How to Analyze a SaaS IPO

If you want to know how to analyze a SaaS IPO, the answer is to stop reading the prospectus as a story and start scoring it as a balance sheet of claims. The narrative is written to sell the offering. The numbers underneath it are filed under securities liability. Your job is to score the four things the S-1 cannot dress up, in a fixed order, before the story has a chance to frame your judgment.

Those four things are growth quality (is the growth efficient or bought), retention (gross and net), the path to profitability (gross margin and the burn hiding behind the adjusted numbers), and valuation versus that profile. The order is a scorecard, not a vibe. Run the same five-test pass on every SaaS IPO and the good ones separate from the hyped ones fast.

This piece shows how to analyze a SaaS IPO using real, recent filings: ServiceTitan, Rubrik, Klaviyo, Instacart, CoreWeave, plus Snowflake and Datadog as post-IPO benchmarks. Every figure ties to a specific filing and period. The framing is analytical: how to think about a filing, not what to do about any stock.

Key takeaways

• Four dimensions, one order. Score growth quality, then retention, then the profit path, then valuation. Management spins prose; it cannot spin these disclosures without consequence.
• Retention is the truth serum. ServiceTitan disclosed 110%+ net dollar retention across ten quarters and Rubrik 133% (their Forms S-1). But only about 5% of public SaaS companies disclose gross retention versus 67% for net (Ordway Labs), and gross is the number that reveals the churn baseline.
• SBC is the true cost of labor. Rubrik reported $913.9M of stock-based compensation in FY2025 against $5.7M the prior period (Rubrik Form S-1). The adjusted figure management hands you is the reason to read the footnote yourself.
• Mix decides the margin. ServiceTitan ran roughly 70% subscription and 25% usage revenue (Form S-1). Adobe posts 89.3% gross margin and Snowflake 67% (their FY2025 10-Ks). Cost structure under the revenue sets the profit ceiling before management allocates a dollar.
• Concentration breaks the narrative. Klaviyo drew 77.5% of ARR from Shopify and CoreWeave 62% of 2024 revenue from Microsoft (their Forms S-1). Platform dependence is the risk the TAM slide hides.

The SaaS IPO Scorecard: four non-negotiable metrics

The clearest way to analyze a SaaS IPO is to fix the tests before you open the document. Call this the SaaS IPO Scorecard: five tests across the four dimensions that decide whether the unit economics work, each with what to read and a pass/fail line you can apply mechanically. For a worked example of reading one company’s revenue mix this way, see how the payments engine dominates how Shopify makes money.

The point of naming it is reuse. Run any SaaS S-1 through the same five rows and you get a consistent operator-grade triage in under an hour, before the narrative has a chance to anchor you. This scorecard is an original analytical asset. The figures used to illustrate each row are sourced to the filings cited throughout this piece.

#	Test (dimension)	What to read	Pass line	Fail / yellow flag
1	Growth quality	MD&A revenue mix + growth rate; sales and marketing as a share of revenue	Durable growth with stable or falling S&M-to-revenue	Growth rate falling fast, or S&M rising faster than revenue (bought growth)
2	Retention	Net and gross dollar retention; cohort definition in footnotes	NRR above 110%; GRR disclosed and in the 90s	NRR omitted or trending down; GRR not disclosed at all
3	Margin / path to profit	Gross margin trend by stream; GAAP loss vs adjusted EBITDA gap; SBC	Gross margin stable; the GAAP-to-adjusted gap is mostly real SBC you can size	Margin compressing; adjusted profit that evaporates once SBC is read in
4	Valuation vs profile	Implied EV/Revenue at the range vs growth + margin + retention	Multiple consistent with peers at the same growth and margin	Premium multiple unsupported by the growth, margin, or retention profile
5	Risk factors	Customer and platform concentration; the specific (not boilerplate) risks	Diversified base; no single ecosystem above ~20% of revenue	One customer or platform above ~20%; a structural risk written to look generic

Read in this order, the filing triages itself. Tests 1 through 3 tell you whether the business model works. Test 4 tells you whether the price respects what tests 1 through 3 found. Test 5 tells you what management does not control and is quietly worried about.

The banker narrative inverts this. It opens with vision and addressable market, then routes the hard numbers into a key-metrics box you are meant to skim. The scorecard reverses the priority on purpose. The same surface-over-story discipline runs through how to read a tech S-1 like an operator, which is the read-order companion to this scoring pass.

Growth quality: organic vs bought (the revenue mix test)

The first test asks a single question: is the growth efficient or bought? Open the MD&A revenue table and the sales-and-marketing line before anything else. Revenue mix tells you which engine drives the business and which one drives the margin, and the two are often not the same.

