Section 1
Nobody wakes up wanting managed IT
Take a 34-person civil engineering firm with two offices. The file server is seven years old. IT support is the owner's nephew plus a break-fix shop on call, roughly $1,800 in a typical month, and it works fine most weeks. Nobody there is shopping. Call them in March and you will be told they are all set, and they will mean it. In September the file server dies on a Thursday afternoon. Backups exist, but nobody has tested a restore in two years. It takes four days to get drawings back. Two client deadlines slip. On the following Monday the owner tells his controller to find three companies that do this properly. Nothing about that firm's technical situation changed between March and September. The server was already old in March. The backup was already untested. What changed is that accumulated pain crossed a private tolerance level. Mark Granovetter's threshold model describes exactly this: each actor holds a position, absorbs pressure without visible change, and then moves quickly once a personal threshold is crossed. Three practical consequences follow for an MSP. First, most of your addressable market is unreachable at any given moment, not because your message is weak but because their threshold has not been crossed. Second, the window between crossing and signing is short, often two to five weeks, and the shortlist forms inside the first 48 hours of it. Third, that shortlist is assembled from memory and from whoever the controller's trusted contacts name, not from a careful research process. So your marketing is not persuading anybody. It is loading a shortlist that gets read months later, under time pressure, by a person who cannot evaluate your stack and will not try. Spec sheets and vendor logos are written for a reader who does not exist.
Section 2
Before the trigger or at the trigger: the only channel question that matters
Once you accept that demand is failure-triggered rather than planned, channel choice stops being a matter of taste. There are only two jobs. Memory channels reach businesses before their threshold is crossed. Their job is to be recalled later. Intent channels reach businesses after the threshold is crossed, while they are actively looking. Their job is to be found and chosen now. Run rough arithmetic on your own metro. Say there are 900 businesses of the right size and type in reach. Suppose 6 percent cross a threshold in a given year and seriously consider changing IT providers. That is 54 live opportunities across the entire market, spread across 12 months. Four or five a month, contested by every MSP in town plus the two national players. Paid search, review profiles and "IT support near me" visibility are all fighting over those four or five. That is why MSP paid search costs what it costs, and why the volume never scales the way an agency deck promises. The other 846 businesses are not opportunities this year. They are memory positions. A monthly plain-English security note, a talk at a trade association, a working relationship with their accountant, and a review profile they will read in nine months are all bids for a slot on a shortlist that does not exist yet. The mistake that wastes the most MSP money is judging memory channels with intent metrics. An owner runs a newsletter for four months, books zero calls, and kills it. That newsletter was never going to book a call in month four. It was going to be the reason a name gets recalled in month fourteen. The reverse error costs almost as much: judging paid search on volume. It is supposed to be small, expensive, and close at a high rate.
Section 3
MSP lead channels compared: cost, lag, and failure mode
Every channel below gets three honest numbers: cost, lag, and failure mode. The failure mode matters most, because that is what quietly drains a budget for a year before anyone notices. Client referrals and reviews. Cost is near zero in cash and real in effort. Lag is 3 to 12 months from a systematic ask to a closed deal. Failure mode: referrals happen when clients talk about you, and a happy client under a well-run contract has nothing to say. Silence is the failure mode of good work. Professional referral partners. Accountants, insurance brokers, commercial realtors, line-of-business software resellers. Cost is time and reciprocity. Lag is 2 to 8 months to a first introduction, longer to a steady flow. Failure mode: one-sided relationships. Send nothing back and you get one referral, then silence. Local search and content. Cost is consistent publishing in-house, or roughly $1,500 to $4,000 a month outsourced. Lag is 6 to 12 months before meaningful movement. Failure mode: ranking for terms nobody types during a crisis, and being compared against three competitors on a listicle page you do not control. Paid search. Cost per click on commercial IT support terms is among the highest in local B2B. Lag is immediate. Failure mode: thin volume, lead aggregators bidding against you, and clicks from residential and one-off repair requests that eat sales time. Outbound. Cold calls, email, LinkedIn. Cost is a person, part time or full. Lag is 2 to 4 months. Failure mode: it reaches the 94 percent who have not crossed a threshold, and repeat contact with people who are not ready burns the list and the caller together. Events. Chamber groups, trade associations, vertical conferences. Cost is $2,000 to $8,000 per event plus days of your time. Lag is 6 to 18 months. Failure mode: attending IT industry events, where the room is peers and vendors, not buyers. Vendor channel. Cost is certification and partner-tier effort. Lag is long. Failure mode: being one of forty partners on a list the vendor controls. Usually a waste for an MSP under about fifteen staff: broad paid social, low-volume display retargeting, booths at IT industry conferences, and untargeted cold email.
