Don’t trust the blended margin. Separate revenue into its real service lines, re-cost labour at fully-loaded rates, and reconcile the PSA, billing and accounting systems against each other, then build an EBITDA bridge where every adjustment traces to a source record. The recurring book, priced on its own economics, is what you’re actually buying. Below is how that works, and how Kautilya applied it to a real $21M MSP under LOI.
Two analysts, ~300 hours, roughly a month — the team behind this engagement. Meet the rest of Kautilya.
A managed service provider almost never makes money the way its P&L suggests. A single blended gross margin averages together service lines that have nothing in common, so the headline number describes none of them. Doing diligence properly means taking that blend apart and rebuilding it from source data until you can see which customers, which contracts and which hours actually carry the business.
That is the general method. What it looks like in practice is best shown through a real engagement. A buyer came to Kautilya under LOI on a fifteen-year-old California MSP, about $21M of revenue across three years, ~90% net retention, and a nagging problem: the blended margin was only ~32%, net profit was thin, and the seller’s P&L couldn’t say why. It could give one blended number; it couldn’t say which customers made money or where the margin lived. The buyer’s investors kept asking. So the method below isn’t hypothetical, it’s what the team actually did.
Step 1, Separate the blend into real service lines
The first move is to stop treating the business as one thing. Split the P&L into its true components and model each on its own, because each line carries a different margin and deserves a different multiple. Buying on the blend means overpaying for the weak lines and underpaying for the strong one.
On this deal, that split is where the whole picture changed.
Under the single ~32% blend sat five distinct businesses. Rebuilt across three years: recurring MSP work at $10.1M and 44.5% margin, resale at $8.7M and 18.3%, contracted services at $1.3M and 14.5%, coaching at $0.9M and 46.2%, and one-time project work at $0.15M and 83.6%. To get there, Kautilya modelled each of the ~300 agreements individually, revenue from invoices, labour from re-costed hours, unit cost from product records, down to a margin per contract. The blend described none of them.
Step 2, Re-cost labour at fully-loaded rates
The seller’s time-tracking export understates the true cost to serve. It shows billable hours at headline wages and hides the benefits, PTO and payroll taxes on them, plus the non-billable time spent winning and supporting the work. Load all of it back in before you believe any margin, especially on the recurring book, where an unloaded number flatters badly.
This is a place where most diligence takes the flattering number. Kautilya didn’t.
A naive allocation flattered the recurring book to 62%. The team refused it and built it down, charging the recurring book its fair share of non-billable labour (about 10 points) and loading wages for benefits, PTO and payroll taxes (about 7 more), landing a defensible ~45%. That independently came within a point of the seller’s own internal figure of 43.8%. It rested on turning 51,063 time entries into fully-loaded cost per agreement and role.
The discipline generalises: a margin built the conservative way, one that survives a lender and doesn’t unwind after close, is worth more to a buyer than a generous one that does.
Step 3, Reconcile the PSA, billing and accounting systems
An MSP runs on a PSA system, a billing or query layer, and an accounting ledger, and they rarely agree at face value. Decision-grade economics start with tying them together so nothing in the model rests on an unchecked number, the standard a lender will hold you to.
Kautilya extracted ConnectWise data (the PSA of record for tickets, agreements, time and invoices) through Metabase queries, validated it back against ConnectWise, then tied it to the Intuit books across accrual and cash. The rebuild rested on 12,267 invoice detail lines, 51,063 time entries, and 8,924 GL transactions, each reconciled to source.
The hard part here is one an off-the-shelf approach misses entirely, and it’s worth understanding if you’re doing this yourself. Joining labour to revenue meant working across two schemas: recurring work keyed cleanly on one identifier and joined directly to time entries, but project labour carried no matching key. The team built a company-account catch-all to attribute that project cost back. No standard join would have caught it, and missing it would have mis-stated the margin on a whole service line.
Step 4, Build an EBITDA bridge a lender will accept
The output is a normalized EBITDA bridge where every adjustment is independently toggleable and traces to a source record, so a buyer and their lender can re-run it scenario by scenario. Done properly, it turns a vague worry into a quantified, fixable line, which is exactly what happened here.
The investors’ question, why is profit thin on $7M, had a concrete answer once labour was fully reconstructed: about $3.2M of labour across three years sat outside customer delivery, internal admin, sales and management time, including 19,140 hours on one internal agreement. That was the bulk of the EBITDA gap and the clearest normalization lever in the deal, reframing a thin-looking business as a profitable core carrying overhead built for a larger revenue base. On its own ~45% economics, the recurring book was worth roughly $4.5M of gross profit over three years.
What the rebuild let the buyer do
This is the payoff of doing diligence at this depth rather than running a checklist. For the first time the buyer could see the business the way an operator does, by customer, by service line, by hour. The 32% blend resolved into its parts: low-margin resale pulling the average down, a coaching line in decline, and a recurring book near 45% once labour was loaded.
That changed what they could do. They stopped underwriting a blended fiction and started underwriting the recurring book that actually drives the value. They could answer their investors with evidence. And because every number traced to a source they and their lender could re-check, they could move with conviction, carrying a re-rated view and a plan to grow the margin, not just the revenue, into their financing conversations. That level of reconstruction is the kind of attention most firms reserve for deals ten times the size, which is precisely why the small-to-mid-market deals where the real economics are buried in the software are the ones Kautilya is built for.
The same source-level discipline applies whether the target runs on ConnectWise or on affiliate and payment-processor data, see how it plays out on a $1.8M digital wellness portfolio.
Frequently asked
What is a good gross margin for an MSP?
A fully-loaded recurring managed-services book typically runs in the mid-40s percent once labour carries benefits and overhead; resale and project work run far lower. A single blended figure usually overstates resale and understates recurring, so separate them by service line before benchmarking, which is exactly what surfaced the real picture on this engagement.
How long does buy-side diligence on an MSP take?
A full source-level rebuild, reconciling the PSA, billing and ledger and re-costing labour to the agreement level, is a matter of weeks with a dedicated team. This one ran roughly a month, about 300 analyst-hours across two analysts.
What are the biggest red flags in an MSP’s numbers?
A single blended margin with no service-line breakdown, labour that doesn’t reconcile to payroll once fully loaded, and revenue that doesn’t tie between the PSA and the accounting system. On this deal all three pointed the same way, the numbers hadn’t been stress-tested at source, and the real story only appeared once they were.