Home › Topics › Justifying Monitoring Spend to Leadership › Before-and-After Stories

Before-and-After Stories That Land With a Finance Director


You have the dividend table. Every cell traces to a timestamp, the sensitivity rows are ready, and the arithmetic is honest. You show it to the finance director, who nods, and then asks the question the table cannot answer: "But what does that actually look like?" The table has persuaded the head. It has not persuaded the room, and the room makes the decision.


The thing that persuades the room is a story, and not a long one. Four sentences about a fault that happened twice, once before monitoring and once after, with a ticket number and an alert timestamp to prove it. Told flat, without a hero, it takes forty seconds and it is the part of the meeting people repeat afterwards. I have watched a four-sentence story about a disk do what three pages of arithmetic could not. It made a finance director say "oh, I see".


By the end of this guide you will know the shape of a story that works. You will know how to mine your own history for pairs. You will know what evidence to attach so it is never "trust me". You will know the rules that keep it credible, and where to put the story in a one-pager or a meeting. And you will have a template you can fill in tonight. There are three worked examples, and one story that went wrong.


In this guide

  • Numbers persuade the head; stories persuade the room
  • The shape of a story that works
  • Mining your own history for pairs
  • The evidence to attach
  • The rules
  • Telling it in the one-pager or the meeting
  • The story template
  • Three worked examples
  • A war story: the one that backfired
  • The story checklist
  • Where to go next

Numbers persuade the head; stories persuade the room


A number tells a leader that something is true, and a story tells them what it would feel like. Both are needed, in that order. The dividend table from the downtime math guide establishes that monitoring removed a few hundred outage minutes this year and what they were worth. The story shows one of those minutes from the inside. It shows who was stuck, what they could not do, and what the same morning looked like the second time.


A story also survives the journey the number does not. A finance director who approves the line will have to explain it to someone above them. And they will not carry the sensitivity table up the stairs. They will carry "the disk thing", and if you have told it well they will tell it nearly as well. Give them something that can be retold in a corridor, because it will be.


A before-and-after story is a short account of one class of fault occurring twice. The first time is before a monitoring change, when it was found late and cost something. The second time is after, when it was found early and cost less. The same fault, twice, is the whole trick. A story about a disaster on its own asks the listener to imagine an alternative. The same disaster with a different ending does the imagining for them.


The shape of a story that works


The story is four sentences, and each sentence has one job. Learn the four jobs and you can build the story from any pair in your history in ten minutes.


Sentence one: what happened before

State the fault, the date, and how it was discovered. "Last spring the finance file server's disk filled during month-end close, and we found out when the finance team could not post." Plain, dated, and specific about the discovery, because the discovery is the part monitoring changes.


Sentence two: what it cost

State the cost in the business's own terms, from the ticket. "Eleven people could not work for three hours, and the close slipped a day." Not "significant disruption". Not a figure you invented. The number of people, the length of time, and the thing that did not happen, all from the incident record.


Sentence three: what is different now

State the monitoring change, in one clause a non-technical listener can follow. "That server now has a disk threshold and a trend alert at eighty per cent." Resist explaining how thresholds work. The listener needs to know that something was put in place, not how it is configured.


Sentence four: what happened the next time

State the second occurrence, with the timestamp and the outcome. "In autumn the same disk crossed eighty per cent on a Tuesday; the alert fired at 09:12, the disk was extended by eleven, and nobody in finance noticed." That last clause is the whole story, and it should be the last thing the listener hears. Do not add a moral. The moral is the silence after it.


Four sentences, forty seconds, done. The urge to add a fifth sentence explaining what it means is the urge to weaken it. (I have never once regretted stopping at four.)


Mining your own history for pairs


A pair is two incidents of the same fault class, one detected by a user and one by an alert, on either side of a monitoring change. Your incident log has them if you have been monitoring for more than a year. The way to find them is to sort by cause rather than by date. The classic pairs are a disk fill, a dead uplink, and an expired certificate. That is because they are common, they recur, and the before version is always memorable.


If your incident records carry a cause class and a detected-by field, the pairs fall out of one query. If they do not, this is the query to make you wish they did, and the field to add at the next post-incident review. Table names differ between systems; treat this as the shape:


-- Incidents grouped by cause, showing how each was first detected,
-- so that a user-found one and an alert-found one of the same class
-- can be placed side by side.
SELECT cause_class,
       detected_by,          -- 'user' or 'alert'
       COUNT(*)              AS incidents,
       MIN(opened_at)        AS earliest,
       MAX(opened_at)        AS latest
FROM incidents
WHERE opened_at >= DATE('now', '-24 months')
GROUP BY cause_class, detected_by
ORDER BY cause_class, detected_by;

Any cause class with a row for "user" and a later row for "alert" is a candidate pair. Pick the pair where the before incident cost the most in the business's terms and the after incident is cleanest. Here, cleanest means the alert fired, the fix followed, and no user ticket exists. An after incident with three complaints makes a story about improvement, which is fine; one with none makes a story about invisibility, which is better.


