Chapter 2

The Customer Count Break

The de-rating named a pivot without dwelling on it: the recurring-customer count, the number that for two decades read as a live measure of how far the network had spread. SPS reports its business on two disclosed dials — the count of recurring revenue customers and the average recurring revenue each one pays — and revenue is their product [1]. Through fiscal 2024 the two dials turned together, and the count did most of the storytelling: each new trading-partner connection SPS lit up on a retailer's behalf showed up as another customer on the network. In 2025 they came apart. The count leapt, then broke; the revenue-per-customer figure has carried the weight ever since. This act is about how that happened, what it leaves behind, and why today's forward estimates depend on a base that is now shrinking.

Two dials, moving together

SPS defines a recurring revenue customer as one with an active recurring contract at period end, and it labels the great majority "1P" — first-party brands, distributors and suppliers wired into a retailer's network — reserving "3P" for accounts whose only tie is an online-marketplace connection. The second dial, annual revenue per user (ARPU, the metric SPS used to call "wallet share"), is annualized recurring revenue divided by the average customer count [2].

For most of the past decade both climbed in near-lockstep. Customers roughly doubled from 24,805 at the end of 2016 to 45,350 at the end of 2024; ARPU rose from about $7,344 to about $13,300 over the same span. Neither moved in jumps — the count added two-to-three thousand net customers a year through community-enablement campaigns, and ARPU stepped up as existing customers bought more of the product line.

Source: recurring-customer count and ARPU as reported in successive Forms 10-K; FY2025 figures per FY2025 10-K MD&A [3].

Nearly all of that revenue is recurring — about $718.0M of $751.5M in FY2025, of which Fulfillment (the core EDI product) is roughly 89% — so the customers-times-ARPU identity is not an abstraction. It is very nearly the whole income statement, and for years a reader could watch the network expand simply by watching the count.

The Carbon6 seam

In fiscal 2025 the count jumped 20%, from 45,350 to approximately 54,600 [4]. Set against the prior decade's two-to-three thousand a year, a jump of 9,250 looked like the flywheel spinning faster. Almost none of it was.

Of those net additions, roughly 8,500 arrived in a single month. SPS closed its acquisition of Carbon6 — a provider of software tools for Amazon sellers — in February 2025, and the deal brought over about 8,500 recurring revenue customers "of which approximately 300 are 1P recurring revenue customers and the remainder are 3P" [5]. Strip out Carbon6 and the organic base added on the order of 750 customers for the year — its slowest pace in a decade. The 20% headline and the underlying network grew almost nothing in common.

The seats also weighed on the second dial. Because Carbon6's Amazon-seller accounts pay far less than a typical 1P customer, adding roughly 8,200 of them mechanically pulled ARPU down: management put the full-quarter drag at "approximately $1,400" in the second quarter of 2025 alone [6]. So the two dials that had always turned together split in opposite directions: the count spiked on acquired volume while the price-per-customer it should have signaled fell. On the quarter it closed, management framed the deal as establishing "SPS as a clear leader in the emerging category of revenue recovery, supporting supplier communities of the two largest global retailers" [7].

That framing sat against a growth promise being quietly retired at the same time. As late as the third-quarter 2024 call, SPS still told investors it maintained "our annual revenue growth expectation of 15% or greater" [8]. Two quarters later the language changed: "beyond 2025, we expect our revenue growth rate, excluding future acquisitions to be at least high single digits" [9]. The multi-year algorithm — a round-number 15% that acquisitions had helped hit — was replaced by an organic frame roughly half as fast, in the same window the count was being padded with acquired seats.

Giving it back

The seats did not stay. By the first quarter of 2026 SPS was managing them off deliberately: it introduced a subscription platform fee for its 3P take-rate accounts and told investors to expect "a projected decline of up to 4,000 3P suppliers in 2026," adding that it did "not anticipate this action to result in a material impact to revenue" [10]. A managed cull, in other words — the low-value cohort would be priced down to a smaller, better book.

