Home › Topics › Translating Monitoring Data for Management › Risk and Cost Framing

Framing Monitoring Findings as Risk and Cost


The finding is real. The graph shows it, the alert log confirms it, and you have explained it in plain words. Then the manager asks the question that stops most technical people cold: "How bad is it?" You said "it depends", which was honest, and the manager heard "not very", which was not what you meant.


"How bad is it?" is a question about risk. Risk has a shape that managers already understand, because they weigh it all day about suppliers, hires, and contracts. They do not need you to teach them risk. They need you to put your finding into the shape they already use: a likelihood, an impact, and a cost. State them honestly and without a decimal point you cannot defend.


This guide shows you how. It covers which of three kinds of likelihood a finding has and the monitoring data that proves it. It covers impact in the organisation's own terms. It shows a downtime-cost estimate from figures you can source rather than statistics you read somewhere. It offers a range instead of a false number, and a risk statement that fits on half a page.


In this guide

  • Risk, in words before numbers
  • The three likelihoods
  • Impact in the organisation's own terms
  • An honest downtime-cost estimate
  • Ranges instead of false precision
  • The risk statement template
  • The risk register
  • The trap of crying wolf
  • A war story
  • The risk-framing checklist
  • Where to go next

Risk, in words before numbers


A risk is how likely something bad is, multiplied by how bad it would be. That is the whole definition. Likelihood is the chance it happens in a period you name, usually the next year. Impact is what it does to the organisation when it does. A thing that is nearly certain but trivial is a low risk. A thing that is rare but would stop the company is a high one. Managers weigh both halves without thinking about it, which is why the framing works.


Say the risk in words before you put a number on it. "The backup internet line has failed twice this quarter, and if it fails on a day when the main line is also down, everyone in the building is offline until one of them is repaired" is a complete risk statement. It has a likelihood (it has happened, twice), an impact (everyone offline), and a duration (until repaired). A manager can act on that sentence with no arithmetic at all.


Words first because a number invites an argument about the number. Present "risk score 12" and the conversation becomes about how you got twelve. Present the sentence and the conversation becomes about what to do. The numbers come later, when someone asks what it would cost, and by then they have accepted the shape of the thing.


Do not write a risk score without the sentence. Not a three-by-three grid, not a colour, not a heat map. Especially not a heat map.


The three likelihoods


Likelihood comes in three kinds, and the monitoring data tells you which one you have. In order of strength they are: it has happened, it is trending toward happening, and it could happen. Name the kind every time. "Could happen" said with the same confidence as "has happened" is the fastest way to lose a manager's trust. The fastest way to earn it is to keep the three apart.


It has happened

The strongest likelihood is a record. The alert history shows the backup line dropping on two dates. The availability report shows the file server unreachable for forty minutes on a Tuesday. This is not a prediction, it is a count, and nobody can argue with a count that has timestamps. Pull the dates from the alert log and put them in the statement. The guide on honest availability math turns outage minutes into a percentage. Here you turn them back into dates, because dates persuade and percentages do not.


It is trending toward happening

The second kind is a graph with a slope. Examples are disk use rising by a steady amount per week or interface errors climbing month on month. Another is an uplink whose afternoon peak is a little higher every fortnight. The event has not happened, but the line is pointed at it, and you can say roughly when it arrives. Draw the line on the graph and mark the date it crosses the limit. A trend is weaker than a record, so say so: "at the current rate, the shared drive fills in about eleven weeks; the rate could change."


It could happen

The weakest kind is a structural exposure with no history behind it. The monitoring map shows one link between two buildings and no second path. The dependency view shows every service behind one firewall; the rack has one power feed. Nothing has failed yet, and nothing is trending, but a single fault would have a large effect. The guide to dependency thinking is how you find these. State them honestly as exposures: "we have no evidence this will fail, but if it does, here is what stops."


Rank your findings by kind before anything else. A "has happened" with a modest impact usually deserves attention before a "could happen" with a dramatic one. The first is a fact and the second is a story. Managers know the difference.


Impact in the organisation's own terms


Impact is what stops, described in the units the organisation counts itself in. A warehouse counts shipments per hour. A clinic counts appointments. A law firm counts billable hours, a school counts lessons, and a call centre counts calls answered. Find out what your organisation counts, because that is the unit your impact statement needs to be in. The fastest way to learn it is to ask someone in finance what number the board looks at first.


The common impact types are people idle, orders not taken, deadlines missed, and letters from regulators. People idle is the easiest to picture and the easiest to cost. Orders not taken is the one sales cares about. Deadlines missed is the one that carries penalties in contracts. A regulator letter is the one that gets the chief executive's attention in a way no graph will.


A five-step impact scale in words keeps everyone honest. Use it for every finding, and reserve the top step for things that actually stop the organisation.


