey Post survey

Surfaces

Skill or behavior change, baseline to exit, linked by one participant ID.

Risk caught

A cohort average that hides who did not change.

4

Engagement & feedback dashboard

Mid-cycle reflection Open-ended feedback

Surfaces

AI-themed barriers and supports — what participants name, while the cohort is still running.

Risk caught

The reason behind a number sitting unread in free-text.

5

Milestone & delivery dashboard

Program plan Delivery log

Surfaces

Sessions delivered against planned, milestones hit, fidelity by site.

Risk caught

Delivery drift no one sees until the report.

6

Cohort comparison dashboard

C01

64%

C02

71%

C03

73%

Sources

Longitudinal records across program cycles (primary), one persistent ID.

Surfaces

How each cohort performs against the last — before and after a program change.

Risk caught

Cohorts that cannot be compared because the instrument changed.

7

Program health summary

Operational Outcomes AI insight

Surfaces

One screen for leadership and funders — is the program on plan, on outcome, with the reason.

Risk caught

A leadership view that is a stale quarterly slide.

The build tools

Build the view with the AI tools you already have

The dashboard view itself — the tiles, the layout, the layered layout — is no longer the hard part. Claude, Google's analytics stack, Microsoft Power BI, Tableau, and Looker all turn clean, well-defined data into a working dashboard in an afternoon.

So the value is not in the tile-building. Tableau, Power BI, and Looker are strong visualization tools, but they sit downstream of a data store that has to be populated first — they render a dashboard once the work is done; they do not produce the participant record. Point an AI build tool at a monthly spreadsheet export and it builds a fast, confident, stale dashboard. Point the same tool at the layer Sopact maintains — the live participant record, qualitative and quantitative on one row, the at-risk thresholds defined — and it builds a dashboard the team runs the program from.

What AI build tools do well

Build the dashboard view fast — tiles, layout, layered layout in an afternoon.
Write the narrative summary that sits above the numbers.
Re-cut a view for the program, leadership, or funder role on request.
Handle the analysis once the data is clean and well defined.

What they cannot do for you

Produce the live participant record every tile drills back to.
Hold one participant identity across intake, mid-cycle, and exit.
Theme the mid-cycle reflections as they arrive.
Fire the alert the morning a participant slips.

The analysis got easy. The participant record did not. That is the layer to own.

Program dashboard vs the rest

Four things get called a program dashboard

A monthly spreadsheet is a snapshot. A PMO dashboard — the project-management sense of "program management dashboard" — tracks tasks, milestones, and budget, not participant change. A BI dashboard renders whatever it is handed. A working program dashboard, in the impact sense, runs one program live off the participant record.

Capability	Spreadsheet (monthly)	PMO dashboard (Asana, Smartsheet)	BI dashboard (Power BI, Tableau)	Sopact
Tracks participant change, not tasks	Partial	No — tracks tasks and budget	Depends on the source	Yes — skills, behavior, outcomes
Live refresh	No — monthly	Yes — for task status	Partial — needs a pipeline	Yes — reads on arrival
Every tile drills to the participant	No	No — drills to a task	Partial	Yes
Reads qualitative reflections	No	No	No — quantitative only	Yes — themed on arrival
Operational + outcome on one record	No	No	No — separate sources	Yes
Tracks the same participant across cohorts	No — manual matching	No	Partial — if a pipeline exists	Yes — Persistent Contact ID
Alerts fire mid-cycle	No	Yes — on task deadlines	Partial	Yes — on attendance and dosage
Rebuilt every reporting cycle	Yes — by hand	No	Often — pipeline upkeep	No — one live source
Best audience	Program staff, as a fallback	Project managers	Data and IT teams	Program staff, leadership, funders

A PMO dashboard answers "are we on track to ship." A program dashboard answers "are participants changing." Both can be useful — this page is about the second one.

See it on your own program

Bring the spreadsheet you refresh monthly.

We trace each tile to the participant record behind it and rebuild one view live — your program, not a demo account.

Book the working session

FAQ

Program dashboards, answered.

What is a program dashboard?+

A program dashboard is the live, always-on view of whether one program is producing the change it was designed to produce. It draws from the participant record, so every number and narrative on screen reflects current state rather than a snapshot from last quarter. A working program dashboard carries operational health on top, outcomes underneath, and the AI-themed reason from open-ended reflections on the side.

What is the difference between a program dashboard and a program management dashboard?+

A program management dashboard, in the project-portfolio sense, tracks tasks, milestones, budget, and resource allocation — it answers whether the work is on track to ship. A program dashboard, in the impact and evaluation sense, tracks whether participants changed: skills gained, behaviors adopted, conditions improved. The two share a name but answer different questions for different audiences. This guide is about the second one.

How do you build a program dashboard?+

Build a program dashboard in six steps: name the decisions the program manager makes each week, write the logic model and data dictionary, collect primary data clean at source under one participant ID, integrate the systems that already hold session data, read every response on arrival, then assemble the layered view and set it to refresh live. The thresholds that fire alerts are defined before collection starts.

