Create a Single Operational Dashboard
Open the dashboard you use now. Pick any tile and ask what you would do differently if the number doubled tomorrow. If the honest answer is "nothing", that tile is decoration. Most business dashboards are crowded with decoration, built by adding whatever the software could chart, and the result is a screen that everyone glances at and nobody trusts enough to act on.
This article is for the general manager of a service business, whether that is a clinic, a field-service company, a property office or an agency. You run people, schedules and cash at the same time. The aim is a single page that tells you on Monday what needs your attention this week, and that you can still believe on Friday.
Start with decisions, not data
The most reliable way to design a dashboard is to list the decisions you make repeatedly and work backward.
Take a notepad and finish this sentence ten times. "Each week I decide whether to ___." Typical answers for a service business include whether to add or move staff to cover next week's schedule, whether to chase a late payer, whether to pause new bookings, whether to reorder supplies, whether to call a customer before they call you, and whether cash will cover payroll in three weeks.
For each decision write down three things.
- The question you need answered, in plain words.
- The number or short list that answers it.
- The threshold at which you would act.
Now you have a specification. A metric without a decision and a threshold does not make the page. That rule alone will cut most dashboards in half.
Leading and lagging indicators
A lagging indicator tells you what already happened. Revenue last month, completed jobs, and complaints received are lagging. They are reliable and they are late. A leading indicator moves before the result does, and it gives you time to respond. Booked appointments for next week, quotes sent but not yet answered, open overdue tickets, and technician hours scheduled against hours available are leading.
You need both, for different reasons. Lagging figures tell you whether your past decisions worked. Leading figures tell you what to do now. A page with only lagging numbers is a rear-view mirror. A page with only leading numbers has nothing to check against.
A simple pairing helps. For each important outcome, put the lagging result next to the leading signal that usually precedes it.
| Outcome you care about | Lagging indicator | Leading indicator |
|---|---|---|
| Revenue | Invoiced revenue this month | Booked and confirmed work for the next 14 days |
| Cash | Cash balance today | Invoices over 30 days old and payroll due date |
| Customer satisfaction | Complaints and reviews received | Jobs running past their promised date |
| Capacity | Utilization last week | Scheduled hours versus available hours next week |
| Sales | Contracts signed | Quotes outstanding and age of oldest quote |
A caution about leading indicators. They are only useful if you have checked that they actually precede the outcome in your business. Keep a note of why you believe each one, and revisit it after a few months of data.
One page means a limit
A single operational dashboard is a discipline of exclusion. Set a hard cap. Six to nine numbers on the first screen is a good range for a manager. More than that and your eye stops reading and starts scanning.
A layout that works for many service businesses has three bands.
- Top band. Three to four headline numbers tied to your most important decisions, each showing its value, its direction against last week, and a simple color for whether it is inside your acceptable range.
- Middle band. The drivers behind those numbers, such as schedule load by day or the oldest overdue invoices, as small tables or simple bar charts.
- Bottom band. A short list of items that need a person today. A list of named exceptions beats any chart.
Drill-down belongs behind the page, not on it. Clicking a headline number should lead to the records that make it up, so that when a figure looks wrong you can see the rows within a minute.
What to leave off
The most useful design decision is the list of things you refuse to show. Strong candidates for exclusion include the following.
- Vanity totals. Cumulative customers served since opening, total website visits, and similar numbers that only go up.
- Anything you cannot act on. If no one in the company could change it, it is trivia.
- Metrics with no agreed definition. If two people would calculate "active customer" differently, hold the number back until they agree.
- Duplicates. The same fact shown as a number, a gauge and a chart.
- Numbers more than one step removed from the work. Averages across the whole company hide the one location or team that is struggling.
- Real-time everything. A live ticker for a figure you review once a day only adds anxiety.
- Sensitive details that do not belong on a shared screen. Individual pay, personal health information, and similar items should live behind access controls, not on a wall display. Rules vary by industry and jurisdiction, so check with the people responsible for compliance.
Keep a parking lot list of rejected metrics with the reason. People will ask why something is missing, and a recorded reason is better than a debate.
Freshness is part of the number
Every figure on the page should say how old it is. A balance that was current at 6 this morning and a booking count that is ten minutes old can sit side by side as long as each is labeled. Without labels, people assume everything is live, and then lose trust the first time a number is stale.
Decide a freshness target per metric, based on the decision it supports.
| Metric | Reasonable refresh | Why |
|---|---|---|
| Today's schedule | Every few minutes | Changes through the day and drives dispatch |
| Overdue invoices | Daily | Reviewed once a day |
| Cash and payroll runway | Daily | Bank data usually posts in daily batches |
| Monthly revenue | Daily or weekly | Only needs to be right when you review it |
Three failure modes cause most freshness trouble.
