Home › Topics › Defining and Tracking Internal SLAs › SLA Reporting Without Spin

SLA Reporting Without Spin


The monthly SLA report is the only regular contact most of the organisation has with the IT team. And it arrives as an email attachment nobody asked to receive. It has one job, and the job is not to make IT look good. The job is to be believed. A report that is believed can carry bad news. A report that is not believed cannot carry anything, including good news, because the good news reads as spin.


Spin is not lying. It is the average that hides the one bad service, the "adjusted" column that quietly removes the worst week, the exclusion nobody mentioned until the number needed it. Each one is defensible on its own. Together they turn the report into a document that only its author reads. The first time a manager compares it with their own memory of the month, the whole thing loses its standing.


By the end of this guide you will have a one-page layout to fill in. You will have a worked example of why averaging across services is the fastest way to lose trust. You will have a way of writing up a miss that explains without excusing. You will also have a trend view that puts one bad month in proportion. And you will have a rule about red that will feel wrong the first time and right ever after.


In this guide

  • The report exists to build trust
  • The monthly report layout
  • The averaging trap, with numbers
  • Showing a miss plainly: explain, don't excuse
  • Trend over months, not a single month
  • The month with an exclusion
  • Why a report that never shows red is not read
  • Getting it to people
  • A short story from Bellweather Logistics
  • Checklist before you send
  • Where to go next

The report exists to build trust


Trust is the only thing the report produces, and it is built by showing red. Anyone can publish a page of green. A page of green with one honest red row, and two plain sentences under it, tells the reader that when this report says green, it means it. That is the asset. Every decision about layout, wording, and what to leave out should protect it.


The reader is not a technician, and the report should not make them one. A department head wants to know three things per service: what was promised, what was delivered, and whether the gap between them is being dealt with. Everything else — check intervals, vantage points, the query that produced the minutes — belongs in the measurement specification. The report can reference that specification in a footer for the one reader in fifty who wants it.


Keep it to a page. A page gets read on a phone in a corridor; a deck gets filed. If the month genuinely needs more explaining than a page allows, the extra goes in the miss review. That is a different document with a different audience. In that case, the report links to it.


The monthly report layout


One table, one row per service, the same columns every month. The columns are the promise, the result, a plain met-or-missed word, the worst single incident, and a note. The note is one line: empty on a good month, and on a bad month the first sentence of the explanation below the table.


ServiceHoursTargetAchievedMet / MissedWorst incidentNote
ERPMon–Fri 06:00–20:0099.5%99.74%Met31 min, 9th, 08:12
EmailMon–Fri 07:00–19:0099.0%99.91%Met12 min, 22nd
Internet accessMon–Fri 07:00–19:0099.0%99.55%Met59 min, 3rd (ISP)ISP outage counted, not excluded: no ISP ticket
File serverMon–Fri 07:00–19:0099.0%98.38%Missed171 min, 14th, 09:10Storage controller failure; see below
Wi-Fi (office)Mon–Fri 07:00–19:0099.0%99.97%Met4 min, 27th
P1 responseBusiness hours15 min, all P19 of 10 within 15 minMissed41 min, 14thSame incident as file server
P2 responseBusiness hours1 h, all P223 of 23 within 1 hMet48 min
<service><hours><target><achieved><Met / Missed><duration, date, time><one line, or blank>

Under the table go three things, in this order. First comes a paragraph for each missed row. Next comes a one-line list of exclusions applied this month with their evidence. Last comes a footer naming the measurement specification and where the figures came from. Then stop. The report's numbers come straight from the monthly compliance worksheet; the report is the worksheet with the arithmetic columns removed and the words added.


The averaging trap, with numbers


Do not average availability across services. Not once. Not "just for the headline number." An average across services is a figure that belongs to no service, and it hides exactly the row the reader most needs to see. The arithmetic shows how much it hides, so here is a month with ten services.


Nine services achieved 99.9%; one service (payroll) achieved 97.0%.
Target for every service: 99.5%.

Average = (9 × 99.9 + 97.0) / 10 = 996.1 / 10 = 99.61%   "target met"

But payroll, business hours, 13,200 minutes in the month:
   13,200 × (1 − 0.970) = 396 minutes down = 6.6 hours
   ...during the week payroll is run.

