Section 1
Start Concrete: One Referral, Traced Backwards
Picture a 14-person MSP. A manufacturer with 60 seats signs a three-year managed services agreement. The owner logs it in the CRM as 'referral' and moves on. Trace it backwards and 'referral' dissolves into a specific, boring chain of people. The manufacturer's controller called their outside accountant after a ransomware scare shut the plant floor down for a morning. The accountant did not know IT providers well, but she knew a commercial insurance broker who had spent six months walking clients through cyber liability applications. That broker had watched applications get approved or declined on specific points: multi-factor authentication, endpoint detection, documented and tested backups. He knew which local providers could produce that evidence quickly because he had seen them do it. He named two firms. One answered the phone the same afternoon. None of that chain was luck in any unknowable sense. Every link was a person in a predictable role, reachable a year earlier, who runs into IT failure as a normal part of their work. The MSP that got the call was on the broker's short list because it had spent that year making the broker's job easier. It sent him a one-page control checklist his clients could complete without a technical translator, and it never once asked him for a lead. The other eleven MSPs in that metro were not less competent. They were not on the path. That is a structural difference, not a quality difference, and it decided a three-year contract. Once you see one chain at this resolution, the standard advice to 'ask your clients for referrals' looks like what it is: a request aimed at the wrong end of the pipe.
Section 2
Map the Nodes: Who Actually Transmits, and Who Only Nods
A network model is simpler than it sounds. Nodes are people or firms. Edges are the relationships between them. Centrality is a plain-English idea: how many useful paths run through you, and how short they are. Being known in managed services is decided by centrality far more than by headcount or ad budget. The first job is naming the nodes that actually transmit. A transmitting node has four properties. It sits close to the moment something breaks. It has credibility with the person who signs. It encounters your kind of problem repeatedly, not once. And it has no competing interest in keeping the work. Run real names through that filter and a short list appears. Outside accountants and fractional CFOs hear about downtime as a cost problem. Commercial insurance brokers hear about it as an underwriting problem, and they hear it at renewal, on a calendar. Attorneys hear it during incidents, acquisitions and disputes. Complementary providers near IT without being IT, such as VoIP integrators, cabling firms, AV installers, ERP resellers and compliance consultants, are inside the building when the environment is already exposed. Vendor channel reps see which of their customers are unhappy with their current partner. MSP peer groups matter differently: peers refer work they cannot serve, in the wrong geography, vertical or size band, and they refer it to someone they have met. Now the nodes that mostly do not transmit, despite feeling productive. General networking groups where nobody is near a trigger. Chambers of commerce, unless you hold a specific role. Happy clients, who transmit rarely and only when a peer complains in front of them. The valuable position is not the one with the most contacts. It is the one connecting groups that do not otherwise talk to each other.
Section 3
Measure the Edges: Which Relationships Carry Traffic
Naming nodes is cheap. Measuring edges is where most MSPs stop, and it is the step that converts a vague relationship list into a system. Build a single ledger, a spreadsheet is fine, one row per named source. Track six columns: date of last real conversation, introductions made in the last 12 months, opportunities created, opportunities closed, revenue booked, and what you gave that source in the same period. That last column is the one people skip and the one that predicts the rest. Within a quarter of honest tracking, most owners find the same shape. Two to four sources produce nearly all referred revenue. A long tail of contacts produce nothing but coffee. Some of the tail is genuinely dead. Some is dormant, meaning the relationship exists but the source has no idea what you do or which client fits you, so they never make the connection when the moment arrives. Dormant edges are the cheapest growth available to an MSP, because the relationship cost is already paid and only the specificity is missing. Fixing a dormant edge is not a lunch. It is a sentence the source can repeat: which company size, which industry, which situation, what happens in week one. Direction matters as much as volume. An edge that only carries traffic toward you is not an edge, it is a favour, and favours run out. The sources that keep producing are the ones where you send work, evidence, or plain usefulness back. Our working notes on how this fits the rest of an MSP's acquisition mix sit at https://bizgrowthaxel.com/msp/, including where referral effort belongs relative to search and packaging. The ledger is not for reporting. It is for deciding where the next ten hours go.
Section 4
Timing Beats Reach: The Threshold Rule for Referral Sources
A second lens changes how you rank sources, and it is about timing rather than structure. Managed services buying is a threshold event. A business tolerates a mediocre provider for years, absorbing slow tickets and vague invoices, then flips inside a week after an outage, a breach, a failed audit or a resignation. Dissatisfaction does not accumulate into a purchase. It sits flat and then jumps. That has a direct consequence for referral strategy. A source who hears about problems early is worth more than a source with a bigger list, and it is not close. Fifty contacts who might mention you at some point are worth less than one broker sitting in a cyber liability renewal, because the broker is present at the moment the threshold is crossed. Reach is a vanity measure here. Proximity to the trigger is the real one. So rank sources by when they hear, not by how many people they know. Insurance brokers hear at renewal and after a claim. Accountants hear at year-end, during an audit, and when a fraud or wire issue surfaces. Attorneys hear during incidents and transactions. ERP consultants hear during migrations, which is when environments break. Commercial real estate brokers and office movers hear about relocations, which force an infrastructure decision on a fixed date. A referral arriving three months before the threshold gets a polite reply and dies. The same referral, from the same person, arriving the day the plant floor stopped, closes. This also explains a common frustration. An MSP builds a warm relationship with a source who genuinely likes them, and gets nothing. Usually the source is real and the timing position is wrong. They are never in the room when it breaks.