ServiceTitan’s Form S-1 (twelve months ended July 31, 2024) disclosed $685M of ARR growing 25% year over year, with roughly 70% subscription revenue and about 25% usage revenue. That mix is the tell. Subscription carries the highest contribution margin; if usage or services grew faster than subscription, blended gross margin would compress even as the top line accelerated.

Contrast that with CoreWeave’s Form S-1 (FY2024), which reported $1.9 billion of revenue, up from $229M in 2023, a 737% increase. The growth is spectacular, but the mix is entirely GPU-infrastructure revenue, the highest-cost line in the comparison set. A 737% growth rate and a thin, capital-intensive margin profile are both true at once. Growth rate alone tells you nothing about whether the model converts to profit.

Efficient growth and bought growth look identical on a top-line chart. The difference is in the cost to produce the next dollar:

• Efficient growth expands existing accounts (high net retention) and adds new logos at falling or flat acquisition cost. Revenue compounds because the base does.
• Bought growth leans on new-logo spend at rising acquisition cost, often masking weak retention underneath. The chart looks the same; the unit economics do not.

The operator question is never how fast revenue grew. It is which revenue grew, and what that does to the margin and the acquisition cost. The mechanics of how seat versus usage pricing changes that retention math are worked through in usage-based pricing vs seat-based pricing.

Why is gross revenue retention more important than net revenue retention?

Gross retention strips out expansion and shows raw churn and downgrades, so it reveals whether the base is stable. Net revenue retention includes expansion, which can mask a leaky foundation. Strong net retention sitting on weak gross retention means expansion is the only thing covering the churn underneath, and that is fragile.

Net revenue retention is the headline number in a SaaS S-1, and the recent filings give a clean benchmark range:

Company (filing)	Retention disclosed	Period
Rubrik	133% subscription dollar-based NRR	as of Jan 31, 2024 (Form S-1)
Datadog	~120% trailing-12-month NRR; GDR mid-to-high 90s	Q2 2025 (Form 10-Q)
Klaviyo	119% NRR (114% in Q1 2024)	June 30, 2023 (Form S-1)
ServiceTitan	110%+ net dollar retention	across ten quarters (Form S-1)

Sources: Rubrik, Inc. Form S-1 (FY ended Jan 31, 2024); Datadog Inc. Form 10-Q (quarter ended June 30, 2025); Klaviyo, Inc. Form S-1 (period ended June 30, 2023); ServiceTitan, Inc. Form S-1 (twelve months ended July 31, 2024).

All four are above 100%, which means each cohort pays more over time even after churn. ServiceTitan’s revenue per active customer rising from roughly $72K to $78K (2023 to 2024, per its Form S-1) is the same expansion at the account level.

Here is the catch. About 5% of public SaaS companies disclose gross retention while 67% disclose net (Ordway Labs analysis). Datadog is one of the few recent filers to give you both, with gross dollar retention in the mid-to-high 90s (Form 10-Q, Q2 2025). When a candidate shows you a flattering NRR and stays silent on GRR, the opacity is the signal. The full distinction between the two numbers and why issuers prefer disclosing only one is the whole subject of gross retention vs net retention in SaaS IPOs.

Profitability path: gross margin and the adjusted-GAAP gap

The third test asks whether the path to profit is real or an accounting choice. Two numbers carry it: gross margin, which sets the ceiling, and the gap between the GAAP loss and the adjusted figure management prefers, which is where the burn hides.

Gross margin is structural, not a sign of who runs the better company. Adobe posts 89.3% gross margin and Snowflake 67% in their FY2025 10-Ks. Adobe ships software whose delivery cost barely moves as another customer signs on. Snowflake delivers data and compute, and the cloud bill rises with every query a customer runs. The mix decides the margin, which is why this connects directly to why gross margin is destiny in SaaS: the margin line caps every budget below it before management allocates a dollar.

What does stock-based compensation reveal about an IPO candidate?

Stock-based compensation reveals dilution and the true cost of labor that never hits cash. A company can post a thin GAAP loss while transferring enormous value to employees and founders ahead of public holders, and the adjusted EBITDA management wants you to use is precisely the figure that hides it.

Rubrik is the cleanest case. Its Form S-1 (FY ended Jan 31, 2024) shows stock-based compensation of $913.9M in FY2025 against just $5.7M the prior comparable period, with $125.7M of unrecognized SBC as of July 31, 2024. That jump is not the business scaling. It is restricted awards vesting on the IPO event.

Instacart’s Form S-1 shows the same pattern at consumer scale: roughly $2.6 billion of stock-based compensation in its first quarter as a public company (Q3 2023), against $764M of quarterly revenue. The SBC charge dwarfed the revenue line for that quarter.

The three indicators to watch:

1. SBC as a share of revenue, trending up. Klaviyo ran SBC at about 13% of revenue (Form S-1), a controlled level. A figure climbing year over year signals dilution outrunning growth.
2. Unrecognized SBC in the footnotes. A large balance is future dilution already promised, telling you what the next several years of the income statement will absorb.
3. An IPO-event spike. A one-time jump tied to vesting on the offering, like Rubrik’s and Instacart’s, is the moment insiders realize value before public shareholders see it flow through earnings.

