September 14, 2026
Day91 Labs Brief — Issue 006: B2B SaaS
Six companies reported, one repriced, and last week's issue got two things wrong. Almost every headline this week needed a second number beside it to mean anything.
1. Sprinklr's 10-Q is out. It lengthened how long it spreads commission costs, citing better retention, in the same half its cash from operations fell 25%.
Evidence tier A — 10-Q opened. This entry closes one open item from Issue 005 and corrects two statements in it.
Sprinklr filed its 10-Q for the quarter ended 31 July 2026 on 3 September. Remaining performance obligations were $1,026.8 million, of which $614.1 million is expected to be recognised in the next twelve months, leaving $412.7 million beyond. The filing discloses that the expected period of benefit for capitalised commissions on new and upsell agreements was increased from five years to six, effective 1 February 2026, on the basis of improved customer retention; the change lowered sales and marketing amortisation by $0.5 million in the quarter. Amortisation of contract acquisition costs was nonetheless $20.6 million against $13.0 million a year earlier, and capitalised costs rose to $183.0 million from $173.0 million in January. Six-month operating cash flow was $88.5 million against $118.6 million. Derived from the cash-flow statement: six-month calculated billings (revenue plus the change in deferred revenue) were $399.6 million against $404.4 million, down 1.2%. The accelerated share repurchase settled on 1 September at 22,118,413 shares for $125 million, an average of $5.65.
Two things. First, the longer amortisation period is justified by retention, while the same company reported subscription net dollar expansion of 102% last week. A longer life on capitalised commissions spreads cost further out and flatters near-term margin; it is a judgement, and the filing states it plainly. Issue 003 found a commission-amortisation contradiction at Thryv. This is the same line of the filing, and it is worth reading at every vendor. Second, the half-year billings figure does not confirm or refute the third-party Q2 billings number carried last week. That figure implies Q1 billings grew about 6%, which is possible, but the 10-Q alone cannot isolate Q2.
Add the commission amortisation period to your competitor file and check it annually. A vendor that lengthens it while cash from operations falls is making its income statement look steadier than its cash. Ask whether the retention evidence behind the change is disclosed; here it is asserted, not quantified. Correction to Issue 005: beyond-twelve-month RPO was $412.7 million, not roughly $416 million. The contract-duration remarks were attributed to 'CFO Manish Sarin' from third-party call coverage, but the 10-Q's Section 906 certification is signed by Anthony Coletta as Chief Financial Officer. The remarks should have been attributed to management until the speaker was confirmed from the transcript.
Sources
2. Oracle's backlog grew $26 billion in a quarter it says it signed more than $30 billion of AI contracts. Revenue is the difference, and the roll-forward is not published.
Evidence tier A — 10-Q opened. The backlog-additions arithmetic is ours because Oracle does not publish the roll-forward.
Oracle's Q1 FY27, quarter ended 31 August 2026: revenue of $19.35 billion, up about 30%. Cloud revenue $11.6 billion, up 62%; cloud infrastructure $7.4 billion, up 121%; cloud applications up 10%. Software revenue $5.55 billion, down about 3%. Remaining performance obligations of $664 billion, up $209 billion year over year and $26 billion from $638 billion in June. Management said more than $30 billion of new AI cloud contracts were signed in the quarter, most structured as prepayments, bring-your-own-hardware or similar. The 10-Q states customer prepayments carrying a significant financing component were $11.4 billion. Operating cash flow was $23.1 billion and capital expenditure was $28.5 billion, confirming free cash flow of negative $5.4 billion. Derived: if RPO were reduced only by revenue recognised, gross additions in the quarter were at most about $45 billion. Oracle does not publish the roll-forward, so cancellations, currency and revenue outside RPO cannot be separated.
Headline backlog growth of $26 billion is net of about $19 billion of revenue coming out the other side. The more useful disclosure is structural: the customer is increasingly paying for the capacity before it is built. That moves financing risk from Oracle to the buyer and shows up as $11.4 billion of prepayment offsetting $28.5 billion of capex. It also means operating cash flow this quarter was partly a customer's balance sheet. Meanwhile the legacy software line — the one that looks most like your business — shrank.
