Home › Topics › Automating Routine Reports › Which Reports to Automate

Which Reports Deserve Automation (and Which Deserve Deletion)


Somewhere on your team there is a report that eats a morning every month. Someone exports three tables from the monitoring tool and pastes them into a spreadsheet. They mend the formulas that broke since last time, screenshot a chart, and email the result to fourteen people. One of them opens it. The ritual continues because it always has.


The obvious fix is "automate it". It is often the right answer, but it is not the first step. Automate a report nobody reads and the nobody-reads-it problem now runs on a schedule. Before you write a script, you need to know which reports exist and who they are for. You also need to know what they cost to produce and whether anything happens when they arrive. That list is the report inventory, and building one is the subject of this guide.


By the end you will be able to build the inventory in an afternoon. You will use a short set of criteria to sort every report into one of four fates: automate, keep manual, merge, or delete. And you will do the effort arithmetic that shows whether automating a given report is worth the work. You will also have a template table you can fill in for your own team. Plus, you will have a list of the reports that almost always deserve automation first.


In this guide

  • Why inventory before you automate
  • Building the report inventory
  • The four fates: automate, keep manual, merge, delete
  • The criteria table
  • The classic automation candidates
  • The effort math
  • The inventory template
  • Where to go next

Why inventory before you automate


Reports accumulate the way cables do behind a rack. Each one was added for a reason. Perhaps it was an outage that made a manager nervous or a client who asked a question once. Perhaps it was an auditor who wanted a printout or a predecessor who liked charts. Reasons expire; reports do not. Nobody cancels a report, because nobody owns the decision to cancel it. And the person who receives it assumes the person who sends it must need to send it. The result is a team producing a dozen reports that could not tell you which three matter.


Automation makes this worse if you skip the inventory. A manual report at least has a visible cost — someone groans about it every month, and that groan is a kind of feedback. An automated report costs nothing to produce, so it never gets questioned. It will run unchanged for years, quietly filling inboxes and shared folders, long after the person who asked for it has changed jobs. Inventory first; then automate what deserves it.


The second benefit is quieter: the inventory reveals duplication. It is common to find three people producing three reports that answer the same question with different numbers. That is because each computes "availability" or "incident count" its own way. Management notices when two reports disagree, and the disagreement costs more credibility than either report earns. Finding those overlaps is worth the inventory on its own.


At Tollgate Foods the inventory turned up nineteen recurring reports for a team of four. Three of them answered "how available were we?" with three different numbers, and the board had been quoting whichever arrived first. Four more were still coming out of the monitoring tool's scheduler for a manager who had left, addressed to a mailbox that no longer existed. The skip test retired seven in one quarter and nobody asked after any of them. The review minuted this as "a rationalisation".


Building the report inventory


First, decide what counts as a report. Use a wide definition: any document, page, email, or spreadsheet that someone produces on a recurring basis, from data, for someone else to read. That includes the Monday status email, the monthly capacity chart, and the quarterly slide for the leadership meeting. It includes the availability figure that goes into a client's service statement and the printout an auditor takes away. It also includes the dashboard screenshot someone pastes into a chat channel every Friday. If it recurs and it carries numbers, it goes on the list.


Where reports hide

You will not find them all by thinking hard. Go looking in five places:

  • Ask each person on the team two questions: "What do you send out regularly?" and "What do you receive regularly?" Both directions matter. People forget what they send because it is routine, and forget what they receive because they do not read it.
  • Search the shared mailboxes and distribution lists for recurring subject lines. A subject that appears every Monday, or on the first of each month, is a report whether or not anyone calls it one.
  • Walk the shared folder where documents get saved. Filenames with dates in them are the tell: anything that exists as a series is a recurring report.
  • Open the monitoring tool's own report scheduler. You may find reports configured years ago by someone who has since left, still emailing an address that bounces. These are reports too, and they are the easiest ones to delete.
  • Check calendar reminders. A recurring appointment called "build the monthly report" is the surest sign that a manual ritual exists and that somebody dreads it.

The six columns

For each report, record six things. Keep the wording plain; this is a working document, not a policy.

  • Name: what people actually call it. "The Monday email," not "Weekly Infrastructure Status Summary."
  • Audience: who receives it, and separately, who reads it. These are rarely the same list. A distribution list of twelve may have one genuine reader.
  • Frequency: daily, weekly, monthly, quarterly, or on demand. Note whether the cadence is a real requirement or just habit.
  • Source of the numbers: which tool, which export, which log, which ticket query — or which person's memory. Be specific. "The monitoring tool" is not a source; "the availability report exported as CSV from the monitoring tool, filtered to production" is.
  • Effort to build: honest hours per cycle, including fixing broken formulas, chasing missing data, and the follow-up questions the report generates. People underestimate this column badly, so ask the builder to time it once.
  • Does anyone act on it? The hardest and most important column. Ask each reader directly: "When did this report last change what you did?" A specific answer ("we ordered the extra disk in spring because of the capacity page") means yes. "It's good to have" means no.