The full pattern of how filings stage and footnote this charge is dissected in stock-based compensation in tech IPOs. The operator reads SBC in, because the dilution is real even when the cash is not.

Valuation vs profile: the revenue multiple in context

The fourth test asks whether the price respects the economics the first three tests found. An EV/Revenue multiple is not high or low in isolation. It is high or low relative to the growth rate, the margin, and the retention behind it.

A clean way to keep this honest is to refuse to judge the multiple until you have scored the profile. A 15x forward revenue multiple is reasonable for a business growing 40% at 80% gross margin with 130% net retention. The same multiple on a business growing 20% at 65% margin with undisclosed gross retention is a different proposition entirely. The number is identical; the thing it prices is not.

The discipline is to build the profile first, then ask what multiple it justifies against peers at the same growth and margin:

Profile input	Where to read it	What it does to the justified multiple
Growth rate + durability	MD&A revenue trend	Faster, more durable growth supports a higher multiple
Gross margin	Income statement, by stream	Higher margin converts more revenue to cash, lifting the multiple
Net + gross retention	Key metrics + footnotes	Higher retention lowers the cost of future growth, lifting the multiple
SBC + dilution	Income statement + footnotes	Heavy dilution lowers the per-share value the multiple implies

The Rule of 40 (growth rate plus profit margin) is a fast first screen for whether a profile earns its multiple. The point of test 4 is not to output a target. It is to check whether the range prices the economics accurately or asks the public to pay for a story tests 1 through 3 did not support.

The five-pass framework: how to analyze a SaaS IPO on every filing

How should operators use the five-test pass on a SaaS IPO?

Run the tests in order: revenue mix and growth rate from the MD&A, net and gross retention, SBC as a share of revenue plus the unrecognized balance, customer and platform concentration, then the specific risk factors. Answer each mechanically before reading the narrative. Form your thesis on the numbers, then read the prose to see whether it confirms or contradicts what you found.

That order is the entire method. It separates signal from narrative before the narrative can frame your judgment. A banker leads with the story so every number you see afterward is interpreted through it. The five-pass framework anchors you on the unit economics first.

Methodology: scoring a SaaS IPO in under an hour

• Inputs: the MD&A revenue table, the key-metrics section, the income statement and SBC footnote, the customer-concentration disclosure, the risk-factor headings, and the implied valuation at the range. Six sources, read in scorecard order.
• Assumptions: that disclosed retention and ARR figures use a stable, issuer-defined methodology across periods, and that the most recent period is representative. Both are checked in the footnotes, not assumed.
• Sensitivity: a single point of NRR, a few points of mix shift toward low-margin revenue, a 10-point move in top-customer concentration, or a few turns of multiple each change the read materially. Treat every metric as a range and a trend, not a fixed point.
• What this misses: issuer-defined metrics can be gamed within the rules, segment-level margin is rarely broken out, and a snapshot S-1 cannot show whether a concentration is diversifying. The cross-check against later filings is the only fix.

A worked scorecard run: ServiceTitan

Run the five tests on ServiceTitan’s Form S-1 (twelve months ended July 31, 2024) to see the method in motion. The figures below are all from that filing; the scoring framing is illustrative, not a recommendation.

Test	What the filing showed	Read
1 Growth quality	$685M ARR, 25% YoY; 70% subscription / 25% usage mix	Durable, subscription-led growth. Pass, watch the usage mix shift
2 Retention	110%+ net dollar retention across ten quarters; RPAC $72K to $78K	Consistent expansion; GRR not separately broken out. Pass with a footnote caveat
3 Margin / profit path	Subscription-weighted mix supports a healthy software margin; read SBC in	Mix supports margin; the burn-vs-adjusted gap is the line to size next
4 Valuation vs profile	25% growth + subscription margin + 110%+ NRR	A premium multiple needs that retention to hold; the profile supports a software multiple, not an infrastructure one
5 Risk factors	Top 10 customers about 10% of revenue	Diversified base, no single-customer dependence. Pass

ServiceTitan scores well on a mechanical read: subscription-led growth, consistent expansion, a diversified base. The one open question the scorecard surfaces is gross retention, which the filing does not break out separately. That open question is exactly what the method is for. It tells you the three things to dig into before the narrative tells you everything is fine.

Bear case: what the scorecard misses

The strongest argument against this scorecard is that it over-trusts disclosed metrics and under-weights the qualitative story, and the skeptics get something real here.