The average says the department met its target with room to spare. The payroll team spent most of a working day unable to pay anyone. Both statements are true, and only one of them will be remembered. And the report that led with the other one will not be opened next month. Headline figures, if the reader insists on one, should be "services that met their target: nine of ten." That is a count, and counts do not hide anything.


If you build the report in a spreadsheet, the cells that summarise the month should count and find the minimum, never average. With the achieved percentages in one column and the met-or-missed words in another:


Services meeting target:  =COUNTIF(met_range, "Met") & " of " & COUNTA(met_range)
Lowest availability:      =MIN(achieved_range)
Worst service:            =INDEX(service_range, MATCH(MIN(achieved_range), achieved_range, 0))

The same trap appears in time, not just across services. A quarter's figure averages three months and hides the one that missed. Report each month on its own row and let the reader see the shape; the trend section below covers how. A summary that is the average of its parts has never once been the part anyone was worried about.


Showing a miss plainly: explain, don't excuse


A miss gets a paragraph, and the paragraph has three parts in a fixed order: what happened, what it cost, what changes. Nothing else. No "unfortunately," no "despite the team's best efforts," no reference to how well the other nine services did. The reader already knows about the other nine; they are in the table.


What happened is two sentences of fact with a time. The service, the start, the end, the cause as far as it is known. If the cause is not yet known, say that, because "cause under investigation" is a fact and "a complex interaction of factors" is a smokescreen.


What it cost is one sentence in the reader's terms, not yours. Not "171 minutes of downtime" but "the drawing office could not open project files for a morning and dispatch reverted to paper for an hour." You are the only person who can translate minutes into that sentence, which is why the report is worth writing rather than generating.


What changes is one or two sentences naming a specific action with an owner and a date, taken from the miss review. If nothing is changing, say why. "This failure mode is inside the target and we accept it" is a legitimate answer. It is also a more credible one than an action item invented to fill the line. The review that produces the action is described in when you miss the SLA.


FILE SERVER: 98.38% against 99.0%. Missed.
What happened: on the 14th the primary storage controller failed at 09:10.
Service was restored at 12:01 from the standby controller; a firmware
mismatch between the two delayed the swap. Total outage 171 minutes,
all inside service hours.
What it cost: the drawing office could not open project files for the
morning, and dispatch ran on paper for about an hour.
What changes: standby controller firmware is now checked during monthly
patching (owner: infrastructure lead, from next cycle), and a test
failover is booked for the next maintenance window.

That paragraph is the template. Copy its shape for every miss, every month, and the reader learns that a miss in this report always comes with a cost and an action. (They will also learn to skip straight to the paragraph. That is fine; it means they trust the table.)


Trend over months, not a single month


One month tells the reader almost nothing about whether a service is reliable, because one month contains one or two outages and their timing is luck. Six months in a row tells them whether the misses are a pattern. Put a small trend table under the main one. Give it one row per service, one column per month, and the achieved figure in each cell. Add a final column counting the months that met the target.


ServiceM1M2M3M4M5M6 (this month)Months met of 6
ERP (99.5%)99.81%99.62%99.90%99.71%99.55%99.74%6
File server (99.0%)99.42%98.71%99.20%98.90%99.31%98.38%3
Internet (99.0%)99.63%99.70%99.12%99.80%99.58%99.55%6

The trend table changes the conversation about the file server. One miss in six is a bad month; three in six is a target that is wrong or a service that is broken. And the report should say which one you think it is. The count in the last column is the single most useful number on the page. And it takes one formula per row: =COUNTIF(month_cells, ">=" & target_cell).


Resist the temptation to add a graph unless the trend is genuinely hard to see in the numbers. Six figures in a row are readable. A line chart with six points and a legend is a decoration that takes up the space the explanation paragraph needed.


The month with an exclusion


An exclusion is a slice of the month the target did not cover, and it goes on the report by name, every time. "Exclusions applied this month: one, ERP, 60 minutes, announced upgrade on the 20th at 07:00, calendar entry dated the 13th." One line, with the evidence that it was decided before the outage. A reader who sees that line every month, including the months when it says "none," never has to wonder whether the figure has been cleaned.


If an exclusion is large, show the number both ways. "ERP: 99.74% after the announced 60-minute upgrade; 99.42% if it were counted." That second figure costs you nothing, because the agreement says the first one is the measure. It buys a great deal, because it proves you are not hiding behind the rule. The idea of an availability budget that planned work draws from, so that upgrades are decided rather than sneaked, is in planned downtime.


