The Sourcing Decision Worksheet

The meeting is at 14:00. The finance director has a proposal from an outside provider and a feeling about it. The operations manager has a different feeling, formed during the last provider. The IT administrator has the numbers, in the sense that she could produce them if anyone asked. She has a feeling she is trying not to have, because the proposal is partly about her job. Nobody has a worksheet. The decision will therefore be made by whoever has the strongest feelings and the latest slot on the agenda. It will be revisited in eighteen months by different people with different feelings.

This is the last article in our Sourcing Support series, and it is the one to bring to that meeting. The worksheet does not replace judgment. It makes judgment show its working, on five axes that can each be scored from one to three. Each score answers a question whose answer is a fact rather than a mood. The three models are defined in The Three Sourcing Models, Plainly. The five functions that stay in-house whatever the score are in What to Keep In-House No Matter What.

It covers what the worksheet decides and what it refuses to, then the five axes and how to score each. Next come the worksheet itself, a guide to reading the total, and three worked examples scored and read. Then it covers what a worksheet cannot tell you and when to score it again. A war story finishes the article: an organization scored it twice and got two different answers, both right.

What the Worksheet Decides, and What It Refuses To

A managed service provider, or MSP, is an outside company that runs some or all of your IT for a recurring fee. That fee is usually shaped as a monthly amount per user or per device. The worksheet decides one thing: which of the three sourcing models fits the shape of your organization as it is now. Those models are in-house, outsourced to an MSP, or co-managed between the two. It answers in a range rather than a verdict. When the answer is co-managed, it also says where to draw the split. The axes that pushed the score are the functions to buy.

It refuses to decide three things, on purpose. It will not tell you whether a particular MSP is any good. It will not tell you whether your in-house team is any good. And it will not tell you the price. The price is a number the provider gives you after the worksheet has said what to ask for. Asking before you know is how organizations end up buying round-the-clock cover for a desk that closes at 18:00. The sheet is for the shape. The judgment is for the people.

Score it with the budget holder, the operations lead and the IT lead in the room, and the ticket numbers printed out. Each axis takes about five minutes to argue about, which is the correct amount. An axis that takes forty minutes is one where nobody has the fact.

The Five Axes

Five things decide the fit, and each gets one axis. Volume is how much support work there is: tickets per month, and whether they would keep a person busy. Hours of coverage is when the work has to be done: the working day, or evenings and weekends, or around the clock. Skills on hand is whether the people who can do the work exist on your payroll and can be kept. Control and regulatory needs is whether a regulator, a customer contract or an auditor requires named employees and records you hold. Budget shape is whether the organization can carry salaried headcount or needs a monthly fee that flexes.

Each axis scores from one to three in the same direction. A one leans in-house, a three leans outsourced, and a two is the territory where either works and co-managed usually works best. Five axes at up to three points each gives a total from five to fifteen, and the reading guide turns the total into a model.

Why these five and not fifteen? Consider the other factors people bring to the meeting: lock-in, the knowledge that leaves with the contract, and whether the current administrator is tired. Every other factor is one of these five in disguise or a question about a specific provider. A worksheet with fifteen axes has fifteen places to put a thumb on the scale.

Scoring the Axes

Volume

The question that decides the volume score is: would the support work, on its own, keep at least one person busy full time every week of the year? If yes, comfortably, score one, because steady volume is cheaper done by your own people: a fee per user scales up with it and a salary does not. If roughly, with slack weeks, score two. If no, score three, because a desk that fills half a day cannot justify a person, and an MSP spreads that half day across its other customers. Get the number from twelve months of ticket history, and count the walk-ups and calls that never became tickets as generously as honesty allows.

If you already run a help desk tool, its reports should give you the count without arithmetic. QuerySolv's reports, for instance, show demand over seven, thirty or ninety days with a busiest-hours heat map, which happens to answer the next axis too. Whatever the tool, use the real count. Memory rounds busy months up and quiet ones away.

Hours of Coverage

The question is: when does a support problem have to be worked on by a person, and what does waiting cost? Score one if the working day in one time zone covers it, with an occasional evening nobody minds. Score two if the business runs outside the working day often enough that waiting until morning sometimes costs money. Think of a second shift, a weekend operation, a site three time zones away. Score three if someone must answer at 03:00 and the answer must be a fix, not a message. Be strict about the word must. Most organizations that believe they need round-the-clock cover need a phone that is answered and a problem that can wait until morning, which is much cheaper.