Section 5
Referral Concentration Is a Balance Sheet Risk
Referrals get treated as a marketing topic. Past a certain point they become a financial one, and MSP owners notice late. If most new logos over the last two years trace back to one broker, one vendor rep or one accounting firm, you do not have a referral network. You have a supplier, and no contract with them. That is a single point of failure outside your control. The person retires. They move firms and the new firm has a preferred provider. Their practice gets rolled up by a buyer who standardises vendors nationally. Any one of those ends a channel in a quarter, with no notice period. The comparison operators already understand is client concentration. Nobody argues that one client at 40 percent of revenue is fine. Source concentration is the same risk one step upstream, and worse in one respect: it is usually invisible, because nothing in the P&L labels it. When an MSP is sold, buyers ask where new business comes from for exactly this reason. A pipeline that depends on relationships held personally by the departing owner is worth less than one that does not, because the acquirer cannot buy the relationship. That is not a valuation formula, it is a diligence question, and it gets asked. Two practical guardrails. Set a ceiling, say no single source above a third of referred revenue, and treat crossing it as a trigger to build a second and third path rather than as good news. Second, move relationships off the owner. If the owner is the only node, the firm has no network. The owner has one, and it walks out with them. Introduce a second person into every important relationship before you need to.
Section 6
The Cost of Position: Slow, Compounding, and Not a Campaign
Here is the part most articles on this topic leave out. Building network position is expensive in the currency MSP owners have least of, and it does not respond to urgency. Expect a senior person, usually the owner, to spend 4 to 8 hours a month on it, and 12 to 24 months before it shows up as booked revenue rather than pleasant conversations. There is real cash cost too. Peer group dues plus travel are a meaningful annual line. Co-marketing with a complementary provider costs time and often a shared event budget. Vendor partner tiers, one of the more reliable routes to becoming central, generally require revenue commitments, certification headcount, or both. Be clear-eyed about where the vendor makes money there. Partner programs exist to buy distribution. Market development funds come with strings about how and where you promote, and a higher tier usually means you have agreed to concentrate purchasing with that vendor. That can be a fair trade. It is still a trade, and it deepens exactly the dependency described above. The cost is worth carrying because it compounds in a way paid demand does not. Every referral you close well gives the source direct evidence that sending you work makes them look good to their own client. That raises the probability of the next introduction, which produces more evidence. Ads do not do this. Ads are rented, and the moment you stop paying the position disappears. Network position, once real, keeps producing for years on maintenance rather than spend. The practical consequence is a sequencing rule. Do not start a referral push the quarter you need pipeline. Start it while another channel carries you, and judge it on a multi-quarter trend rather than any single month.
Section 7
Where This Breaks: Channel Rewiring, and the Thing the Model Cannot See
Two honest limits, and the second matters more. The first is a structure break. Vendor channel consolidation can rewire who is central overnight. Distributors merge. A vendor restructures its partner program and the tier you built around disappears. Your key channel rep gets reassigned or leaves. The complementary provider whose introductions you rely on is acquired by a national firm that already owns an IT division and has no reason to refer to you. None of it is gradual or announces itself in advance. An MSP whose position depended on one relationship inside one channel can be structurally stranded while its service quality is unchanged. The defence is not loyalty but diversity of network type: hold real position in at least two structurally different networks, one vendor channel and one professional services cluster, so a rewiring in one does not take the whole pipeline. The second limit is the honest one, and it applies to everything above. This model works at the level of structure. It tells you where to stand, which relationships are worth ten hours, and which are decoration. It cannot see the single largest driver of any individual deal, which is one person deciding to stake their own reputation on you in a specific conversation. Whether the broker says 'call these two' or 'call this one, I trust him' is idiosyncratic, personal, and unmodellable at this resolution. Two MSPs can occupy identical network positions and get different outcomes for reasons no framework captures. So use the network view as an allocator of effort, not a forecast. It should change where your time goes. It should not make you confident about any particular deal, and it should make you judge results over four quarters rather than one.
Section 8
The Fitness Test
You are ready to build a referral system if you can name the last ten referred deals and trace each one to a specific person, if you have at least one service line you can describe in a sentence a non-technical accountant could repeat accurately, if a senior person can genuinely protect 4 to 8 hours a month for the next four quarters, and if you have another channel carrying enough pipeline that you are not depending on this one to make payroll next quarter. You are also ready if you already have three or more sources producing occasionally and never made the pattern explicit. That is the cheapest version of this work. You are not ready if you need booked revenue inside 90 days, because network position does not move that fast and you will quit before it does. You are not ready if you cannot say which client is a good fit in plain language, because a source cannot transmit a description you have not written. You are not ready if the owner is the only person with external relationships and has no intention of changing that, because you will build an asset that cannot be sold. And you are not ready if your onboarding is inconsistent, because referrals amplify whatever experience you actually deliver. A source who sends you one client that goes badly stops sending clients, and that damage runs backwards through the network faster than any good result runs forward. Fix delivery first, then earn position. If you fail the test on timing alone, do not abandon the idea. Start the ledger now, at low intensity, so the compounding has already begun when you can afford to spend properly.