Five VMS Stocks, One Yardstick: Who Ends Up Owning the Cash Flow

Of the €102.4m in group profit that Topicus.com reports for the first half of 2026, €38.0m does not belong to Topicus.com shareholders [1]. At Asseco Poland it is PLN 443.7m out of PLN 873.3m, more than half [3]. At Lumine it is not a single dollar: the half-year accounts show no minority interest position at all [2]. Three offshoots of the same model, three completely different answers to the question of how much of the machine belongs to the shareholder of the listed parent. That question is the reason for this comparison.

Vertical market software attracts quality investors with the same promise: many small, sticky niches, high switching costs, recurring revenues and a capital allocation system that channels cash out of mature businesses into new acquisitions. The template is called Constellation Software, and the pattern is so persuasive that it is now being copied in dozens of offshoots and clones. That is where the trap lies. When five companies work from the same script, the impression arises that they are interchangeably good compounders. They are not. Constellation is the mature quality anchor, Topicus the operationally strongest European offshoot, which in June I still considered the cleanest, Lumine the smaller compounder with a growth path still open, CHAPTERS the young platform with the greatest leverage and the most open flanks, and Asseco the special case that shows why operating strength and shareholder value are two different things. They differ not in the idea but in the execution, and above all in how much of the value creation reaches the shareholder in the end and at what price he buys it.

The comparison has gained additional edge since my individual analyses, because the market has cleared house brutally over the past twelve months. Constellation fell around 39% from its high, Topicus around 40%, Lumine has more than halved from over CAD 50 to CAD 22.65 [7]. The trigger was not an operating collapse but a new worry: whether AI-assisted software development hollows out the moats of the thousand niche vendors on which the entire model rests. The question is therefore no longer only which business is the best, but where quality, a clean structure and price coincide for the first time in a long while after this re-pricing, and whether the AI risk is real or overstated.

I place the five names side by side and measure them against a single yardstick, the original itself: who compounds most cleanly, today and over the next decade?

The yardstick: what makes a clean compounder

Before I compare five companies, it has to be clear what I am measuring them against. In my view a clean compounder meets four conditions at the same time, not just individually.

First, it needs a business with genuine economic substance: high ROIC on the operating core capital, high switching costs, recurring revenues. Second, it must be able to reinvest capital at returns well above the cost of capital; not only the core business but also the capital paid for acquisitions has to earn. Third, the cash generated must actually reach the shareholder of the listed parent and must not seep away through minorities, leverage or holding structures. And fourth, the point at which most cases fail, the price must still leave enough margin of safety that normal operating fluctuations do not immediately turn into a loss of capital.

For these four conditions there is a living textbook example. Constellation Software is not simply another name on this list but the reference against which I measure the other four.

Constellation: the living yardstick

What distinguishes Constellation is not a single outstanding metric but the simultaneity of all four conditions over decades. On the capital currently tied up the group earns 19.4% ROIC, and in the more conservative Leonard logic, measured against everything ever deployed in M&A, still 16.2% [9]. With a spread of roughly 9.6 to 12.9 percentage points, both figures sit far above the cost of capital.

I find the third lens even more revealing: strip out the acquired intangibles entirely and the invested capital of the core business turns negative. Customers pay in advance, and the negative working capital helps finance operations. Economically that means operating capital efficiency is effectively infinite, and the source of value is not the operations but the decision on what happens to the surplus cash. That is the pure form of the compounder model, which the other four only approximate.

It is remarkable that the machine has not stalled at a market value of around CAD 63bn. Free cash flow grew at around 23% a year over eight years, with only a single, marginal year of decline [6]. That consistency through the full cycle, including the interest rate turn, is for me the strongest empirical argument against the AI bear: so far nothing has slowed the cash generation.

That is what makes Constellation a useful anchor for me. Every question put to the four offshoots comes down to the same thing: how close do you get to the original, on quality, on cleanliness, on price? And Constellation itself teaches the humility with which I look at the whole field: even the best business is at today’s levels fairly paid for at best.

The five at a glance

Prices and multiples are stated uniformly as at 8 June 2026 [7], the ROIC figures come from my individual analyses [9], the minority ratios from the 2026 half-year accounts [1][2][3] and the 2025 CHAPTERS annual report [5]. I show the Topicus multiples in their corrected form; the derivation follows in the valuation chapter.

