Scale and concentration risk

Portfolio Economics: Why Microsoft's Math Isn't Yours

Microsoft's partner base absorbs churn by sheer scale. A sub-$1B ISV's real working partner list is short enough to count, and losing three names off it is not the same math as losing three out of five hundred thousand.

Two different math problems

What Microsoft's scale is

What pruning looks like at that scale

The sub-$1B math

Depth over breadth

Before you find out the hard way

A line has been circulating in partner-strategy commentary this year: Microsoft, and firms of similar scale, can absorb a bad partner relationship, or a dozen of them, without the business missing a beat. A company running below $1B in revenue, the argument goes, cannot copy that same tolerance for churn. The instinct behind that argument is correct. The way it usually gets stated is not specific enough for anyone to act on.

What is missing is which number applies. Losing three partners means one thing measured against Microsoft's own base and something else entirely measured against the partner list a smaller company depends on this quarter. Both are real numbers. They are not the same calculation, and a sub-$1B ISV that treats them as one is doing someone else's math with its own money.

Two different math problems

The question sounds like one number. It is two.

Microsoft's partner base now runs past 500,000 firms. Losing a handful of them costs the company nothing anyone would notice on an earnings call. A sub-$1B ISV's working partner list, the names its go-to-market motion depends on this quarter, is usually short enough to count on two hands and change. Losing the same handful of names does not cost that company nothing. It costs a fifth of the whole thing.

Those are two different calculations, not two versions of the same statistic. The first is a fraction of a number too large to feel: three firms out of five hundred thousand is roughly 0.0006 percent, a rounding error inside the portfolio. The second is a fraction of a number small enough to name individually: three partners out of a working list of fifteen is twenty percent, the kind of swing that shows up in a board deck.

Treating these as the same math, because both involve the phrase "lost three partners," is exactly where the enterprise playbook stops applying to a company this size.

0.0006%
vs 20%

The same "lost three partners." Two entirely different numbers, depending on whose portfolio it comes out of.

Microsoft's scale absorbs the first without comment. A sub-$1B ISV's go-to-market motion does not absorb the second the same way, and assuming it does is the risk.

What Microsoft's scale is

The number that makes "sign everyone, prune later" rational

Microsoft closed fiscal year 2026 (ended June 30, 2026) at $331.8 billion in revenue, up 18 percent year over year, per its own July 29, 2026 earnings release. Nicole Dezen, Microsoft's Chief Partner Officer, has described the company's partner network as "the largest partner ecosystem in the industry, numbering 500,000 and growing," a figure Microsoft has reaffirmed as recently as an April 21, 2026 partner blog post.

At that scale, portfolio pruning is close to costless. No single partner, and no small cluster of partners, moves the topline in a way anyone downstream would trace back to a name. That is not a criticism of how Microsoft runs its partner program. It is the arithmetic that makes "sign broadly, prune the ones that do not produce" a rational strategy, for a company built at this scale.

500,000+

Partners in Microsoft's network, per Chief Partner Officer Nicole Dezen, reaffirmed April 21, 2026

Microsoft's FY26 revenue closed at $331.8B, up 18 percent year over year, per its own July 29, 2026 earnings release. Neither figure moves because of what happens to any single partner.

What pruning looks like at that scale

Giants restructure constantly, and the business does not miss a beat

Microsoft's own partner-level churn is not published with enough precision to cite a comparable number directly, so the clearest public illustration of what "scale absorbs pruning" looks like in practice comes from an adjacent giant, not Microsoft itself. Accenture cut approximately 11,000 positions, about 1 percent of its global headcount, in its fiscal fourth quarter of 2025 (the three months leading into its year ended August 31, 2025), moving from roughly 791,000 employees to roughly 779,000, as part of a company-wide AI-driven restructuring. CEO Julie Sweet described the move on that quarter's earnings call: "We are exiting on a compressed timeline people where reskilling, based on our experience, is not a viable path for the skills we need."

799,000

Accenture's headcount as of its fiscal Q3 2026 (three months ended May 31, 2026), per the company's own investor fact sheet, up from the 779,000 posted at fiscal 2025 year-end and past the original 791,000. Growing this far beyond a cut this size within a year is a range of motion a sub-$1B company's short partner list does not have.