Never add an exclusion in the report that is not in the agreement. The month might have had a category of outage nobody had thought to exclude — a power cut to the whole building, say. In that case, it counts and the row shows red. The explanation paragraph says the target may need a new exclusion, to be discussed at the review. That is a miss with a proposed fix. Quietly excluding it is the thing this whole guide is about.


Why a report that never shows red is not read


A report that is always green becomes wallpaper. The reader learns in three months that the numbers never change and nothing in it needs a decision. From then on the email is deleted unopened. That is a reasonable response to a document that has never once said anything. Green is not information. Green is the absence of information, and people are good at noticing absence.


Worse, a page of green is not believed, because the reader has their own memory of the month. They remember the morning the file server was down. If the report does not, then either the report measures something other than their experience, or it is being tidied, and neither conclusion helps you. One honest red row a quarter does more for the standing of the report than a year of green.


If the services really did meet every target, say so plainly. Then use the space for something true: the closest call ("the ERP used 34 of its 66 allowed minutes"). Or cover the trend or the exclusion line reading "none." A good month reported honestly still has content. I have sent reports that were entirely green and entirely unread, and I did not notice the second part for some time.


Tip: put the "met / missed" column in plain words, not just colour. Colour is lost in a forwarded email, a printout, and on the screen of the one reader who cannot tell red from green. "Missed" survives all three.


Getting it to people


The report goes out on the same day every month, to the same list, from the same address, whether the month was good or bad. Consistency is part of the trust. A report that appears on the fifth in good months and the twelfth in bad ones has told the reader the result before they open it. The mechanics of sending a report on a schedule without a person remembering to do it are in scheduled delivery. And the same monitoring system that produced the numbers can usually send its own availability report on a timer as the raw attachment.


The one-page version for management is a slightly different document with the same table. The whole art of putting monitoring numbers in front of people who do not run networks is its own pillar at monitoring data for management. The short version is that the SLA report is already most of the way there. A promise, a result, and a plain sentence about the gap is the format managers wanted all along. Which reports get opened at all, and why, is in reports people actually read.


Send it as a page, not a link to a dashboard. A dashboard shows now; the report shows the month, signed off, unchanging. Six months later, when the target is renegotiated, the six reports are the evidence, and a dashboard has no memory of what it showed in March.


A short story from Bellweather Logistics


Bellweather Logistics added a column called "adjusted availability" to its monthly SLA report after a bad quarter. The column removed any outage that a subsequent review had classified as "not reasonably preventable." The raw column kept showing misses; the adjusted column showed green for eleven consecutive months. And the adjusted figure was the one that went in the board summary. In the twelfth month the operations director, who had spent an afternoon in a depot with no scanners, asked which column the depot had experienced. The adjusted column was retired at that meeting, along with a certain amount of goodwill. And the following report showed three red rows and a paragraph under each. It was the first one she replied to.


Checklist before you send


Read the finished page against this list. Anything that fails is fixed before it goes out, not explained afterwards.


  • One row per service, with the same columns as last month, in the same order.
  • Every row shows target, achieved, and a plain "Met" or "Missed" in words.
  • No cell anywhere on the page is an average across services or across months.
  • Every "Missed" row has a paragraph below the table with what happened, what it cost, and what changes.
  • Every "what changes" line names an owner and a date, or says plainly that nothing changes and why.
  • The exclusions line is present, even when it reads "none," and every exclusion has its evidence.
  • No exclusion appears that is not in the agreement.
  • Large exclusions show the figure both with and without them.
  • The trend table shows at least six months and a count of months met.
  • The footer names the measurement specification and the source of the figures.
  • The whole thing is one page and reads on a phone.
  • It is going out on the usual day, to the usual list, regardless of the result.

Where to go next


A report built this way is dull in a good month and useful in a bad one, which is the correct way round. The numbers it carries come from measuring SLA compliance from monitoring data. The targets it reports against were set in choosing what to promise. And the document all of it serves is described in what an internal SLA is. When the red row appears, the review that follows it is in when you miss the SLA. The full set is collected in Defining and Tracking Internal SLAs.


Put this into practice with SysPeep Monitor

The table in this guide fills straight from an availability report. The monitoring software behind this site keeps availability reports and status history for every service check in one web console. It can send scheduled reports on the same day each month. So the honest number arrives on time whether the month was good or bad. The free trial runs on a single Windows machine.

Download Free TrialLearn About SysPeep