The table makes visible a pattern that runs through the entire comparison: operating quality is abundant in the VMS universe. What is scarce are clean capital structures and attractive prices, and after the share price slide this second point shifts in the buyer’s favour for the first time in a long while.

Business quality: Constellation and Topicus lead

At the pure business level the original and its European offshoot lead. In the Leonard logic Constellation delivers a 16.2% return on all M&A capital ever deployed, plus a core business that ties up so little capital that it is arithmetically capital-negative [9]. Topicus combines a recurring revenue share of 71% with an operating ROIC of 20.4% after adjusting for Asseco and a cash generation so robust that it makes the reported profit line almost meaningless [9]. Free cash flow in 2025 came to around 965% of reported profit and grew at roughly 26% a year from 2018 to 2025, without a single year of decline [8]. That the Leonard ROIC of 16.9% sits below the operating core is a reminder that on a total capital basis, that is including Asseco, some of the shine is lost.

Asseco is the paradox of the quintet: on operating core capital the group earns 30.0% ROIC, the highest core return of all five [9]. It does the shareholder little good, because it is diluted on the way up. Including the purchase price paid, ROIC falls to 12.9%, and a large part of what remains belongs to minorities.

Lumine and CHAPTERS are the early-stage representatives. Both have the right DNA, that is specialised niches, recurring revenues and decentralised management, but on a smaller base and with a shorter history. At CHAPTERS, HGB accounting with its enforced goodwill amortisation completely conceals the economic reality of a core business earning around 19% ROIC: reported EBIT is negative while free cash flow is clearly positive [5].

Capital allocation: where the gap between core and purchase price decides

The metric I find most interesting is not the ROIC of the core business but the gap between core ROIC and ROIC on the purchase price paid. It shows whether a compounder deploys capital with discipline or wastefully [9].

Constellation and Topicus show the healthiest profile here. At Constellation the proof was delivered long ago: even in the conservative Leonard calculation, which counts every dollar ever put into M&A, ROIC stands at 16.2%, a spread of around 9.6 percentage points over the cost of capital, consistent across decades. Topicus follows with a Leonard ROIC of 16.9%; the distance to the operating core is moderate and the level stays above the cost of capital. That is the mechanics of an allocator that pays for quality but does not overpay.

Lumine is the special case in this table, not because discipline is missing but because the proof is not yet measurable. Since the 2023 spin-off the standalone history has been too short and the balance sheet too much in motion because of the large acquisitions (most recently Synchronoss) to calculate a reliable Leonard ROIC on the capital paid. The indications point to discipline: Constellation DNA in the playbook, clean cash flow conversion, moderate leverage. But that is the difference between indication and proof: at Constellation and Topicus the acquisition return is measured, at Lumine it is so far a matter of trust. That is part of why I consider the 12x premium harder to justify.

Since the 2026 half-year report, however, that trust has a name and a price. On 13 February 2026 Lumine acquired Synchronoss Technologies outright for $309.3m in cash, financed through credit facilities; financial debt rose as a result from $209.9m to $481.7m [2]. In the first half Synchronoss contributed revenue of $64.3m and a net loss of $3.9m [2]. The question of Lumine’s capital allocation discipline is therefore no longer abstract: it has acquired a concrete, dated and quantifiable test object whose outcome will become visible over the next four to six quarters.

Asseco reveals the opposite pattern: an excellent core business (30.0%) pushed down to 12.9% on the capital paid by expensive historic acquisitions. CHAPTERS is the sharpest case. On the purchase price paid, ROIC falls to around 4%, below the WACC of 7.9% [9]. As long as that figure sits below the cost of capital, every further debt-financed acquisition arithmetically widens the gap rather than closing it, all the more so because at 3.87x net debt/EBITDA CHAPTERS is the most highly leveraged of the five [7].

Who owns the profit? The Asseco trap

This is where the field separates most clearly. However good a compounder is operationally, if the profit it generates does not reach the shareholder of the listed parent, the business quality is worth only half as much to the investor.

That is the central reason why Asseco is not a clean compounder despite the highest core ROIC and the optically lowest multiple. The relevant cash flow is not the group cash flow but the parent-attributable cash flow after minority outflows, and that is considerably more ordinary than the headline picture.

CHAPTERS suffers from the same problem, only in milder form. Minorities account for around 29% of consolidated equity, and CHG owns only 61.7% of the largest EBITDA block [5].