Treat any vendor's backlog figure as a stock, not a flow. Ask for the additions. If you are a buyer being offered capacity on prepayment terms, you are now part of your vendor's capital structure; price that like credit risk, not like a discount.
Sources
3. Adobe's own full-year guide implies its smallest net new ARR quarter of the year is the one that is usually its largest.
Evidence tier A — 8-K exhibit opened for ARR, RPO and cRPO share. The implied Q4 figure is ours. Two analyst characterisations are marked as third-party.
Adobe's Q3 FY26, quarter ended 28 August 2026: revenue of $6.76 billion, up 13%. Total Adobe ARR of $27.50 billion, up 11.2%. RPO of $22.16 billion, up 8%, with current RPO at 67% of the total and growing 9%. Operating cash flow a Q3 record at $2.52 billion; about 9.5 million shares repurchased. AI-first ending ARR exceeded $650 million. The full-year target is total ARR growth of 10.2% against a beginning book of $25.66 billion. Derived: that implies ending ARR of about $28.28 billion, so Q4 net new ARR of about $0.78 billion, against $1.84 billion added across the first three quarters. Interim CFO Steve Day said RPO and current RPO typically step up in the fourth quarter. Analysts on the call cited a 36–37% year-over-year decline in net new ARR and the first single-digit RPO growth since early fiscal 2023; neither figure is in the release located. Anil Chakravarthy becomes CEO on 1 December.
Management's seasonal defence is reasonable for RPO, which does bunch into Q4 renewals. It is harder to apply to ARR, because the company's own guide already embeds Q4 and still lands at about $0.78 billion. The freemium pivot is the stated cause: more users, less paid ARR per user in the near term. Revenue at 13%, ARR at 11%, RPO at 8% — the further out the metric, the slower the growth. That is the GitLab ranking from Issue 005, running in reverse order.
If you compete with a vendor moving to freemium, expect price pressure at the low end before you see their ARR recover. Back-solve the implied final quarter from any full-year ARR guide; it is one subtraction and it is rarely reported.
Sources
4. Braze's near-term backlog grew slower than its revenue, and it is halfway through moving customers to new pricing.
Evidence tier A — Release read for retention and customers; RPO and organic growth from the call.
Braze's Q2 FY27, quarter ended 31 July 2026: revenue of $227.2 million, up 26%; management gave organic growth of 24% excluding $6.6 million from BrazeAI Decisioning Studio. Total RPO of $1.1 billion, up 27%; current RPO $691 million, up 24%. Trailing twelve-month dollar-based net retention of 110%, against 108% a year earlier; 112% for customers above $500,000 of ARR. Those 361 large customers are 65% of total ARR, against 62%. Non-GAAP gross margin of 68.6%, against 69.3%, attributed to premium messaging volumes and Decisioning Studio headcount. Management attributed current-RPO deceleration to lapping the OfferFit acquisition and renewal-cohort seasonality, and said the company is roughly 50% through migrating customers to new pricing and packaging, which is reclassifying some revenue. Derived: 24% organic growth only reconciles if the prior-year base also contained about $2.1 million of acquired revenue. Shares fell about 12% after hours.
Third trailing-twelve-month retention figure in two issues, after Asana. It smooths the pricing migration out of view: a customer moved to new packaging mid-year shows up in that number slowly. The more current signals are the ones that moved — near-term backlog below revenue growth, gross margin down on message volume, and concentration up three points. A mid-migration pricing change is exactly when those three should be watched together.
If you are running your own repricing, publish or at least track the share of the base migrated, and read net retention only for the migrated cohort. Blended, it will tell you nothing until the migration is over.
Sources
5. SailPoint's ARR grew 25% and revenue 17%. Almost none of the new ARR came from anything but SaaS.
Evidence tier B — The wire release is a pointer to the IR site; the full release was not opened. Figures from call coverage; the non-SaaS residual is ours and sensitive to rounding.
