Scheduled Delivery: Getting Reports Where They're Read
A report that is generated but never seen is worth the same as one that was never generated, minus the electricity. That sounds obvious, and yet the most common failure in report automation is not the numbers or the layout. It is delivery. The report lands in a folder nobody visits. An inbox rule files it under "Automated". It arrives as an attachment the reader's phone cannot open. Or, the quiet classic, the job stops running one day and nobody notices for a quarter.
Delivery is a design problem in its own right, separate from what the report says. It has four parts. The first two are the engine that produces the report on schedule and the format the output takes. The other two are the channel that carries it to the reader and the way you know it arrived. Each part has a small number of sensible options. And the right choice depends almost entirely on how the audience reads, not on what is easiest to configure.
This guide walks through all four. By the end you will be able to choose between the monitoring tool's own scheduler and a scripted job. You will be able to pick HTML, PDF, or CSV for the right reasons. You will know how to match each audience to the channel where they will actually open the report. You will know how to name and retain the output so it is useful later. And — the part most people skip — you will be able to make the report job itself something that alerts when it fails.
In this guide
- The two engines
- Output formats and when each fits
- Delivery channels
- Matching channel, format, and audience
- Naming and dating conventions
- Retention: generated reports are evidence
- Monitoring the report job itself
- The scheduling checklist
- Where to go next
The two engines
Something has to wake up on schedule and produce the report. In practice there are two candidates, and most teams end up using both. (The third candidate, a person with a calendar reminder, is the one you are retiring.)
The monitoring tool's own scheduler
Most monitoring tools can generate their built-in reports on a schedule and send them somewhere, usually by email. This is the engine to reach for first, because it needs no code and already has the data. And it will keep working through upgrades that would break a script. Its limits are the flip side of its convenience. You get the report types the tool offers, in the layouts it offers, with the filtering it supports. For availability summaries, top-talker lists, and capacity charts, that is often exactly enough. Check the product documentation for the report types and scheduling options your tool provides, and configure the recurring ones before building anything by hand.
A script on a schedule
For anything the tool cannot format, you write a script and run it from the operating system's scheduler. That might be a report that combines monitoring data with a ticket export. It might be a client page that needs their logo and contract wording, or a one-page management summary with a computed headline. On Windows the scheduler is Task Scheduler; on Linux it is cron. The script pulls the data, computes what it needs, writes the output in the chosen format, and hands it to the delivery channel. It is more work to build and more work to maintain, but it produces precisely the report you designed.
A scheduled entry for a monthly script looks something like this (illustrative; adjust paths and timing for your environment):
# Linux cron: 06:30 on the first day of each month 30 6 1 * * /opt/reports/monthly_management.py >> /var/log/reports/monthly.log 2>&1 # Windows: create a monthly task that runs a PowerShell script schtasks /Create /SC MONTHLY /D 1 /ST 06:30 /TN "Reports\Monthly management" ^ /TR "powershell.exe -NoProfile -File C:\Reports\Monthly-Management.ps1"
The two engines divide cleanly by a simple question: can the tool produce this report as designed? If yes, schedule it there. If not, script it. Do not script a report the tool already offers just because scripting feels more controllable; you will be maintaining that script for years. I wrote one to replace a report the tool already offered, and I am still patching it.
Output formats and when each fits
Three formats cover nearly every case, and each has a natural home.
HTML: for email bodies and web pages
HTML is the format of the email body itself, and of any report that will be read on a screen rather than filed. Its great advantage is that the reader sees the report the moment the message opens. There is no attachment to download and no viewer to launch. And it reads acceptably on a phone if you keep the layout simple. Use it for anything with a short shelf life — the daily ops summary, the weekly status, the management one-pager. Keep it plain: a heading, a few short tables, inline styles rather than anything clever. And use no images that depend on an external server the reader's mail client will block.
PDF: for the record
PDF is the format for anything that must look the same in a year as it does today. Client statements, audit evidence, and anything that will be attached to a contract dispute or a compliance file belong in PDF. That is because the layout is frozen, the content cannot be casually edited, and every viewer renders it identically. PDF's cost is friction — it is an attachment, it is opened deliberately, and it is not read on a phone in a corridor. That friction is fine for a document meant to be filed, and wrong for a document meant to be glanced at.
CSV: for people who will pivot it
CSV (comma-separated values: a plain text table, one row per line) is for readers who want the numbers rather than your presentation of them. Those readers include the capacity planner who will build their own chart and the finance colleague who will join it to cost data. They also include the auditor who wants the raw alert list. Give them CSV and let them work. CSV also serves as the handoff format between your own automation steps. A script that saves a CSV per data source, which a second step turns into HTML or PDF, is easier to debug than one monolith.
Many reports want two formats at once: an HTML body for reading, with a CSV or PDF attached for those who need more. That is a good default for management and client reports, and it costs nothing extra once the numbers are computed.