Topicus: the correction of 4 September 2026

An earlier version of this article said: “At Constellation, Topicus and Lumine this seepage is smallest.” I no longer hold that statement in that form. The half-year accounts of Topicus.com show a minority ratio that sits considerably closer to the Asseco finding than to Constellation’s cleanliness [1]:

The structural reason lies one level down: Topicus.com holds only 64.33% of Topicus Coop, the operating vehicle. The Joday Group holds 29.38%, the Ijssel Group 6.14% [1]. Accordingly, 83.5m basic shares stand against 129.8m fully diluted shares, a dilution potential of 56% [1].

37% against 47% is therefore a difference of degree, not of structure. Both companies fully consolidate subsidiaries they do not fully own; at Asseco the effect is larger, at Topicus it is constructed differently (a chain of holdings rather than many individual subsidiaries), but it is of the same order of magnitude. The clean group thus shrinks to Constellation and Lumine.

At Asseco the finding has become sharper still in the first half of 2026. Of PLN 873.3m in group profit only PLN 429.6m was attributable to the shareholders of the parent company; PLN 443.7m, or 50.8%, went to minorities, after 57.2% in the prior-year half [3]. On the balance sheet side, PLN 5,178.0m of PLN 12,136.6m in equity belongs to minorities, that is 42.7% [3]. A single figure from the notes shows the core of the structure: at Matrix IT Group, the largest earnings contributor, the minority ratio is 87.69% [3]. Asseco fully consolidates this company and economically owns a fraction of it.

That reverses one of this article’s statements: on this axis Lumine is the cleanest of the four offshoots, not Topicus. The half-year accounts show no minority interest position at all, and the entire group profit belongs to Lumine shareholders [2]. That is the structure I originally attributed to Topicus.

What Asseco is doing about the problem is notable at the same time: in the first half of 2026, PLN 190.5m was spent on acquiring minority interests, after PLN 85.3m in the prior-year period [3]. That is the condition I formulate in the closing section: Asseco only becomes interesting as a share once the minority ratio falls structurally. The company is working on it. With a minority book value of PLN 5.2bn, however, a buy-back volume of PLN 190m per half-year is a very long road.

None of this devalues the Topicus case. The share that reaches the shareholder is known, disclosed and stable (36.7% in the prior-year half, 37.2% now). What it does devalue is the claim that Topicus is comparable to Constellation on this dimension. Anyone buying Topicus buys roughly two-thirds of a compounder machine, not the whole of it.

The profit line lies, and at three of them in different ways

A recurring motif: at these companies reported net income is a poor guide. The reasons differ.

At Topicus, a write-down to acquisition cost of €221.7m in the course of the Asseco accounting pushed the 2025 result from €92m down to €42m, an accounting effect, not an operating deterioration [9]. At Asseco the effect works the other way: reported group earnings jumped to PLN 3,628m in 2025, but the bulk came from the sale of the Sapiens majority stake, a one-off [9]. At CHAPTERS, HGB accounting enforces straight-line goodwill amortisation that produces a negative EBIT while free cash flow remained positive and is growing at a 64% CAGR [5]. The reported LTM ROIC of −1.9% at TIKR is precisely this artefact [7].

The lesson for all five: anyone valuing these companies on simple earnings multiples is analysing the accounting, not the business. Cash flow and ROIC on the right capital are the more reliable lenses. It is no coincidence that the NTM MC/FCF multiple, that is the price on free cash flow, is the most informative valuation line in the overview table.

The shared risk: AI against the moats

However different the five are in structure and maturity, they share one risk, and it is the reason for the whole segment’s share price collapse. The VMS model rests on thousands of small niche vendors being protected by high switching costs, deep process integration and a lack of alternatives. Since 2026 the market has been questioning those moats: if AI-assisted software development lowers the cost of replacing or rebuilding a niche solution, the protection that carries pricing power and customer retention shrinks.

This risk does not hit the five equally. Constellation and Topicus are the most broadly insured through diversification across thousands and hundreds of end markets respectively; no single disrupted niche market moves the whole. Lumine is concentrated on the communications and media industry and is therefore more narrowly exposed. Asseco sits in regulated, often public-sector end markets where replacements are slow and political, which paradoxically protects it. CHAPTERS, for its part, is trying to use AI offensively as an opportunity by turning the structured data of its OpCos into new products.