Tip: the "acts on it" question feels rude to ask, so people skip it. Ask it anyway, and ask it kindly — "I'm trying to make sure this report is still giving you what you need" gets an honest answer. A report that nobody acts on is not a failure of the reader. It is a report that has outlived its reason, and finding that out is the whole point of the exercise.


The four fates: automate, keep manual, merge, delete


Once the inventory exists, every report gets exactly one of four fates. Resist the urge to invent a fifth ("keep for now, decide later"). That is how the inventory becomes another report nobody acts on.


Automate

A report earns automation when it is regular, stable, and acted on. Regular means a fixed cadence, so a schedule can produce it. Stable means the shape is the same every cycle: the same sections, the same tables, the same charts, with only the numbers changing. Acted on means a specific person makes a specific decision because of it. All three conditions matter. A regular, stable report nobody uses is a deletion candidate. And a report that is acted on but rebuilt differently every time is not ready for a script yet. For that report, stabilize its format first, then automate.


Keep manual

Some reports should stay handmade. The first group is anything rare: a report produced twice a year takes less effort to build by hand than to automate and maintain. The second group is anything where the value lives in judgment rather than numbers. Examples are a post-incident review, a budget justification, and the answer to a pointed question from the board. The commentary is the report; the numbers are supporting evidence. Note, though, that even a manual report can have automated ingredients. The narrative stays human, but the availability table pasted into it can come from a script.


Merge

Look for clusters of reports that answer one question. The weekly availability email, the monthly uptime chart, and the client's service statement all answer "how available were we?" They use the same data, computed three slightly different ways, by three different people. Merge them into one computation with three views. The numbers finally agree, one script replaces three rituals, and every reader gets the depth appropriate to them. Merging is where credibility comes back, because one source of truth is by definition never in disagreement with itself.


Delete

The fourth fate is for reports nobody reads. You will suspect several, but suspicion is not proof, so run the test: skip a cycle and see who asks. Tell your own manager you are running the experiment, so it is not mistaken for you forgetting, and then simply do not send the report. If nobody asks within a full cycle, retire it formally — announce that it has ended, archive the old copies, and remove the calendar reminder. If someone does ask, you have learned who the real audience is. Often it is one person, who turns out to need one number. That person would be perfectly happy with a much smaller report or a line in an existing one.


The criteria table


Here is the decision in one table. Read each row for the report you are classifying; the column where most rows land is its fate.


CriterionAutomateKeep manualMergeDelete
CadenceFixed and frequent (daily to monthly)Rare, irregular, or on demandAny, but overlaps another report's cadenceAny
FormatSame sections every cycleReshaped each time to fit the questionSections duplicate another report'sIrrelevant
Source of numbersA tool export, an API, a log, a ticket queryInterviews, judgment, one-off investigationThe same sources as a sibling reportIrrelevant
Acted on?Yes, by a named person, for a named decisionYes, but the action depends on the commentaryYes, but the readers overlap with another report'sNobody can name an action
EffortHours per cycle worth recoveringLow, or automation would cost more than it savesEffort is duplicated across reportsAny effort at all is too much
The tell"I check it every time it arrives.""It depends what happened that quarter.""Wait, whose numbers are right?""Oh, does that still go out?"

Two rows deserve a comment. The "acted on" row is decisive. A report that fails it goes to Delete regardless of how it scores elsewhere, because a beautifully automated report with no reader is still waste. And the "source" row is a readiness check rather than a verdict. A report whose numbers come from a person's memory cannot be automated until you find where those numbers really originate. Finding that source is the first step in retiring a hand-built spreadsheet.


The classic automation candidates


Certain reports appear in nearly every IT team, and they nearly always pass the three tests. If your inventory is long and you need somewhere to start, start here.


The weekly availability report

Which services and devices were down, for how long, and whether the week met whatever commitment exists. This is regular, its format never changes, and someone in management reads it because it is the number they get asked about. The only design decision is how you compute the figure. Agree on that once, using the principles in honest availability math, and then let the schedule produce it.


The monthly capacity trend

This shows disk usage, bandwidth utilization, and CPU load on the systems that matter as a trend rather than a snapshot. That way, anyone can see what is heading toward a limit. The report is acted on because it drives purchases. The monitoring history already holds every number it needs. And the connection between the trend line and the purchase order is the subject of utilization to purchase timing.


The incident summary

How many incidents, how long they took to detect and resolve, and which systems caused the most of them. Its readers act on it by prioritizing fixes. The metrics themselves and how to present them are covered in reporting incident metrics. Here the point is only that the summary is a textbook automation case.


The client report

If you support other organizations, each one expects a periodic statement of their service, their incidents, and their agreed levels. The format is fixed by contract, the cadence is fixed by contract, and the client certainly reads it. The specifics of building one are in client reporting and SLAs.


The audit evidence set

Auditors want proof that controls operate: that monitoring ran, that alerts fired, that backups completed. Gathering it by hand the week before an audit is miserable; producing it on a schedule means it is always ready. What goes into such a set is laid out in evidence packs.