Skills on Hand

The question is: do the people who can do this work exist on the payroll now, and can you keep them and cover them? Score one if yes on both counts, including cover for leave. Score two if most of the work is covered and one or two domains, such as the network or the backup system, depend on one person or on nobody. Score three if the skills do not exist in-house and hiring them is not realistic. Skills are the axis most often scored by pride rather than by fact. Write the domains down and put a name beside each; a domain with no name is a three for that domain, whatever the overall score.

Control and Regulatory Needs

The question is: must named employees make the identity and security decisions, with the records staying in your hands? Does a regulator, a customer contract, an auditor or your own risk appetite require it? Score one if yes, in writing, from someone who can enforce it. Score two if there are obligations a contract clause can meet, which is where most organizations sit: confidentiality, data location, a right to audit. Score three if nothing beyond ordinary good practice applies. A one here pins certain functions in-house whatever the total says; the reading guide treats it as an override rather than a vote.

Budget Shape

The question is: can the organization carry salaried headcount, with the recruitment, cover and training that come with it? Or does it need a monthly fee that flexes with headcount and can be stopped? Score one if headcount is normal here and the organization would rather own than rent. Score two if either works, or if it depends on the year. Score three if a permanent post is close to impossible and a flexing fee is the only shape finance will sign. This axis is about the shape of the money, not the amount. The amount comes later, from the provider, for the functions the other four axes said to buy.

The Worksheet

Score every axis, add them up, then read the total against the guide in the next section. Print it; a worksheet on a screen gets edited after the meeting by whoever has the login.

Reading the Total

Five to seven points says in-house. The work is steady, the hours are ordinary, the people exist or can be hired, and the organization can carry them. Buy nothing but cover: a named contact for leave and emergencies on a small retainer, which is the lightest form of co-managed there is and does not deserve the name.

Eight to eleven points says co-managed, and the split is drawn by looking at which axes scored three. A three on hours means buy the hours: after-hours and weekend cover, with your desk keeping the day. A three on skills means buy the domain: the network, the servers, the backup, whichever had no name beside it. The design of the seam, whichever way it is drawn, is in Designing a Co-Managed Support Desk.

Twelve to fifteen says outsourced, with the five keep-in-house functions kept anyway. These are identity and access decisions, security decisions, vendor relationships, the knowledge of how the business works, and the records, including the ticket history. Someone on your payroll owns those even when nobody on your payroll performs them. A fully outsourced desk with those five kept is a sound model. Without them it is a model with a very expensive exit.

Two overrides sit above the total. The first is the control override already mentioned. The second is the mixed score. A sheet with a one and a three on it is telling you about a split, not an average. A nine made of three threes and two ones is a co-managed desk with a sharp seam, not a mild preference. Read the ones and threes before the sum. And whatever the total, keep the ticket record in your hands, in a queue you can export and that survives a change of model. A self-hosted desk whose history sits on your own server does that by construction. QuerySolv is one, and also our product, so discount our enthusiasm and check the reasoning on its own.

Worked Example One: Forty People and One Admin

Kestrel Payroll has forty staff on one site, one administrator, Dana, and a proposal from an MSP to "become its IT department". Volume: about ninety tickets a month, which is half of Dana's week, the other half going on projects and the audit; score two. Hours: the office runs 08:30 to 17:30 and the payroll run is a daytime job; score one. Skills: Dana covers the desktop estate, the network and the two servers. A contractor who knows the payroll application is on call. Every domain has a name; score one. Control: the firm processes payroll data for its clients under terms that require named staff and in-house records; score one. Budget: Dana is already a salaried post and a second is thinkable in two years; score two.

Total: seven. The reading is in-house. What the sheet does flag, between the two on volume and the one on skills, is the thing it is not designed to flag. Every domain has a name, and it is the same name. That is a cover problem, not a sourcing problem, and it is solved with a named contact for leave and a runbook, not a contract. The administrator who is a department has a pillar of her own at The One-Person IT Department.

Worked Example Two: Two Sites, Three Hundred Users, One Regulator

Bluewater Bank has three hundred users across a head office and an operations center ninety minutes away, and a regulator to answer to. Volume: about five hundred and twenty tickets a month, which keeps two people busy with room for a third; score one. Hours: the operations center opens at 07:00 and the trading desk is active until 20:00, and a failure at either end costs money that afternoon; score two. Skills: the desk is staffed and the servers have a named owner. But the network at the operations center depends on one person, Marcus. The backup system depends on the vendor's documentation; score two. Control: the regulator requires named employees for identity and security decisions and records held by the bank; score one. Budget: headcount is normal, and so is a fee; score two.