So far the empirical counter-argument comes from the cash generation itself: across the entire interest rate cycle and through the current AI hype, free cash flow at Constellation, Topicus and Lumine has continued to grow. Nothing has slowed the machine to date. That does not make the risk smaller, but it does put into perspective the speed at which a moat erosion thesis would have to show up in the figures. For me that is the decisive open question and the reason why even the original is only fair today rather than cheap.

The tokenisation test: five questions, five very different answers

The AI risk can be framed more precisely than through the blanket question “are the moats eroding?”. The software sector is going through a concrete, measurable upheaval in parallel: the shift from per-seat pricing to token, usage and outcome models. IDC expects around 70% of vendors to abandon pure seat pricing by 2028 (the full dossier here) [9]. The dossier yields a checklist of five questions that can be applied to any software name. Applied to the five compounders, it shows that the market has punished the group uniformly while the risk is anything but uniformly distributed.

Three findings stand out. First: Constellation and Topicus pass the test almost completely. Their billing units were never the seat for routine knowledge work, their maintenance revenues are committed rather than usage-dependent, and both can layer AI functions as usage-priced modules on top of the captive customer base without ever having to go through the expensive pricing migration of horizontal SaaS vendors. That is the offensive reading which gets lost in the blanket AI discount.

Second: Asseco does surprisingly well on this axis. Regulated, public-sector end markets are the most deflation-resistant form of moat there is. That sharpens the finding from the structure chapter: the 5.85x multiple really is a pure minority discount, not a hidden AI risk. Anyone buying the discount buys the structure, not the disruption.

Third: CHAPTERS collects the most open flags in the field, and not because anything negative has been demonstrated but because the decisive disclosures are missing. Which licence metrics do the top OpCos use (German Mittelstand software is traditionally licensed per workstation, the German translation of the seat)? How high is the ARR ratio? And will the AI modules of the Momentum initiative be priced separately by usage or bundled into existing maintenance fees? Those were my three questions for the annual general meeting. At 23.3x EV/EBITDA not even the pricing resilience that the multiple has long presupposed can be verified.

Valuation: the share price slide changes the picture

This is where most has happened since the individual analyses. Back then my finding for almost every name was: good quality, but no margin of safety. After a year in which the whole segment lost 25 to 40%, that statement needs differentiating [7].

The surprise is the relative valuation, though in the opposite direction to the one the first version of this article assumed. The raw figures suggest Topicus trades at 10.65x NTM EV/EBITDA, more cheaply than Constellation (11.97x) and Lumine (12.03x) [7]. Those raw figures are calculated on an inconsistent basis, and the error can be demonstrated directly from the data provider’s cash flow extract.

The arithmetic error in the Topicus valuation

The data provider’s multiples (10.65x NTM EV/EBITDA, 14.12x NTM MC/FCF) are calculated against the basic share count of 83.5m; the free cash flow behind them, however, is the consolidated figure [7][8]. Reported for 2025 are free cash flow of €403.09m and a “cash flow per share” of €4.83 [8]. The division reveals the divisor: 403.09 ÷ 4.83 = 83.5m shares. For the last twelve months the same calculation gives 414.73 ÷ 4.96 = 83.6m. So it is the basic share count throughout.

This consolidated cash flow, however, belongs to Topicus shareholders only to the extent of 64.33% [1]. There are two consistent routes, and they have to deliver the same result:

• Attributable FCF against basic shares: 403.09 × 64.33% ÷ 83.5 = €3.11
• Consolidated FCF against diluted shares: 403.09 ÷ 129.8 = €3.11

The cross-check works to the cent. Correct free cash flow per share is €3.11, not €4.83; the market data figure is 56% too high. All multiples derived from it are therefore too low by the same factor.

Derivation of EV/EBITDA: of the 10.65x, 0.59x relates to net debt and 10.06x to market capitalisation. Only the market capitalisation part is corrected by the factor 1.555: 10.06 × 1.555 + 0.59 = 16.2x. For MC/FCF the factor applies to the whole metric: 14.12 × 1.555 = 22.0x.

That overturns the central statement of the first version. It said: “Topicus now trades at 10.65x NTM EV/EBITDA, more cheaply than Constellation and more cheaply than Lumine.” On a consistent basis the opposite is true. At 16.2x EV/EBITDA and 22.0x MC/FCF, Topicus is the second most expensive name in the field after CHAPTERS, distinctly dearer than the original (12.0x / 13.8x) and dearer than Lumine (12.0x / 14.2x).