Disclosed metrics are management-chosen definitions. Net revenue retention is not a GAAP term, so companies define the cohort, the window, and the inclusions themselves. A 119% NRR computed on a favorably drawn cohort is not comparable to a 119% computed conservatively, and the S-1 does not always give you enough to normalize them. The same caution applies to ARR, active customers, and adjusted gross margin, all issuer-defined.

The bear case also notes that the scorecard can talk a reader out of a genuinely great business. A high-concentration filing looks fragile on test 5, but concentration is how category-defining companies often start. CoreWeave’s 62% Microsoft dependence is a risk and the reason the revenue exists at all. Reading concentration purely as a red flag would have flagged many companies that diversified successfully from exactly that position. Concentration is a risk and a growth vector at the same time, and the scorecard does not weigh the second.

Here is the honest weighing. The scorecard is the right triage because it anchors on what management is legally constrained to disclose, but it is a filter, not a verdict. It tells you where to look harder. Reading the definitions in the footnotes, judging whether a concentration is a launchpad or a leash, and sizing whether a structural risk is near-term or distant is the part no scorecard automates. Use the five tests to find the three things that matter in a 300-page document, then do the slow work on those three.

What operators should take from this

If you read filings as a founder, operator, or analyst, the transferable discipline is the order of operations and the questions each test forces. Here is the playbook, six concrete moves you can run on the next SaaS S-1 that drops.

1. Run the five-test pass before reading the narrative. Score growth quality, retention, the profit path, valuation, and risk first. Form your thesis on the numbers, then read the prose to see whether it confirms or contradicts what you found.
2. Separate gross retention from net. Find the gross dollar retention if it exists; if it does not, treat the silence as the answer. Net retention sitting on undisclosed gross retention is expansion covering churn you cannot see.
3. Read SBC into the model, never out of it. Treat stock-based compensation as a real cost and track the unrecognized balance as future dilution. The adjusted figure management prefers is the one to distrust.
4. Refuse to judge the multiple until you have the profile. Build growth, margin, and retention first, then ask what multiple peers at that profile carry. A number without a profile behind it is not analysis.
5. Filter risk factors for specificity. Highlight only sentences that name a counterparty, a dependency, or a mechanism unique to this issuer. Klaviyo’s Shopify dependence and CoreWeave’s Microsoft concentration are the specific risks; the rest is liability boilerplate.
6. Cross-check against the first 10-Q. An S-1 is the pitch; the first post-IPO filing is the proof. Re-run the same five tests and watch the deltas: did NRR hold, did SBC normalize after the spike, did concentration improve.

That sixth move is the one most readers skip. The S-1 is a snapshot; the subsequent filings are where the truth lives. Datadog and Snowflake, two post-IPO benchmarks cited above, are useful precisely because their public filings let you watch retention and margin evolve over several years, well past the single frame an S-1 freezes.

Where this framework is vulnerable to management gaming

A method this clean deserves its own counterexamples, because management knows the scorecard too and can stage the filing around it.

Metrics can be defined favorably. NRR, ARR, and adjusted gross margin are issuer-defined. A company can choose the cohort window, the customer definition, and the COGS classification that flatter the number. The scorecard tells you which metric to find; it does not guarantee the metric is computed the way you would compute it. This is why gross retention disclosure matters so much: it is harder to dress up than net.

Timing can be engineered. A company can pull a strong quarter into the S-1 window or push a soft one out, so the trailing figures look better than the run-rate. The single-snapshot nature of an S-1 is exactly why the cross-check against the first 10-Q is non-negotiable.

Concentration can be temporarily masked. An issuer can sign a late, large customer to dilute a concentration percentage just before filing, or structure reseller relationships to spread revenue across nominal accounts that share an ultimate parent. The disclosed top-10 figure can understate the true dependency. The full anatomy of how this hides in filings is covered in customer concentration risk in SaaS filings.

Risk factors can hide the real risk in plain sight. The most material risk is sometimes written in the same flat tone as the boilerplate, precisely so it does not stand out. Specificity is the filter, but a careful drafter can make a structural risk read like a generic one.

None of this breaks the method. It bounds it. The SaaS IPO Scorecard is the fastest way to find the three things in a filing that decide the business. What it cannot do is the slow, skeptical, footnote-level work on those three. That part is still the operator’s, and it is the part that does not scale.

---

Analysis, not investment advice. Figures are drawn from the public SEC filings cited inline by company and form type (Forms S-1 for ServiceTitan, Rubrik, Klaviyo, Instacart, and CoreWeave; Forms 10-K and 10-Q for Snowflake, Adobe, and Datadog). Retention-disclosure prevalence is from the Ordway Labs analysis. Frameworks here are for understanding how to read business filings and tradeoffs, not for making buy or sell decisions.

Want the full toolkit for analyzing a SaaS IPO, the SaaS IPO Scorecard, the retention and SBC worksheets, and the valuation-profile model used above? It’s in the Tech Business Analysis Playbook.