- Silent failure. The refresh job stops and the page keeps showing yesterday's values with no warning. Show a visible timestamp, and make the whole tile turn gray if data is older than its target.
- Time zones and cutoffs. A day that ends at midnight UTC is not the same as a business day in Puerto Rico. Choose one time zone for reporting and say so on the page.
- Late-arriving data. Invoices, timesheets and payments get entered or posted days after the event. Numbers for last week can legitimately change this week. Mark recent periods as provisional.
Ownership, or the dashboard will rot
Every number on the page needs a named owner. Not a department, a person. The owner is responsible for three things.
- The definition. What is counted, what is excluded, and which system is the source of record.
- The quality. Noticing when it looks wrong and fixing the source, not just the chart.
- The action. Being the one who responds when it crosses its threshold.
Write these into a short specification for each metric. One row per metric is enough.
| Field | Example for overdue invoices |
|---|---|
| Name | Overdue receivables over 30 days |
| Decision it supports | Whether to call or pause work for a customer |
| Definition | Sum of unpaid invoice balances more than 30 days past due date, excluding disputed invoices |
| Source of record | The accounting system |
| Refresh | Daily by 7 am Atlantic time |
| Owner | Accounts lead |
| Action threshold | Any customer over 45 days, or total over the agreed limit |
When the number is questioned, the specification settles it. When an owner leaves, the specification keeps the number alive.
A worked example
This is hypothetical, with invented figures, to show the design process.
A general manager runs a field-service company with twelve technicians. In her notepad exercise she lists her weekly decisions. Does next week's schedule have enough hours for the work booked? Which customers are overdue and by how much? Are any jobs at risk of missing their promised date? Is cash on hand enough to cover payroll in two weeks?
She builds a page with seven numbers.
- Scheduled technician hours next week versus available hours, with a flag if scheduled exceeds 90 percent of available.
- Jobs booked for the next 14 days, compared with the same period last year.
- Jobs running past their promised date, as a count with a list.
- Total of invoices over 30 days old, with the three largest customers named.
- Cash balance and days until the next payroll.
- Quotes sent and unanswered for more than a week.
- Complaints received this week, as a count with the text of each.
She leaves off total jobs completed since the company opened, a monthly average rating across all customers, a social media follower count, and a chart of fuel prices. She records the reasons in a parking lot list.
In the first month she notices that her scheduled hours figure looks too high on Mondays. The cause is that jobs rescheduled on Sunday are still counted in their old slot until a nightly sync runs. The fix is a short refresh during working hours, and a note on the tile showing the last update time. The lesson is that a dashboard teaches you about your data as much as your business, which is why ownership and labels matter.
Think also about the island context. During hurricane season a dashboard that depends on one office network is a risk. Make sure it can be viewed from a phone, and decide ahead of time which three numbers you want on a day when power or connectivity is limited, such as jobs at risk, staff available, and cash.
Building it without overbuilding
You do not need a data warehouse to start. For a business of this size, a reasonable path has these stages.
- Paper first. Draw the page by hand and review it with the people who own each number.
- Spreadsheet or the tools you already pay for. Many accounting, scheduling and CRM products have built-in dashboards. Try those before buying anything.
- Pull from the source of record. Connect directly where you can and avoid copying numbers by hand, since each copy is a chance for drift.
- Reconcile. For each metric, compare the dashboard value to the source system on three different days and write down any difference.
- Add a refresh and a stale-data warning.
- Only then consider a dedicated tool or a database. If you outgrow the first approach, you will know exactly what you need.
Questions for a vendor
- Can I see when each tile was last updated, and can I set an alert if data goes stale?
- Can I define a metric once and reuse it, or is every chart its own calculation?
- How are users and permissions managed, and can I restrict sensitive tiles?
- Can I export my metric definitions and data if I leave?
- What happens to price and speed as my data grows?
What to do on Monday morning
- List ten recurring decisions you make.
- For each, write the question, the number, and the threshold.
- Cross out everything that lacks a decision or a definition.
- Pair each outcome with a leading and a lagging indicator.
- Name an owner for every remaining number.
- Sketch the page on paper and keep it to nine numbers or fewer.
If you want help turning that sketch into something that refreshes reliably, Ozcorp InfraOps works with small teams on exactly this, and I am happy to review your metric list with you whether or not you ever hire anyone.
Osvaldo D. Rivera, PhD
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.