The margin of safety that the 40% share price decline supposedly created at Topicus does not exist in that form. It was an artefact of dividing the cash flow of a company that belongs to others by a good third by a share count that counts only its own shareholders. Anyone who does not correct that error believes Topicus to be 30% cheaper than it is. That is the core of the case, not a footnote. The 40% share price decline did release real valuation air at Topicus; it did not create a margin of safety relative to Constellation. Anyone buying the shares because of the supposed discount to the original is buying a metric, not a mispricing.

What else the cash flow extract shows

Three figures from the same extract fit the rest of the half-year findings [8]. First: free cash flow rose from €403.09m (2025) to €414.73m over the last twelve months, plus 2.9%. For 2018 to 2025 I quote a growth rate of around 26% a year above; the machine has therefore become noticeably slower, and it did so before any AI effect would be visible. Second: the FCF margin fell from 26.0% to 24.4%. Third: spending on acquisitions fell from €273.85m in 2025 to €97.98m over the last twelve months, minus 64%. That matches the weak M&A half-year and, in a model whose return depends on the reinvestment rate, it is the most uncomfortable of the three figures.

What is confirmed, by contrast, is my statement about the quality of the profit line: free cash flow of €403.09m against reported net income of €41.76m is exactly the 965% I cite above [8].

The other three on the price axis

Lumine is the other side of the same coin: despite the share price halving, at 12.03x it remains the dearest of the three “clean” businesses and therefore still carries a premium that is harder to justify now that the discovery discount has gone [7]. On multiples, Asseco is by far the cheapest at 5.85x EV/EBITDA and a 6.1% dividend yield, but that is the structural discount, not a discount on substance [7]. And CHAPTERS is by some distance the most expensive name in the field at 23.29x EV/EBITDA; even after a 26% fall in the share price the market is pricing in a successful compounder that has yet to grow into its valuation [7].

On the price basis: the price table in this article deliberately remains as at 8 June 2026 so that all five names stay comparable. For context: Topicus was quoted at CAD 104.28 on 7 August 2026, around 1.6% above the CAD 102.67 used here [7]. The relative valuation statement does not change as a result.

Two questions, two rankings

A distinction is worth drawing here that the market likes to blur. These five names do not answer the same investor question. Anyone looking for the safest compounder arrives at a different order from anyone looking for the largest possible multiple of their money. Both questions are legitimate, but they lead to different shares, and most mistakes arise where the two get mixed up.

Question one: who is the safest, cleanest compounder? What counts here is the simultaneity of quality, a clean structure and a sensible price.

My answer to this question after the half-year reports differs from the one in the first version. Constellation is the best company with the burden of proof already discharged, but even after the crash it offers no margin of safety, with a fair value around 8% below the price. Topicus remains operationally the strongest European offshoot, yet it loses on both axes where it previously led: 37% of the profit goes to minorities, and after correcting for dilution it is dearer than the original at 16.2x EV/EBITDA rather than cheaper. It remains a very good business at a price that forgives nothing. Lumine has the cleanest structure in the entire field, no minorities and no dilution caveat, but at 12.03x it stays the dearest of the three clear businesses, and organic growth in the first half of 2026 was only around 4.5% [2]. Asseco and CHAPTERS fall back on this axis, not because the operating business is weak but because at Asseco almost half the profit goes to minorities and CHAPTERS earns less than its cost of capital on the purchase price paid.

Question two: which share could multiply the most from today’s starting point? Here the order reverses, but not as uncritically as it first tempts one to think.

In theory, smaller compounders that have yet to complete their proof have the longest reinvestment runway and, if successful, the strongest leverage. On that logic CHAPTERS moves to the front, with Lumine right behind. Both are small enough for individual successful acquisitions to change the profile noticeably, an asymmetry that the mature Constellation can structurally no longer offer.

At this point, though, I stay disciplined. CHAPTERS already trades at 23.3x EV/EBITDA, the most expensive multiple in the field. The return source “multiple expansion” is weakest precisely where the multiple is already highest. The upside at CHAPTERS therefore does not depend on a re-rating but on the company growing into its valuation: on margins normalising, on ROIC including intangibles breaking through the WACC threshold and on the interest burden staying bearable. That is possible, but it is a bet on execution, not on a mispricing. Lumine is the more measured version of the same idea: smaller than Topicus, with a more open growth path, but without the leverage question marks of CHAPTERS and without its definitional problems in the income statement. The price for that is that the entire growth path currently hangs on a single acquisition.