Delivery channels
The channel is where the report physically arrives, and the rule is to match the channel to how the audience already reads. Do not make anyone go somewhere new for a report; go to where they are.
Email to a list
Email is the universal default and usually correct for management, clients, and anyone outside the technical team. It reaches phones, it works across organizations, and it creates a searchable trail. Its weaknesses are filtering rules that bury automated mail, distribution lists that drift out of date, and attachments that get stripped. Send to a list rather than to individuals, so membership changes do not require editing the job. And put the answer in the subject line where possible.
A shared folder or document system
Saving each report to a shared folder or document library gives you the archive for free. Every period sits next to the last, in order, findable. This is the right primary channel for audit evidence and client statements, and a good secondary channel for everything else. Its weakness is that folders are pull, not push — nobody visits one to see if something new appeared. A shared folder keeps its news to itself. Pair it with a short email or chat notification that says the report is there.
A chat channel post
For the ops team, a post to the team's chat channel is often the best channel of all. It arrives where they already spend the day, it can be discussed in place, and a bad number gets a reply within minutes. Keep chat posts short: the headline and key numbers, with a link to the full report. That is because a wall of tables in chat is scrolled past. Chat is a poor channel for anything senior or external.
An intranet page
A page on the internal web that always shows the latest version of a report is ideal for anything people look up rather than receive. That includes the current availability figures, the capacity trend, and the list of planned maintenance. The page is the reporting cousin of a dashboard. Its weakness is the same as the shared folder: nobody is notified. Use it as the "where to see more" link inside your emailed reports.
Matching channel, format, and audience
Putting the pieces together, here is the combination that fits each audience in most teams. Treat it as a default to adjust, not a rule.
| Audience | Cadence | Primary channel | Format | Secondary |
|---|---|---|---|---|
| Operations | Daily or weekly | Chat channel post | Short HTML summary with a link | Full report on an intranet page |
| Management | Monthly | Email to a list | One-page HTML body | PDF copy in the shared folder |
| Clients | Monthly | Email to the client contact | PDF attachment, short HTML cover note | Copy retained in the client's folder |
| Auditors | Quarterly or on request | Shared folder or document system | PDF report plus CSV of the underlying records | Email notification that the pack is ready |
| Analysts and planners | As the data changes | Shared folder | CSV per data source | None needed |
Notice the pattern: the more senior or external the reader, the more the channel shifts toward email and the format toward PDF. The more hands-on the reader, the more it shifts toward chat and CSV. If you remember nothing else, remember that ops wants to discuss it and management wants to file it.
Naming and dating conventions
A generated report is a file, and files need names that sort, search, and explain themselves a year later. Decide the convention once and build it into the script or scheduler. A pattern that works almost everywhere is audience-report-period, with the period written so that alphabetical order is chronological order:
management-monthly-summary-YYYY-MM.pdf client-acme-service-statement-YYYY-MM.pdf ops-weekly-status-YYYY-Www.html audit-alert-response-evidence-YYYY-Qn.csv
The YYYY-MM form (four-digit year, two-digit month) is the important part. It sorts correctly, it is unambiguous across countries, and it means a folder listing is a timeline. Weekly reports use the ISO week number so that "week 7" does not become "the 12th" in someone else's head. Quarterly reports name the quarter explicitly. Avoid spaces, and avoid "final" and "draft" (a generated report has no drafts). Also avoid the word "latest" in any filename, since it stops being true the moment the next one is generated.
Use the same period label inside the report and in the email subject. "Monthly management summary — YYYY-MM: all commitments met" is a subject line that answers the question before the message is opened. It is also trivially searchable later.
Retention: generated reports are evidence
It is tempting to treat generated reports as disposable, because you can always regenerate them. You often cannot. Monitoring history gets purged as the database grows. A device that was decommissioned no longer appears in a fresh report. The calculation method may have changed since. The report as it was sent is the record of what you told people. That makes it evidence — for client disputes, audits, and your own "when did this start?" investigations.
So keep every generated report in the shared folder, named by the convention above. Keep it for as long as your compliance or contractual obligations require and at least long enough to cover a full annual cycle of questions. Availability reports in particular become part of the record that proves what your uptime actually was. That subject is taken further in the uptime evidence file. Storage is cheap; the argument you cannot settle because the report was overwritten is not.
Two practical notes. Make the archive folder write-once in spirit: the job writes there, people read there, nobody edits there. And keep the CSV alongside the PDF where possible. That way, when someone asks "how was that number calculated?", the answer is a file rather than a reconstruction.
Monitoring the report job itself
Here is the failure that catches everyone once. The job that produces the management report stops. For example, a password expired, a mail relay changed, or the server was rebuilt without the scheduled task. And nothing complains. No error reaches anyone, because the thing that would have reported the error is the thing that stopped. The report simply does not arrive, and since it is one email among hundreds, nobody notices its absence.