One quarter later the plan was overtaken by a full exit. On June 30, 2026, SPS sold the entire 3P revenue-recovery business, received about $8.8 million in cash, and booked a loss of approximately $23.5 million; the same acquisition had cost $210.2 million in total consideration seventeen months earlier [11]. The divestiture removed "approximately 7,300 customers," dropping the recurring count to roughly 46,600, and because those were the lowest-ARPU accounts, their removal pushed ARPU the other way — up to about $15,100 [12]. After the sale, the network is once again entirely first-party: "all recurring revenue customers are classified as 1P" [13].

Sources: Carbon6 addition, FY2025 10-K [14]; divestiture customer reduction and loss, Q2 FY2026 call and 10-Q [15], [16].

Read as a round trip, the arithmetic almost cancels. Roughly 8,500 seats in, roughly 7,300 out; the count ended the second quarter of 2026 at about 46,600, barely above the 45,350 it left FY2024 at. What changed hands in between was $210 million of consideration, a $23.5 million book loss, and the "clear leadership" claim, retired within eighteen months of being made. The customer count that looked like a 20% expansion in 2025 was, on the network that matters, a wide detour back to roughly where it started.

The break, in one picture

The clearest way to see the seam is to put the two dials side by side across the years they diverged. The count spikes on Carbon6, then craters on the divestiture; ARPU never pauses.

Source: recurring-customer count as reported; FY2025 per 10-K MD&A [17]; Q2 2026 per Q2 FY2026 call [18].

Source: ARPU as reported; FY2025 per 10-K MD&A [19]; Q2 2026 per Q2 FY2026 call [20].

The count line now bends the wrong way for a network-effects story: the unit that historically signaled expansion has turned down. The ARPU line does the opposite, and its recent lift is partly cosmetic — removing the cheapest accounts raises the average of what remains without anyone buying more. Both moves trace to the same Carbon6 seam.

What is left when the seats clear

Underneath the detour, the core did not break — it was obscured by it. Asked what the business looks like without the Amazon piece, management was explicit: "if you remove Amazon revenue recovery from Q1, the rest of the business is already growing high single digits" [21]. The Fulfillment engine — roughly nine-tenths of recurring revenue — kept compounding through 2025; the reported growth rate was flattered on the way up by acquired seats and depressed on the way down as those seats churned and were sold, but the 1P network in the middle grew at a rate the headline never cleanly showed.

Precisely how much of the past growth was organic cannot be measured from the outside, because SPS does not measure it internally. In a 2024 response to an SEC comment, the company stated that it "monitors recurring revenue customers and wallet share as the primary metrics" and "does not actively track revenue broken out by new and existing customers" [22]. The one disclosed anchor is the acquisition pro-forma: had Carbon6 been owned for all of 2024, that year's revenue would have been about $680.9 million rather than the $637.8 million reported — roughly $43 million, or 6.8%, of acquired top line that the deal layered onto the comparison [23]. A reader can bound the organic run-rate — high single digits by management's own framing, a mid-single-digit forward consensus by the market's — but cannot pin it, and the company's own metrics are built the same way.

Where forward growth comes from

The reset is already in the numbers rather than only in the guidance. Consensus embeds forward revenue growth of about 5.3% — down from the 17.8% SPS delivered in FY2025 — and the first half of 2026 grew roughly 5.7%, so the step-down is realized, not projected. What keeps even that mid-single-digit figure positive is a specific arithmetic: with the count now falling year-on-year, the entire top line leans on ARPU. Consensus survives by pairing double-digit revenue-per-customer growth against a customer base that shrinks — the two dials that split in 2025 are now modeled to keep pointing in opposite directions, and only ARPU is additive.

Source: FY2025 growth as reported; H1 FY2026 from Q2 FY2026 results; FY2026 consensus per aggregated analyst estimates.

For a decade, network expansion and a rising customer count were the same fact, legible at a glance. They have come apart. The count can fall while the 1P base deepens — more products sold to the same trading partners — and that is what the recent ARPU lift, cosmetic component aside, is built to show. Whether richer wallet share on a slowly growing 1P base can stand in for the customer-count growth that used to do the work is the thread the rest of the report follows: first into the margin the full-service model earns on those relationships, and then to the people now deciding how the network's cash is spent.