The investment case therefore consists in first being clear about your own question. Anyone looking for stability holds Constellation as the anchor and finds in Lumine the cleanest complement. Anyone deliberately trading uncertainty for potential looks at Lumine and, with eyes open to the conditions, at CHAPTERS. In both cases Asseco remains the special case: operationally strong, but only interesting as a share once the minority ratio falls structurally.

Bear versus bull for the overall case

What has come in since June

On 4 September 2026 I incorporated the 2026 half-year reports of Topicus.com (published on 5 August 2026), Lumine Group and Asseco Group as well as the annual general meeting materials of CHAPTERS Group of 13 July 2026 [1][2][3][4]. Constellation has not been re-examined in this version; the CSU figures stand as at 8 June 2026. I have pulled the correction of the Topicus minority ratio and the Topicus multiples into the respective chapters above. The further findings by company follow here.

Topicus: four findings from the half-year

Beyond the minority ratio, the half-year accounts of 5 August 2026 deliver four findings that sharpen the Topicus part of this comparison [1]. None of them overturns the case; taken together, though, they qualify the label “cleanest European offshoot”.

The Asseco stake is losing significant value. Its book value fell from €489,851k (December 2025) to €454,665k as at 30 June 2026 [1]. More serious is the market value: on the basis of the Asseco share price it fell from €1,038,894k to €747,381k, minus 28% in six months [1]. The buffer between market and book value, which until now defused every impairment discussion, thus melts from around €549m to €293m. It remains comfortable; the direction is new.

The positive free cash flow comes from a dividend. In the first half of the year Topicus received an Asseco dividend of €59,023k [1]. Without that item the half-year’s free cash flow would have stayed negative. This is faultless in accounting terms and explicable operationally, and the payment is real and recurring. But it means that the half-year’s cash generation did not come from the core VMS business but from a financial stake whose market value fell 28% over the same period.

M&A volume remains low. In the first half of 2026, acquisitions were made with total consideration of €62,734k (of which €46,251k in cash, the rest holdbacks and contingent purchase price payments) [1]. After the balance sheet date a further €56,123k was committed [1]. That is more than the €40.2m the press release named, but for a serial acquirer of this size it is a weak half-year. In a model whose entire return depends on the reinvestment rate, low deployment is not a side issue.

The tax rate is rising. 18% in the half-year (Q2 alone: 21%) against 15% in the prior-year half [1]. Three percentage points on the tax rate are no detail in a compounder model with a long reinvestment runway; they are a permanent deduction from NOPAT.

Lumine: clean, but expensively bought

The 2026 half-year moves the Lumine picture in two opposing directions [2]. On the structural side Lumine is better than assumed, on the capital allocation side more open.

Growth is almost entirely bought. Revenue rose 22.3% to $443.4m [2]. Of that, $64.3m comes from Synchronoss Technologies, acquired in February 2026. Strip out that contribution and $379.1m remains against $362.6m in the prior-year half, organically around 4.5%. For a VMS compounder that is a normal figure, not a weak one; but it means the headline of 22% growth describes the shopping, not the business.

The purchase was large and is still unproven. For Synchronoss, Lumine paid $309.3m in cash, more than fifty times what was spent on acquisitions in the whole of the prior-year half ($6.8m) [2]. It was financed by a drawdown of $380.0m on the credit facilities; financial debt rose from $209.9m to $481.7m [2]. In the half-year Synchronoss contributed revenue of $64.3m and a net loss of $3.9m [2]. In the purchase price allocation only $28.5m was assigned to goodwill; the lion’s share went into technology ($111.0m) and customer assets ($175.0m), which are amortised on a scheduled basis [2]. That is why amortisation of intangible assets jumped from $52.3m to $67.6m.

Cash generation has declined. Operating cash flow fell 25.5% to $88.3m; after capital expenditure and lease repayments, around $82.7m of free cash flow remains, after $113.7m in the prior-year half, a fall of 27% in a half-year in which revenue rose 22% [2]. The EBITDA margin also gave way: from 35.0% to 31.0% [2].