The "what changed" report

Less famous but very useful: a weekly list of devices newly discovered, interfaces added or removed, and thresholds edited. Ops teams act on it immediately ("who plugged that in?"), and the monitoring tool already knows every entry.


The effort math


Automation has a cost, and the honest way to justify it is arithmetic you can show to whoever approves your time. You do not need precise numbers; you need your numbers. So use the method below and fill in your own figures rather than borrowing anyone else's. I once borrowed a colleague's and got a payback of three weeks, which should have told me something.


Start with the manual cost. Take the hours per cycle from the inventory and multiply by the cycles per year:


manual_hours_per_year = hours_per_cycle * cycles_per_year

Then estimate the automation cost in two parts. The first is the one-time hours to build the script or configure the scheduler. The second is the ongoing hours per year to maintain it, because sources change, exports move, and mail servers get replaced. Be generous with the maintenance estimate; automated reports break more often than their authors expect.


automation_hours_year_one = build_hours + maintain_hours_per_year
automation_hours_later    = maintain_hours_per_year

saved_per_year   = manual_hours_per_year - maintain_hours_per_year
payback_in_years = build_hours / saved_per_year

If the payback is under a year, automate without further debate. If it is several years, the report is either rare enough to keep manual or cheap enough to leave alone. Do the same sum for every report in the Automate column and sort by payback; that ordering is your roadmap.


The formula leaves out three things you should mention when you present it, because they all favor automation more than the raw hours suggest. First, errors: a hand-built report is wrong more often than an automated one. And a wrong number in front of management costs more than the hours to produce it. Second, delay: a manual report arrives when someone has time; an automated one arrives when it is due. Third, the bus factor: the manual report lives in one person's head, and when they leave, the report leaves with them. A script at least leaves behind a script.


Note also what the sum says about deletion. A deleted report saves its entire manual cost at zero build cost. Its payback is immediate. This is why the inventory pays for itself even if you automate nothing. Removing two unread reports frees more hours than automating one that is read.


The inventory template


Below is the table to copy. The rows are illustrative examples of the kind of thing an inventory turns up, not real figures; replace them with your own. The last column is filled in only after you have done the criteria and the arithmetic.


ReportAudience (receives / reads)FrequencySource of numbersEffort per cycleActed on?Fate
Monday status emailWhole IT list / two team leadsWeeklyMonitoring tool downtime export, ticket queueExample: about an hourYes: sets the week's prioritiesAutomate
Monthly capacity workbookIT manager / IT managerMonthlyMonitoring tool disk and interface history, pasted by handExample: half a dayYes: drives purchase requestsAutomate
Uptime chart for leadershipLeadership list / unknownMonthlySame data as the Monday email, computed differentlyExample: two hoursUnclear: nobody named an actionMerge into the status report
Client service statementEach client contact / each clientMonthlyMonitoring tool availability per client siteExample: an hour per clientYes: contractualAutomate
Post-incident reviewTeam and affected owners / sameAfter major incidentsTimeline, logs, interviewsVariesYes: produces action itemsKeep manual (automate the timeline table)
Printer fleet summaryFacilities list / nobody foundMonthlyPrinter SNMP page countsExample: an hourNo: last acted on unknownDelete (run the skip test first)

A few things to notice in the example. The uptime chart for leadership is not deleted even though nobody named an action, because its content is useful. The chart just duplicates the status report, so it merges. The post-incident review stays manual, but the effort column drops once its timeline table is generated rather than typed. And the printer summary is the classic candidate for the skip test. It is sent to a list and read by nobody anyone can identify. It costs an hour a month to establish that the printers remain printers. (They do.)


Running the exercise

Practically, the whole thing fits in a week of part-time attention:

  • Day one: collect the report list from the five hiding places. Do not classify yet; just find them.
  • Day two: fill the six columns, asking builders for honest effort and readers for the last action taken.
  • Day three: classify each report using the criteria table, and mark merge clusters.
  • Day four: do the effort math for the Automate column and sort by payback.
  • Day five: start the skip test on every Delete candidate, and pick the top of the Automate list to build first.

Keep the inventory itself as a living document and revisit it a couple of times a year. New reports will appear — that is fine, as long as each one arrives with an audience, a source, and a reason.


Where to go next


With the inventory done, you know which reports to build, which to fold together, and which to let go. The report that went to fourteen people can now go to the one who opened it, and nobody's morning goes with it. The rest of this pillar covers the building. Report templates per audience shows how one data source becomes four different reports for ops, management, clients, and auditors. Scheduled delivery gets the finished report to where its reader will actually see it. Killing the hand-built spreadsheet is the step-by-step migration for your most painful manual report. And reports people actually read is about making sure the automated version is worth opening. All five live under Automating Routine Reports.


Put this into practice with SysPeep Monitor

Several of the classic candidates above — availability, capacity trends, what changed — are exactly the numbers a monitoring system already holds. SysPeep Monitor produces reports from the same web console as its 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