Impact levelWhat it meansExample
1. Nobody noticesIT fixes it before anyone is affectedA backup job fails once and reruns successfully
2. A team works around itSlower or awkward, but work continuesWi-Fi is down in one meeting room
3. A team stopsOne department cannot do its main jobThe warehouse scanners lose the network
4. The organisation stopsMost staff cannot workThe internet connection and its backup both fail
5. The organisation stops and someone outside noticesCustomers, patients, or a regulator are affectedThe customer portal is down during the ordering peak

Write the level number and the words together, every time: "impact 3, the warehouse stops." The number lets you sort; the words stop the number from being mistaken for a fact.


An honest downtime-cost estimate


A downtime cost is only honest when every number in it comes from your own organisation. You may have seen a figure somewhere for the cost of downtime per minute. That figure belongs to whoever measured it, with their staff, their margins, and their outage. It does not transfer, and a finance director will know it does not the moment you say it. Build your own, out loud, from three parts.


DOWNTIME COST (one outage)

  people idle     = staff affected x loaded hourly cost x hours down
  lost business   = transactions per hour x share actually lost
                    x value per transaction x hours down
  recovery        = overtime, engineer call-out, expedited parts

  total           = people idle + lost business + recovery

Loaded hourly cost is what an hour of a person costs the organisation. It is salary plus the overheads that come with employing them, divided by working hours. Finance has this number and will usually give you a rounded figure if you explain why you want it. If they will not, use salary divided by working hours and label it as a floor, because it is one.


Share actually lost is the part people forget. When the order system is down for three hours, not every order in those hours vanishes; many are placed later. Some are not. Ask sales what fraction they believe walks away. They will give you a guess. A labelled guess from sales is worth more than a precise number from you, because it is theirs.


Here is the arithmetic for a fictional warehouse, Northgate Retail, with every input made up and labelled as an example. The numbers are in whatever currency you use; only the shape matters.


EXAMPLE INPUTS (all invented for illustration)

  staff affected         40 pickers and packers
  loaded hourly cost     30 per hour            (from finance, rounded)
  hours down              3                     (last outage lasted 2h50)
  orders per hour       120                     (from the order system)
  share actually lost     1 in 3                (sales team's estimate)
  margin per order        8                     (from finance)
  recovery              240                     (4h engineer call-out at 60)

  people idle    = 40 x 30 x 3          = 3,600
  lost business  = 120 x (1/3) x 8 x 3  =   960
  recovery       =                        240
  total          =                      4,800  for one three-hour outage

In a spreadsheet, with the inputs in cells B2 to B8 in that order, the total is one formula:


=B2*B3*B4 + B5*B6*B7*B4 + B8

Show the inputs, always. The estimate is persuasive because a manager can see each number and knows where to argue if they disagree. Hide them and the total becomes something you claimed; show them and it becomes something you worked out together. The guide to where spending actually helps uses the same habit for the other side of the ledger.


Tip: multiply the one-outage cost by how often it happens. A cost of 4,800 sounds like one thing when it happened once last year and a different thing when it happened twice last quarter. The "has happened" count from the alert history is the multiplier, and it is the number that turns an estimate into an argument.


Ranges instead of false precision


Give every cost as a range with a stated lean: "between X and Y, most likely nearer X." A single figure invites the finance director to check it, and one wrong input sinks the whole estimate. A range invites them to agree with the shape. It is honest, because you do not know whether the next outage lasts two hours or six.


Build the low end from the things that definitely stopped and the shortest plausible duration. In this example, those are the forty pickers, the two hours of the last outage, and no lost orders. Build the high end from everyone the system touches and a bad day. Here, that is the eighty people in the building, six hours, and a third of orders gone. Then say which end you believe and why. "Most likely nearer the low end, because the last two outages were both under three hours and both during the quiet part of the day."


Round the results. "Between 2,400 and 14,000, most likely around 5,000" is a sentence a manager can carry into another meeting. "4,812.50" is a sentence that gets checked, and nothing good has ever come from being checked to the last decimal. A range also survives being wrong. An outage that costs 6,000 against "most likely 5,000" is a fair estimate. Against "4,812.50" it is a mistake.


The risk statement template


Everything above fits into eight lines. Write one statement per finding, keep it to half a page, and use the same headings every time. That way, a manager who has read one can read the next in thirty seconds. This is the template; the section after it shows one filled in.


RISK STATEMENT

Finding:                 (one plain sentence: what is wrong)
Evidence:                (graph or report name, dates, source system)
Likelihood:              (has happened / trending toward / could happen,
                          with the count, the projected date, or the
                          single point of failure)
Impact:                  (level 1-5 and the words; who stops, what stops)
Cost range:              (low - high, most likely nearer which, with
                          the inputs listed or attached)
Recommendation:          (what you propose, in one sentence)
Cost of recommendation:  (quote, hours, or "none")
Decision needed by:      (date, and what happens after that date)

A filled example for a fictional clinic, Harbourview Clinic, where the appointment system depends on one internet line:


Finding:                 The appointment system needs the internet, and
                         the clinic has one internet line with no backup.
Evidence:                Network map (monitoring console), single WAN
                         link; alert history shows two outages this year,
                         47 min and 2 h 10 min.
Likelihood:              Has happened (twice in the last year).
Impact:                  4, the organisation stops: reception cannot
                         book, check in, or see the day's list.
Cost range:              Example inputs: 12 front-desk and clinical
                         staff, loaded cost 35/h, 1-3 hours idle;
                         roughly 400-1,300 per outage, most likely
                         nearer 400. Missed appointments not costed;
                         reception estimates 5-10 per hour.
Recommendation:          Add a second internet line from a different
                         provider with automatic switch-over.
Cost of recommendation:  Quote from procurement attached; about 6 hours
                         of my time to install and test.
Decision needed by:      End of the month, before the winter clinic
                         schedule starts and appointment volume doubles.