At Meridian Parts the monthly availability report had gone to the finance director by scheduled email for two years. Then the mail relay was replaced and the job's outbound settings were not carried across. The job kept running, wrote its PDF to the archive every month, and sent it nowhere. Nine months later the finance director mentioned, in passing, that she had assumed the report was discontinued. Nine PDFs were sitting in the archive folder, correctly named, in order, unread. The team now gets a copy, and an alert when the file does not appear.
A report that silently stops is a small outage, and the fix is the same as for any outage: monitor it. Specifically, alert on the report not arriving, not merely on the job failing. That is because the most common failure is that the job never ran at all. Three approaches, from simplest to most robust:
- Check for the output file. A tiny scheduled check runs an hour after the report is due and alerts if this period's file is missing from the archive folder. This catches "never ran" as well as "ran and crashed."
- Send the job's heartbeat to the monitoring tool. Have the report script's last action be a signal to your monitoring system. That could be a small file it watches, a check it can query, or whatever the tool supports. Add a rule that alerts if the signal is older than the report interval. Silence becomes an alert.
- Watch the delivery end. Include the monitoring team's address on every report list. Treat "did we get it?" as a Monday-morning checklist item until the automated checks are trusted.
A file-presence check, in outline, is short enough to write in an afternoon:
# PowerShell outline: alert if this month's report is missing
$period = Get-Date -Format "yyyy-MM"
$expected = "\\fileserver\reports\management-monthly-summary-$period.pdf"
if (-not (Test-Path $expected)) {
Send-MailMessage -To "ops-alerts@example.internal" -From "reports@example.internal" `
-Subject "MISSING: monthly management report for $period" `
-Body "Expected $expected and did not find it. Check the report job." `
-SmtpServer "mail.example.internal"
}
Also log every run: what started, what it pulled, how many rows it processed, what it sent, and to whom. A run that processed zero rows and sent a beautifully formatted empty report is a subtler failure than a crash. Only the log will reveal it. Add a sanity rule if you can: refuse to send if the row count is far below the previous period's. That way, an empty data pull produces an alert rather than a confident-looking blank.
Tip: test the failure path deliberately. Once the job is working, disable it for a period and confirm that the missing-report alert actually fires and reaches a human. A monitoring check that has never been seen to alert is a hope, not a control. Then re-enable the job, and put a reminder in the team's calendar to repeat the test whenever the report server or mail system changes.
The scheduling checklist
Run through this list for every report you put on a schedule. It takes ten minutes and prevents most of the failures above.
| # | Item | Done when |
|---|---|---|
| 1 | Engine chosen | Tool scheduler if the tool can produce the report as designed; script otherwise |
| 2 | Schedule set for when it is read | Arrives before the meeting that uses it, in the reader's working hours, after the period's data is complete |
| 3 | Format matched to the reader | HTML to glance at, PDF to file, CSV to work with; two formats where both are needed |
| 4 | Channel matched to the reader | Chat for ops, email for management and clients, shared folder for audit and archives |
| 5 | Recipients are a list, not names | Membership changes do not require editing the job |
| 6 | Naming convention applied | audience-report-period, sortable date, no "latest" or "final" |
| 7 | Subject line carries the answer | Period and headline visible without opening the message |
| 8 | Archive copy retained | Every generated report saved to the write-once folder with its CSV |
| 9 | Missing-report alert in place | Something fires when the report does not arrive, and the alert has been tested |
| 10 | Run log and sanity rule | Each run logged; an empty or tiny data set blocks sending and alerts instead |
| 11 | Owner named | One person is responsible for the job, and it says so in the report footer |
| 12 | Credentials documented | The account the job runs as, and where its password lives, are written down for whoever inherits it |
Item 2 deserves a word. The best time to deliver a report is shortly before the moment it will be used, not the moment the data is available. A monthly management report that arrives at midnight on the first is buried by nine. One that arrives at eight on the morning of the monthly review is read. Look at the reader's calendar, not the job's.
Where to go next
Delivery is the last mile of report automation. A report that arrives where its reader already looks, on the morning it is needed, with an alarm on its absence, has earned its electricity. The earlier miles are covered by the rest of this pillar. The guide to which reports deserve automation decides what to schedule at all. The guide to report templates per audience shapes what each channel carries. The guide to killing the hand-built spreadsheet is the migration guide for the scripted engine. The guide to reports people actually read is about making the delivered report worth opening. All of them live under Automating Routine Reports.
Put this into practice with SysPeep Monitor
The first engine above — the monitoring tool's own scheduler — is the one to configure before writing any script. SysPeep Monitor produces reports from the same web console as its live dashboards and can email them to the people who need them. Check the product documentation for the report types and scheduling options. The free trial installs on any Windows machine in minutes.
Download Free TrialLearn About SysPeep