Section 4
Referral position is a network, not luck
Most MSP owners treat referrals as weather. Good months, bad months, nothing to be done. That is a modelling failure. Referrals are a network problem, and networks have structure you can see and change. Stop thinking about referrers as people who like you and start thinking about them as transmitting nodes: contacts who sit close to the moment a threshold gets crossed and who are asked for a name. Score every candidate on three things. How many businesses in your target profile do they touch in a year? How likely are they to be in the room when the trigger fires? When asked, do they name one provider or hand over three? Score the obvious nodes honestly. An accountant touches 80 to 300 small businesses and hears about the fraud attempt, the failed payment, and the office move. A commercial insurance broker handles the cyber renewal and the questionnaire nobody can complete, which is a threshold event manufactured on a fixed annual schedule. A commercial realtor knows about every office move six months early, and a move is the cleanest MSP trigger there is. A line-of-business software reseller in your vertical is present during migrations. A peer CEO roundtable is where the complaining happens out loud. A general networking group scores badly on all three questions. Twenty weak ties across twenty industries produce far less than three reciprocal relationships with high-centrality nodes. Then do the unglamorous part: give first, give specifically, and give repeatedly. A broker who receives two qualified introductions from you will name you, singular, at renewal season. A broker you took to coffee once will say "I know a few IT companies." The mechanics of turning that into a repeatable system are covered in [the MSP referral system that is not luck](https://bizgrowthaxel.com/blog/msp-referral-system-that-is-not-luck/).
Section 5
Sell outcomes, package clearly, price predictably
A shortlist position gets you into the room. What loses the deal from there is almost never technical. It is fog. The buyer in a threshold moment is scared and rushed. They have just learned their IT situation was worse than they thought, so they now distrust their own judgment on the subject. Handed three proposals with different scopes, different pricing units and different exclusions, they cannot compare them. So they default to the cheapest, or to whoever their accountant named, or to doing nothing for another quarter. Package against that. Two or three named tiers, priced per seat or per site, described by outcomes rather than components: response time commitments, after-hours coverage, what happens during an incident, what the security floor includes, how often somebody senior reviews the roadmap. Put the technology inside the tier, not on the front of it. Publish indicative pricing, even as a range. It costs a few tyre-kicker calls and buys credibility with every buyer who has been burned by open-ended invoicing, which is most of them. The entry offer that works in a threshold market is a substantive paid assessment ending in a written findings memo a non-technical owner can read. Free network scans are read as a sales pretext because they usually are. A paid assessment filters, demonstrates competence in the only way that matters, and produces the document that makes the migration conversation obvious. One warning on winning fast. Threshold buyers sign in a hurry and judge in a hurry too. A deal closed on urgency with vague scope becomes a client who is unhappy inside the first fortnight, which is a different and more expensive problem, covered in [why MSP clients leave in a week](https://bizgrowthaxel.com/blog/why-msp-clients-leave-in-a-week/). Clarity at proposal stage is retention work disguised as sales work.