That is a workforce number, not a partner-churn number, and the two are not the same mechanism. It is offered here as the closest available public example of the pattern this brief is about: a firm large enough that pruning one percent of its base, whether that base is employees or partners, does not show up as a disruption anyone outside the company would notice. The pattern, not the specific headcount, is what carries over to how Microsoft treats its own partner base.

The sub-$1B math

A working partner list short enough to hurt when it shrinks

Most sub-$1B ISVs do not have 500,000 partner relationships. They have a working list: the partners driving referrals, co-sell motion, or field introductions this quarter. For a lot of companies at this stage, that list is closer to a dozen or two than a directory. Fifteen is a plausible size for that list, not a benchmarked industry average.

If the working number is fifteen, three partners going quiet carries real weight: the same twenty percent whether those partners leave loudly or simply stop returning calls. Nothing about Microsoft's scale, or Accenture's, transfers to a company running a list this short. A different denominator calls for a different playbook, not a smaller copy of the one built for 500,000 partners.

1 in 5

What losing three names off a working list of fifteen partners means

Fifteen is illustrative, not a Microsoft-published or industry-surveyed figure. It is a plausible size for a sub-$1B ISV's real working partner list, used here to make the ratio concrete.

Depth over breadth

The leverage a company this size has

A sub-$1B ISV cannot out-portfolio Microsoft or a firm the size of Accenture. Microsoft's model works because it is wide enough to average out a weak partner without anyone feeling it. A sub-$1B ISV's model has to work on different terms, because the statistical averaging that protects Microsoft only starts to kick in at a scale most sub-$1B companies will not reach.

What breadth copies

Signing more partners to imitate Microsoft's or Accenture's scale, betting that volume will cover for depth. This borrows the wrong half of the giant's playbook, the size without the reach to match it.

What depth requires

Knowing, partner by partner, which names on a short list carry the motion, tracked individually and renewed deliberately, and treating any one of them going quiet as the twenty-percent event it really is.

Borrowing Microsoft's comfort with churn without borrowing its portfolio size is the trap. The fix runs the other direction from signing more names to imitate the giant's breadth.

Before you find out the hard way

Five things to check before a quiet partner becomes a quarter-defining problem

Count your real list

Write down the specific partners producing referrals, co-sell motion, or field introductions this quarter. Most sub-$1B ISVs find the working number is smaller than the CRM suggests.

Weight it, not just count it

Two partners out of fifteen often carry more of the pipeline than the other thirteen combined. Know which two before either one goes quiet, not after.

Watch for quiet churn

A partner rarely announces it is stepping back. Set a real check-in cadence with your top names, not just a renewal-date reminder on the calendar.

Don't borrow the comfort

Microsoft's tolerance for partner churn is a function of scale you do not have. Adopting its comfort with losing partners, without its portfolio size, is the trap.

Know your concentration number

If three specific names leaving would cost more than three-fifteenths of your pipeline, that concentration, not the raw count, is the risk worth managing.

If three of your fifteen partners went quiet tomorrow, would you know?

The IMS CRI checks six co-sell readiness dimensions in about ten minutes, no email required, including Field Traction: whether named Microsoft sellers know and recommend you. That is a different relationship than the partner list this brief is about, but the same discipline applies. Name the partners carrying this quarter's motion first, then let the Index tell you whether Microsoft's field knows you at all.

Measure your readiness

Microsoft can lose three partners out of five hundred thousand and never notice. A company running a working list of fifteen cannot make that same bet and call it the same risk.

Published FY27 (September 2026). Confirmed directly: Microsoft FY26 revenue of $331.8B, up 18 percent year over year (Microsoft's own July 29, 2026 earnings release); Microsoft's 500,000+ partner network, per Chief Partner Officer Nicole Dezen (blogs.microsoft.com/blog/2026/04/21/accelerating-frontier-transformation-with-microsoft-partners, checked directly Sept 9, 2026); Accenture's approximately 11,000-position reduction in its fiscal Q4 2025 (year ended August 31, 2025), and its subsequent headcount of 799,000 as of fiscal Q3 2026 (newsroom.accenture.com/fact-sheet), Julie Sweet's quote sourced to that quarter's earnings call. The "working list of fifteen partners" used throughout this brief is illustrative, not a Microsoft-published or IMS-surveyed benchmark; the exercise in Section 6 is how to find your own number.