What that means for the assessment of Lumine: I carry Lumine as “clean cash flow, good asymmetry, but the dearest of the clean businesses”. After the half-year report the first part is better documented than before: no minorities, no dilution question, the entire profit belongs to shareholders. The second part gains new content. The 12x premium was hard to justify because the acquisition return was unproven. Now the object of proof lies concretely on the table: $309m for a business that made a loss in the first half. Whether Lumine is a disciplined allocator will be decided at Synchronoss, and over the next four to six quarters, not in ten years.

CHAPTERS: the annual general meeting did not answer the three questions

I had formulated three questions for the annual general meeting: which licence metrics do the largest operating companies use? How high is the ARR ratio? And will the AI modules be priced separately by usage or bundled into existing maintenance fees?

The materials from the annual general meeting of 13 July 2026 answer none of these three questions [4]. There is no disclosure of the OpCos’ licence models, no ARR ratio and no statement on the pricing of the AI functions. My valuation statement therefore stands unchanged: at 23.3x EV/EBITDA, pricing resilience remains unverifiable.

What the materials do deliver is highly revealing in one place and uncomfortable in another.

Revealing: the adjustment ratio. Adjusted operating EBITDA of the holdings came to €49.12m pro forma in 2025 [4]. The adjustments on it amounted to €22.69m; reported EBITDA of the operating companies stood at €26.44m [4]. Around 46% of the headline metric therefore consisted of adjustments, after 14% in 2023. The company discloses this itself and forecasts normalisation to below 15% for 2026 [4]. Until then the rule holds: anyone valuing CHAPTERS on “adjusted operating EBITDA” is valuing a question of definition to nearly half its extent. At group level, after holding costs, €12.27m of reported EBITDA before share-based compensation remains, out of a €49m headline [4].

Uncomfortable: the 2025 income statement. The net loss for the year came to €40.2m, of which €10.6m was attributable to other shareholders [4]. Adding back the HGB-driven amortisation of purchase price allocation and goodwill (€43.9m) leaves an adjusted group profit of −€0.6m, still negative; adjusted earnings per share stand at −€0.03, or −€0.05 excluding the securities result and share-based compensation [4]. The company reported €0.28 in each of 2023 and 2024. The road back is ambitiously scheduled: €0.80 to €1.10 for 2027e [4].

Net debt of 2.9x is lower than the 3.87x quoted above; the difference arises from the reference figure. The 2.9x is measured against pro-forma adjusted operating EBITDA of €49.1m [4], the 3.87x from the market data source against a narrower EBITDA definition [7]. Against reported group EBITDA of €12.3m it would arithmetically be more than 11x. The range of these three figures is itself the finding: at CHAPTERS the level of leverage depends on which earnings definition you accept, and at 23.3x EV/EBITDA that question of definition is the most expensive open flank in the field.

Conclusion

Five companies, one script, five very different realities. The popular assumption that every Constellation clone is automatically a good compounder does not survive scrutiny. Operating quality is abundant in the VMS universe; Constellation, Topicus and Asseco even show excellent core returns. But quality alone does not make a good share, and above all it does not make the same share.

The case is decided on the levels above, and the soberest answer depends on the question you ask. Anyone looking for the safest, cleanest compounder ends up with Constellation as the anchor and Lumine as the structurally cleanest alternative, both carried by proven or cleanly attributable quality, but both at fair rather than giveaway prices. Topicus, the “best relative entry point” in the first version of this article, falls back on this axis: not because of the business but because two metrics that argued for it did not survive a check against the primary figures. Anyone deliberately trading more uncertainty for more potential looks at Lumine and CHAPTERS, with CHAPTERS combining the greatest theoretical upside with the most expensive multiple and the highest number of conditions still to be proven. Asseco remains the instructive special case: a good business whose value to the shareholder hangs on the ownership structure, not on operating strength.

Over all of it stands the shared AI risk, which explains why even the best business in the field is only fair today rather than cheap. The tokenisation test shows, however, that this risk is not evenly distributed: Constellation and Topicus stand structurally more on the beneficiary side of the pricing shift, while resilience at CHAPTERS remains unproven for lack of disclosure. My soberest conclusion is therefore: Constellation is the best company. Lumine has the cleanest ownership structure in the field and, with Synchronoss, the most concrete open proof at the same time. Topicus is operationally the strongest European alternative, but structurally less clean than originally claimed here. CHAPTERS is the name with the greatest leverage and the most conditions, not one of which the annual general meeting resolved. Patience in this universe is not a sacrifice of returns. It is part of the return, and after the re-pricing it is in one place a little less necessary than it was a year ago.