Section 6
From break-fix referrals to a predictable pipeline
Most MSPs grow to somewhere between $600k and $1.5m on heroic referrals and then flatten. The flattening is structural. Referral volume is capped by client count, and client count is capped by referral volume. Breaking that loop takes four unexciting things. Capture. Every inquiry, assessment, association contact and partner introduction lands in one system, not the owner's inbox. If the record of a conversation exists only in somebody's head, it does not exist. Threshold demand is unpredictable in timing, so the only defence is that nothing gets lost while you wait. Speed. A buyer inside the window is comparing two or three providers within days. Same-day response and a same-week assessment slot win deals on responsiveness alone. Responsiveness is literally the product you are selling, so the sales process is a live demonstration of it. A three-day reply loses on message, not just on timing. Nurture. This is where the memory channel lives. A monthly note in plain English, no acronyms, no fear-selling, sent to every prospect who ever spoke to you and never bought. Most will not reply for a year. In a threshold market that is fine. You are not trying to move them. You are trying to be the name available when something else moves them. Measurement. Track what the model implies rather than the vanity numbers. Time to first response. Share of new deals arriving through named partner nodes. How many prospects have had at least six touches. Assessment-to-proposal rate. Monthly lead volume is noise in a market with four or five live opportunities a month, and chasing it makes you change tactics for reasons that are pure randomness. If you want that machinery specified for an MSP rather than a generic service business, [our MSP growth work](https://bizgrowthaxel.com/msp/) is built around exactly this sequence.
Section 7
When many thresholds move at once, and what this model cannot see
The threshold model assumes tolerance levels move slowly and independently. Sometimes they do not, and when that happens the channel economics described above invert for a few months. Three things move a lot of thresholds at once. A metro-wide or sector-wide breach, where one compromised local supplier puts a hundred businesses on alert in the same week. An insurer or regulator mandate, where cyber renewal suddenly requires controls most local businesses do not have, creating a dated and unavoidable trigger across a whole client base. And a supply shock on the provider side, where a large local MSP is acquired, cuts service quality, or fails, releasing dozens of unhappy clients at once. In those windows the arithmetic changes. Search volume for commercial IT terms spikes, so paid search becomes cheap relative to available volume rather than expensive. Outbound response rates jump because you are contacting people who have just crossed a threshold. Memory channels pay off all at once, because every position you loaded over two years gets read in the same fortnight. Shift budget toward intent channels hard and temporarily, and have assessment capacity ready before the window opens rather than after. These windows last weeks, not quarters. Now the honest limit. This whole model is population-level. It tells you where demand comes from, how it is timed, and which channels do which job. It cannot see individual relationship dynamics, and that is where a meaningful share of MSP deals is actually decided. One vouch from a CFO who worked with you six years ago at a different company will beat your entire channel mix, will not appear in any attribution report, and cannot be manufactured on a schedule. Build the system because the system is what you control. Do not mistake it for the whole picture.
Section 8
The fitness test: is this the right build for you right now?
This approach is not universally correct. It is correct under conditions. Check yourself against them honestly before spending anything. You are ready for this if you can already deliver on the promise. If your current clients would not describe your response times as reliable, more leads is just churn bought at retail price. You are ready if you have a defined target profile: a size band, a geography, ideally one or two verticals. Threshold marketing depends on being memorable to a specific group, and you cannot hold a memory position in the mind of "any business." You are ready if you can commit 12 to 18 months to the memory channels without killing them at month four. And you are ready if somebody, even part time, owns follow-up, because the whole system rests on nothing being dropped. You are not ready if your delivery is currently under water. Fix the service before the pipeline. You are not ready if the next 90 days depend on this working, because nothing here produces meaningful volume in 90 days. If you need revenue this quarter, work your existing client base for expansion and go directly at the two or three partner nodes closest to trigger moments. That is the only fast path in this model, and it is fast because it borrows somebody else's memory position instead of building your own. And you are not ready if you expect clean attribution. When the trigger arrives months after the impression, most of what works will never be traceable, and demanding traceability steers you toward the channels that are easiest to measure rather than the ones that build the position.