Total: eight. The reading is co-managed at the in-house end, and the twos say what to buy. Buy early-morning and evening monitoring and response, and a second pair of hands on the branch network and the backup system. The one on control pins identity, incident declaration and the records in-house, which is where they were going anyway. Bluewater keeps the desk, keeps the decisions, and buys the edges of the day. The MSP's engineers work as scoped agents in Bluewater's queue, with Marcus as owner of record for the network, which also fixes the cover problem.

Worked Example Three: The Retailer Whose Quarter Triples

Acme is a retailer with a hundred and twenty staff for three quarters of the year. It has two hundred in the peak quarter, when the stores extend their hours. Score it twice, trough and peak, because any axis that moves by two between seasons is a seam. Trough: volume about a hundred and forty tickets a month, which would not fill a week, score three. Hours: store hours six days a week, score two. Skills: one administrator, Tomas, with the point-of-sale system supported by its vendor, score two. Control: card payments handled by a processor, score two. Budget: either works, score two. Trough total: eleven.

Peak: volume about four hundred and fifty a month, with seasonal staff to set up, score one. Hours: evenings and Sundays with a warehouse night shift, score three. Skills and control unchanged at two. Budget: finance wants the extra cost to stop when the quarter does, score three. Peak total: twelve.

The two totals straddle the line between co-managed and outsourced, and that is the reading. A total that crosses the line for one quarter is not a reason to outsource the year. It is a reason to buy the difference. Acme keeps Tomas as the desk and the front door all year. It buys after-hours and overflow cover for the peak quarter under a co-managed contract with a seasonal clause. The MSP's people work in Acme's queue so that the peak's tickets stay at Acme when the quarter ends. The trough's three on volume is the other thing to notice: half of Tomas's job is not tickets, and the worksheet only scores the half that is.

What the Worksheet Cannot Tell You

The worksheet cannot tell you whether a particular MSP is any good. A provider's quality is in the engineer who answers, the SLA that measures a human response rather than a portal timestamp, and the knowledge they hand back. The way to find out is references from organizations your size, a short paid pilot on one domain, and the questions in Working With an MSP Without Losing the Plot. The worksheet tells you what to ask for, not whom to ask.

It cannot score culture fit. Will your users accept help from people who are not in the building, and will your managers stop walking up to the desk? A fully outsourced desk at an organization where every request arrives by shoulder-tap will score well on the sheet and badly in the first month. The sheet does not know about the shoulders. Ask the question out loud, and write the answer under the total.

It cannot tell you about the person scoring it. The administrator scoring the sheet about her own job will, in my experience, score it honestly under two conditions. The sheet must be in the room. The decision must be visibly about the shape of the work rather than about her. That is the strongest argument for using it at all. A worksheet makes the decision about the organization, and takes it off the person who has to live with it.

Revisiting the Decision

Score the sheet again once a year, and sooner when any axis moves by one, because the shape of an organization changes faster than its contracts. Volume moves with headcount and with the systems you adopt. Hours move when a shift is added or dropped or a site opens. Skills move every time someone is hired or leaves. Control moves with a new customer or regulator. Budget moves with the annual plan. Keep the scored sheets with the contract, dated, so that the next meeting starts from the last answer rather than from feelings.

Revisiting is not the same as switching. A total that drifts from seven to eight does not mean cancel anything. It means look at which axis moved and whether a small purchase, or a small hire, fixes it. The expensive mistakes are made on a total nobody has looked at for three years. The meeting is called by whoever has the strongest feelings, which is where this article came in.

A Short War Story: Northgate Retail Scored It Twice

Northgate Retail scored the worksheet before signing a fully outsourced contract. Volume was one. Hours were three because the distribution center ran an overnight shift. Skills were three because the last administrator had left and nobody had been hired. Control was two, budget three: a total of twelve, and the contract matched it. A year later, at the renewal, Sam, hired as IT lead in the meantime, scored it again with the same people. The hours axis had moved to two. The overnight shift had become an early one and nothing now had to be fixed at 03:00. Skills had moved to one, because the contract's knowledge clause had put two technicians through a year of working the estate. Budget had moved to two; the total was eight. Co-managed after-hours cover replaced the fully outsourced desk at the renewal. Northgate's own desk covered the daytime, with the MSP's engineers as scoped agents in Northgate's queue for evenings and weekends. The MSP agreed, because the sheet was on the table and the arithmetic was not an opinion. Both scores had been right. The organization had changed under the contract, which organizations do.

I have sat in the meeting with no worksheet, and I have sat in the one with it. The second one was shorter, and its decision was still in place two years later, which the first one's was not. Bring the sheet. The feelings will still be there, and that is fine; they just have to show their working.

Frequently Asked Questions