Fault classBefore (user-detected)After (alert-detected)Evidence for each
Disk fillDatabase or file server stops; discovered by the department that uses itTrend alert at a threshold; disk extended before fullBefore: ticket number, people affected, duration. After: alert timestamp, capacity graph
Dead uplinkWhole floor or site loses connectivity; discovered by phone callsInterface-down alert; cable or port replaced before the shift startsBefore: ticket, time of first call, time of fix. After: alert timestamp, interface graph
Expired certificatePortal or mail stops with a warning nobody understands; discovered by customersExpiry alert weeks ahead; renewed on a quiet dayBefore: ticket, customer complaints, duration. After: alert timestamp, renewal date
Provider outageArgued about by email; credit not claimedLatency graph attached to the provider ticket; credit claimedBefore: email thread. After: provider ticket number, credit note

One story per meeting is enough. And the disk story is usually the strongest, because everyone in the room has run out of space on something. Keep the others for next quarter, and for the moment someone says "but that was just one time".


The evidence to attach


Every story carries three pieces of evidence so that no sentence in it is "trust me". They are not read out. They sit under the story on the page, or on the next slide, and their job is to exist. A finance director who can see a ticket number rarely asks to see the ticket. And the one who does will find it says what you said.


  • For the before: the ticket number, with the time it was opened, the time it was closed, and the number of people it lists as affected. If your before incident predates the ticket system, an email with a timestamp will do, and say so.
  • For the after: the alert log entry, with the timestamp the alert fired and the timestamp it cleared. This is the sentence-four proof, and it is the one piece of evidence a sceptic will actually check. That is because sentence four is the surprising claim.
  • A graph that shows the after incident: the disk climbing to the threshold and dropping after the extension, or the interface going flat and coming back. One graph, with the date on it, small enough to fit under four sentences.

If any of the three is missing, the story is still tellable, but say which one is missing. "We do not have a ticket for the first one because it was before the help desk; the finance manager remembers it" is honest and, told without embarrassment, credible. What you may not do is round the missing piece into existence.


The rules


The rules are about credibility, and they all point the same way: the deadpan version is stronger. A story that sounds like a story is discounted. A story that sounds like a log entry is believed.


  • No heroics. Nobody "raced in" or "worked through the night". The alert fired, the fix was made, the time is on the log. A story with a hero has a villain waiting in the next sentence, and the room will start looking for one.
  • No villains. Not the vendor, not the provider, not the department that filled the disk with video files. The fault is the subject; people are not.
  • No blaming the previous administrator. Even when it is warranted. Especially when it is warranted. The before incident happened "before we had a threshold on it", not "before I arrived".
  • No exaggeration. Three hours is three hours, not "most of the day". Eleven people is eleven, not "the whole department". The evidence underneath will contradict any inflation, and one contradicted number discounts every true one after it.
  • No jargon in the four sentences. "Disk threshold" is fine. "SNMP trap on hrStorageUsed" is not, however accurate. The evidence lines can carry the technical detail; the sentences cannot.
  • Say the date. A story without a date is an anecdote. "Last spring" and "in autumn" are enough in the telling; the ticket has the rest.

Tip: rehearse the story once out loud, with a clock. If it takes more than a minute you have added a sentence. Find it and take it out. The usual culprit is the one that explains how the alert works.


Telling it in the one-pager or the meeting


The story has two homes. On paper it is a boxed paragraph with the evidence lines beneath it, next to the number it illustrates. The layout is in the monthly one-pager guide. The story fits its "what changed" box without alteration. In a meeting it is spoken, once, after the number and before the ask, with the evidence on the slide behind you, unread.


Timing matters more than delivery. Tell it after the dividend number, so the listener already believes the scale. Tell it before the budget request, so the request arrives in the silence after "nobody noticed". Do not open with it, because a story told first sounds like a pitch. And do not close with it, because the close belongs to the decision. The guide on presenting monitoring data in meetings covers the rest of the room. This guide's only addition is: number, story, ask, in that order.


Tell it the same way every time. A story that changes between the one-pager and the meeting is a story someone will notice changing. Write the four sentences once, keep them in the evidence file, and say them verbatim. Finance directors have heard a great many stories, most of them about why a project was late. One that is identical on paper and out loud is a novelty they tend to remember.


The story template


Fill in every line. Leave nothing blank. If a piece of evidence does not exist, write "none: [reason]" in its place so the gap is visible and deliberate.


BEFORE-AND-AFTER STORY: [fault class, e.g. disk fill on finance file server]

1. Before:      [When] the [system] [failed how], and we found out when
                [who] [could not do what].
2. Cost:        [N people] could not [do what] for [duration], and
                [the business consequence].
3. Change:      [System] now has [the monitoring change, one clause].
4. Next time:   [When] the same [fault] happened; the alert fired at
                [HH:MM], [the fix] was done by [HH:MM], and
                [what users experienced, ideally nothing].