What is a program-level outcomes dashboard?+

A program-level outcomes dashboard is the layer of a program dashboard that surfaces outcome indicators — skills gained, jobs secured, conditions improved — tied to the participant record. Every aggregate can be filtered by cohort, site, or demographic, and opened to the participant-level data that produced it. Without record-level grounding, an outcomes dashboard becomes a static summary nobody can interrogate.

What is a program health dashboard?+

A program health dashboard is the operational layer of a program dashboard: enrollment, attendance, dosage, drop-off, completion, and alerts. It answers the most basic question — is the program delivering on plan — and it is the foundation the outcome and reporting layers sit on. A program with broken delivery cannot produce outcomes, so the health layer catches the delivery problem before it becomes an outcome problem.

What is a program evaluation dashboard?+

A program evaluation dashboard is the live view of the evidence base a program evaluation interprets. The evaluation produces a periodic written judgment; the dashboard exposes the underlying participant record continuously, so the judgment can be checked and updated. Both pull from the same record — the evaluation is the analytical work, the dashboard is the always-on surface.

What are program dashboard examples?+

Seven program dashboard examples cover most of what a program manager needs: an enrollment and attendance dashboard, a participant progress and at-risk dashboard, an outcome dashboard with pre-post comparison, an engagement and feedback dashboard, a milestone and delivery dashboard, a cohort comparison dashboard, and a program health summary. Each is one view of the same program, drawing from one participant record.

Should a program dashboard use primary or secondary data?+

Lead with primary data — intake forms, attendance logs, pre-mid-post surveys, mid-cycle reflections you collect directly — because a program dashboard is mostly a live view of the program's own delivery and outcomes. Integrate secondary data when a learning management system or attendance system already holds session data, or when reporting against a funder's framework. The data dictionary maps the two together.

How do AI dashboards improve program visibility and oversight?+

AI dashboards improve visibility by processing open-ended text — reflections, transcripts, documents — into themes, sentiment, and pattern alerts as it arrives. A traditional dashboard shows that 71 percent passed a credential. An AI dashboard shows, next to that number, that the most-named barrier in mid-cycle reflections shifted from transportation to childcare between cohorts. Oversight scales because the AI layer surfaces what would otherwise sit unread in a thousand free-text fields.

Can you build a program dashboard in Tableau or Power BI?+

Tableau, Power BI, Looker, and Domo are strong visualization tools, but they sit downstream of a data store that must be populated first. For a program dashboard the upstream work — collecting data on a participant record, linking baseline to outcome by stable ID, theming the narratives — is the harder problem. BI tools render the view once that work is done; they do not produce the participant record on their own.

What is the difference between a program dashboard, an evaluation, and a report?+

All three pull from the same participant record. A program evaluation is the periodic analytical work of judging whether the program produced its intended outcomes. A program report is the structured artifact that packages the findings for a specific audience. A program dashboard is the live, continuous view of the same evidence base. Cadence, output, and audience differ; the underlying data is one record per participant.

How does Sopact build a program dashboard?+

Sopact holds the participant record and the dashboard layer in one place — the record intake writes to is the record the dashboard reads from. Ratings, narratives, documents, and transcripts feed both the operational and outcome layers, and the AI layer themes the narratives as they arrive. The dashboard updates as records update, with role-based views for program staff, leadership, and funders from one source.

Related guides

Guides that pair with the program dashboard.

Bring your messiest dashboard

We'll rebuild it as a live view, on screen.

Sixty minutes with someone who builds these for a living. Bring the spreadsheet you refresh monthly, the BI report nobody opens, or the program tile you wish was live. We trace each tile to the participant record behind it, show where primary and secondary data connect through the data dictionary, and rebuild one view live. No slideware, no demo accounts — your program, read live.

Bring a dashboard Watch the 3-minute demo

No slideware. No demo accounts. Your own records, read live.

Format

Working session · 60 min

With

Unmesh Sheth · Founder & CEO

Bring

A spreadsheet, a BI report, or the tile you wish was live

Leave with

One view rebuilt live, and a map of where every tile comes from

Program Dashboard: How to Build One + 7 Examples

Build a program dashboard that catches drop-off mid-cycle.

What is a program dashboard?

"71% passed"

"Tiles, refreshed nightly"

"Operational health, outcomes, and the AI-themed reason — every tile drilling to the participant, live."

Why most program dashboards fail

Tiles with no participant behind them

Nightly batch, mid-cycle blindness

The reason sits unread

A separate BI build every cycle

Immediate, continuous, and learning — not a nightly batch

Read on arrival

One participant, one record

The dashboard you run from

A dashboard is only as reliable as the data underneath it

The primary-data approach

Integrating primary + secondary

How to build a program dashboard, step by step

What a live program dashboard looks like

Seven program dashboards, and the data behind each

Build the view with the AI tools you already have

What AI build tools do well

What they cannot do for you

Four things get called a program dashboard

Program dashboards, answered.

Guides that pair with the program dashboard.

We'll rebuild it as a live view, on screen.

Company

Resources

Agents & Solutions