Update log

• 05.09.2026 — Language revised to house style, list of sources added; figures and assessment unchanged. The correction and addendum boxes of 4 September have been moved into the chapters “Who owns the profit?”, “Valuation” and “What has come in since June”.
• 04.09.2026 — Correction and addenda based on the 2026 half-year reports: Topicus minority ratio 37.2% (Topicus.com holds 64.33% of Topicus Coop), free cash flow per share €3.11 instead of €4.83, corrected multiples 16.2x EV/EBITDA and 22.0x MC/FCF; Lumine with no minority interest position, Synchronoss acquisition for $309.3m; Asseco minority ratio 50.8%; CHAPTERS annual general meeting with no answer to the three open questions, adjustment ratio 46%. Prices still as at 8 June 2026.
• 09.06.2026 — First published, based on the five individual analyses, price basis 8 June 2026.

Sources

1. Topicus.com Inc., Condensed Consolidated Interim Financial Statements Q2 2026 (half-year to 30.06.2026), published 05.08.2026. Group profit and minority interests H1 2026/H1 2025, holdings in Topicus Coop (64.33% / 29.38% / 6.14%), basic and diluted share count, book and market value of the Asseco stake, Asseco dividend, acquisition volume, tax rate. → ‹TODO: Link› (press release: https://topicus.com/news/constellation-software-inc-and-topicuscom-inc-announce-results-for-topicuscom-inc-for-the-second-quarter-ended-june-30-2026)
2. Lumine Group Inc., Condensed Consolidated Interim Financial Statements Q2 2026 (half-year to 30.06.2026), published 04.08.2026. Revenue, Synchronoss contribution, purchase price allocation, financial debt, operating and free cash flow, EBITDA margin, recurring revenue share, net profit with no minority interest position. → ‹TODO: Link› (press release: https://www.globenewswire.com/news-release/2026/08/04/3338792/0/en/Lumine-Group-Inc-Announces-Results-for-the-Second-Quarter-Ended-June-30-2026.html)
3. Asseco Group, Semi-annual Report H1 2026 (consolidated financial statements for the six months to 30.06.2026). Group profit and minority interests, minority equity, Matrix IT minority ratio, expenditure on the acquisition of minority interests. → ‹TODO: Link›
4. CHAPTERS Group AG, Presentation for the 2026 Annual General Meeting, Hamburg, 13.07.2026. Pro-forma 2025 key figures, reconciliation of adjusted operating EBITDA, adjustment ratio, 2025 income statement, balance sheet and financial debt as at 31.12.2025, guidance upgrade, 2027e EPS target, AI slide. → https://www.chaptersgroup.com/wp-content/uploads/2026/07/20260713_CHAPTERS_HV.pdf
5. CHAPTERS Group AG, Annual Report 2025 (HGB consolidated financial statements), published May 2026. Goodwill amortisation, minority interests in equity, share of the fintech segment, free cash flow, interest burden. → https://www.chaptersgroup.de/wp-content/uploads/2026/05/CHAPTERS-Group-AG-Geschaftsbericht-2025.pdf
6. Constellation Software Inc., Financial Report Fourth Quarter and Fiscal Year 2025 (Q4 2025 Shareholder Report), published April 2026. Free cash flow and its history, deferred revenue. → https://www.csisoftware.com/wp-content/uploads/2026/04/Q4-2025-Shareholder-Report.pdf
7. TIKR Terminal, market data as at 08.06.2026 for Constellation (TSX: CSU), Topicus (TSXV: TOI), Lumine (TSXV: LMN), Asseco Poland (WSE: ACP) and CHAPTERS Group (XETRA: CHG): prices, NTM EV/EBITDA, NTM MC/FCF, net debt/EBITDA, LTM ROIC, dividend yield, declines from the high; Topicus quotation 07.08.2026 (CAD 104.28).
8. TIKR Terminal, cash flow extract Topicus.com (financial year 2025 and last twelve months): free cash flow, cash flow per share, net income, FCF margin, acquisition spending.
9. Individual analyses and topic dossier on frinvesting.de: ROIC readings, fair values and structural findings from the articles on Constellation Software, Topicus, Lumine Group, Asseco Poland and CHAPTERS Group as well as the IDC forecast from the dossier Tokenisation; each with their own sources and reporting dates.