Evidence
  Before:   ticket [number], opened [date time], closed [date time],
            affected [N]
  After:    alert [id or name], fired [date time], cleared [date time]
  Graph:    [metric] on [system], [date range], attached
  Missing:  [none, or which item and why]

Told in:  [monthly one-pager, month] / [meeting, date]
Same wording each time:  yes

Three worked examples


All three organisations are fictional, and every detail is invented to show the shape. Notice what is not in them: no adjectives about how bad it was, no names, no explanation of how monitoring works.


Harbourview Clinic: a disk fill

Last winter the appointments database server's disk filled overnight, and we found out when reception could not book anyone in at eight. Reception worked from a paper diary for two and a half hours, and forty appointments were re-entered by hand that afternoon. That server now has a disk threshold and a trend alert at eighty per cent. In summer the same disk crossed eighty per cent on a Wednesday; the alert fired at 14:20, the disk was extended by 15:05, and reception never knew.


Evidence: before, ticket 4471, opened 08:04, closed 10:36, affected: reception (six staff); after, alert "db01 disk C: 80 per cent", fired 14:20, cleared 15:07; graph of disk use on db01.example.com across the week, attached.


Acme: a dead uplink

Two springs ago the uplink from the warehouse switch to the core failed at about six in the morning, and we found out when the warehouse phoned at seven-fifteen because the scanners had stopped. The first shift picked by paper for an hour and forty minutes, and the morning's despatch went out late. The warehouse uplink now has an interface-down alert that goes to the on-call phone. This spring the same port failed at 05:52; the alert fired at 05:53, a spare cable was in by 06:30, and the first shift started on time.


Evidence: before, ticket 2218, opened 07:15, closed 08:55, affected: warehouse (fourteen staff); after, alert "core-sw1 Gi1/0/24 down", fired 05:53, cleared 06:31; graph of interface status for that port, attached.


Ashgrove Insurance: a certificate

Last year the customer portal's certificate expired on a Saturday morning, and we found out from customers, by email, on Monday. The portal showed a security warning for the whole weekend, and the service desk logged thirty-one customer contacts about it. Every certificate we own is now monitored, with an alert thirty days before expiry. This year the same certificate reached its thirty-day mark on a Thursday; the alert fired at 00:10, the renewal was done the following Tuesday during a quiet hour, and no customer saw a warning.


Evidence: before, ticket 3390, opened Monday 08:22, closed Monday 10:40, customer contacts: thirty-one; after, alert "portal.example.com certificate expires in 30 days", fired 00:10, cleared on renewal; renewal record dated, attached.


Each story is under a hundred words, has a number in sentence two that came from a ticket, and ends with the users not noticing. Read aloud, each takes about thirty seconds, which leaves plenty of room for the silence to do its work.


A war story: the one that backfired


Kestrel Freight's administrator told the finance director a story about a failed storage controller. It was a good one: the alert had fired "within seconds", the standby had taken over "in under five minutes", and the depot "never knew". The finance director, who had run depots, asked to see the ticket. That was a reasonable request. The ticket showed the alert firing eleven minutes after the fault and the failover completing forty minutes after that. Two calls from the depot came in between. Every number in the story had been true in spirit and wrong in fact. The director approved nothing that day. The three entirely accurate stories in the following quarter's report were each met with "is that one exact?". The administrator now keeps the four sentences on a card and reads them from it, timestamps included. The director has described that as "an improvement".


The story checklist


Run every story through this before it goes on paper or into a meeting. Any no is a sentence to rewrite or evidence to find.


  • The story is about one fault class that occurred twice, on either side of a monitoring change.
  • It is four sentences: before, cost, change, next time.
  • Sentence one says how the before incident was discovered.
  • Sentence two's numbers (people, duration, consequence) come from the ticket.
  • Sentence three names the change in one clause with no configuration detail.
  • Sentence four has a timestamp from the alert log and ends with what users experienced.
  • There is no hero, no villain, no predecessor, and no adjective describing how bad it was.
  • The ticket number, the alert entry, and one dated graph are attached, or the missing one is named.
  • Spoken aloud, it takes under a minute.
  • It is told after the number and before the ask.
  • The wording is identical in the one-pager and the meeting, and is kept in the evidence file.

Where to go next


The dividend proves the scale; the story proves it was real. Build it from a pair in your own history. Keep it to four flat sentences. Put the ticket and the alert timestamp underneath. And tell it in the gap between the number and the ask, the same way every time. The finance director will carry it up the stairs, and it will arrive intact, which is more than most arguments manage.


What monitoring spend actually buys supplies the plain sentences a story illustrates. And the downtime math guide supplies the number it follows. The monitoring budget line is where the story earns its keep, paired with the dividend so the line is not the first cut. And when leadership says no explains why you keep collecting stories in the year you did not get the money. All of them are part of Justifying Monitoring Spend to Leadership.


Put this into practice with SysPeep Monitor

Sentence four of every story is an alert timestamp and a graph. SysPeep Monitor keeps its alert history, dashboards, and reports in one web console. So the evidence under a story is a screen you can open rather than a screenshot you have to find. The free trial at /download/ installs on a single Windows machine. You can use a lab copy to see what its alert log records before you build your first story from it.

Download Free TrialLearn About SysPeep