Notice what the example does not do. It does not cost the missed appointments, because nobody gave the author a figure they trusted. It names them and gives reception's estimate instead. Leaving a line honestly uncosted is stronger than filling it with a number you would have to defend.


The risk register


A risk register is the list of risk statements you currently hold open, one row each, reviewed on a schedule. It exists so the manager can see all of them at once and so you can prove, later, that a risk was raised and when. Keep it short. A register with forty rows is a to-do list, and nobody reads a to-do list that belongs to someone else.


IDFinding (plain words)LikelihoodImpactCost range (example)RecommendationOwnerReview
R1One internet line, no backupHas happened (2)4, organisation stops400–1,300 per outageSecond line, different providerIT managerMonthly
R2Shared drive fills in about 11 weeksTrending3, a team stopsNot costed; saving failsAdd storage or archiveITFortnightly
R3Server rack on a single power feedCould happen4, organisation stopsSame as R1 per hourSecond feed; electrician quoteFacilitiesQuarterly
R4Backup job has no success checkCould happen5, outside noticesNot costed; data lossAlert on missing reportITClosed

Three habits keep the register useful. Review it on the date in the last column, not when something breaks. Close rows when the risk is gone and say so, because a closed row is proof the process works. And keep the plain-words column plain; a register full of port names is a register the manager stops opening. The guide on alert fatigue as a risk is a good first entry, because it is the risk that makes all the others harder to see.


The trap of crying wolf


If everything is level four, nothing is. Suppose a manager receives a register where every row says "the organisation stops". They learn that the register is a mood rather than a measurement, and stop reading it. You have a credibility budget with every manager, and each finding you overstate spends some of it.


Reserve the top two impact levels for things that meet the words. "The organisation stops" means most staff cannot work; it does not mean the print server is slow. Say "nobody has noticed" when nobody has noticed. Say "could happen" and not "will happen" when there is no history. Managers do not need every finding to be urgent; they need to be able to tell which ones are.


Report the risks that went away. When the second line is installed and R1 closes, tell the manager in one line. Say that the thing they paid for removed the risk they were told about. It takes ten seconds and it is the only proof most managers ever get that the register was worth reading. A register where things only ever get added is a list of complaints with a column layout.


A war story


The administrator at Bellweather Logistics kept a risk register in which every row was marked critical. That included the reception printer and a complaint about the colour scheme on the wall screen. The finance director read it in full the first month, skimmed it the second, and from the third month onward filed it unopened. Row nineteen was the single power feed to the server rack, marked critical like everything else, with no cost and no date. The feed failed on a Thursday, and the depot stopped shipping for an afternoon. The review found the risk had sat on the register for seven months, between the printer and the colour scheme. The register was described as "thorough."


One row, written to the template above, would have done what nineteen could not. "Could happen; impact 4, the depot stops shipping; between X and Y per hour; a second feed costs a quote from the electrician; decision by the end of the quarter." That is a sentence a finance director can say yes to. "Critical" is not.


The risk-framing checklist


Run each finding through this before it goes on the register or into an email.


  • The risk is stated as one sentence in words before any number appears.
  • The likelihood is named as one of the three kinds, with the count, the projected date, or the single point of failure that proves it.
  • The impact uses the five-level scale, with the level number and the words together.
  • The impact is in the organisation's own units (shipments, appointments, orders, billable hours).
  • Every cost input comes from your organisation and is listed beside the total.
  • No industry statistic appears anywhere.
  • The cost is a range with a stated lean and rounded figures.
  • Uncosted items are named as uncosted, not left out.
  • The statement ends with a recommendation, its cost, and a decision date.
  • The top two impact levels are used only for findings that meet their words.

Where to go next


A risk statement is the third rung of the translation ladder from Speaking Business, Not Packets, written out in full. Once you have a few, The Monthly One-Pager shows where they go on the page a finance director reads. Building a Budget Case From Monitoring Data turns the recommendation line into a funded one. Everything in this series lives under Translating Monitoring Data for Management.


Put this into practice with SysPeep Monitor

Every likelihood in this guide comes from a record. There is alert history for "has happened", trend graphs for "trending toward", and the network map for "could happen". SysPeep Monitor keeps all three in one web console, built from what it discovers and polls over SNMP, WMI, and SSH. It can email the reports behind 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