SailPoint's Q2 FY27, quarter ended 31 July 2026: ARR of $1.231 billion, up 25% and $11 million above the guidance midpoint. SaaS ARR of $847 million, up 36%, and 97% of net new ARR. Net new SaaS ARR of $66 million, up 34%. Revenue of $308.8 million, up 17%. Net revenue retention of 113%. AI-driven ARR above $70 million and more than 30% of net new ARR. Management cited a roughly $5 million revenue timing headwind from the SaaS mix. Full-year ARR guidance was raised by $11 million to $1.38 billion; full-year revenue guidance of about $1.27 billion was not raised. Derived: total net new ARR of roughly $68 million, leaving about $2 million from term, maintenance and other subscription.
An eight-point gap between ARR and revenue growth, explained by the company: term licences recognise a large share up front, SaaS recognises ratably, so moving the mix defers revenue. That makes ARR the better demand read here, which is the opposite of the GitLab case. It also means the non-SaaS base is roughly flat, so the migration is doing most of the work. The AI share of net new ARR — more than 30% — is the figure to watch; it is small in dollars and large at the margin.
When a vendor's ARR and revenue diverge, find out which way the mix is moving before deciding which to trust. Upfront-to-ratable depresses revenue; ratable-to-consumption can inflate ARR. Same gap, opposite meaning.
Sources
6. Navan's booking volume grew ten points faster than the revenue it earns on it.
Evidence tier A — Release read. Take rate is derived from two disclosed figures.
Navan's Q2 FY27, quarter ended 31 July 2026: total revenue of $233 million, up 35%. Usage revenue of $211 million, up 35%; subscription revenue of $21 million, up 39%. Gross booking volume of $3.0 billion, up 45%. Payment volume of $1.3 billion, up 34%. GAAP gross margin 74%. Full-year revenue guidance raised to $927–933 million, about 32% growth; Q3 guided at $253–255 million, about 30%. Derived: usage revenue was about 7.03% of gross booking volume, against about 7.55% a year earlier — a decline of roughly 52 basis points.
A usage-priced business whose volume outgrows its revenue is taking a smaller cut per unit. That can be deliberate — larger customers negotiate lower rates, and Navan is moving up-market — or it can be mix. The release does not say which. Either way, it is the metric that tells you what usage-based pricing does at scale: the unit rate compresses as the customer grows.
If you are moving to usage pricing, model the rate curve, not just the volume curve. Your largest customers will ask for Navan's trajectory. Build the step-downs into the rate card before they are negotiated one deal at a time.
Sources
7. Atlassian added a usage meter on top of every seat. First primary-source pricing change in three weeks of looking.
Evidence tier A — Atlassian's own pages read. Details marked as partner-reported are not on the pages opened.
Atlassian announced on 1 September 2026 an expansion of usage-based pricing across its cloud products. Metered capabilities — including Rovo credits for AI interactions, automation steps and Assets objects — now sit alongside seat subscriptions as a separate usage layer. Most paid cloud plans include a monthly allowance, pooled at the organisation level and sized to plan and seat count; admins can adjust limits. Billing for usage beyond allowance begins on 3 December 2026 for most meters. Atlassian's usage page states that the published rates apply to usage and forecasts from 31 August 2026. Partner-reported, not on the pages read: automation is now counted in steps rather than runs, Enterprise plans lose unlimited automation, Collections carry allowances up to roughly ten times higher, one meter charges only for fully AI-resolved tickets, and separate cloud seat-price changes take effect on 13 October.
Seats are not being replaced; they are becoming the allowance multiplier for a consumption meter. That is the hybrid shape Asana described in Issue 005 and Microsoft applied to GitHub Copilot in June. The pooled allowance is the clever part: it makes the first bill look unchanged for most customers while establishing a per-unit rate for everything above it. The bundle incentive — larger allowances in Collections — turns the meter into an upsell lever.
If you buy Atlassian, pull your meter usage from the admin console now; the data has been populating since 1 September and you have until December. If you sell against Atlassian, the allowance ceiling is your pricing talking point, but only with a customer's actual usage in hand.
Sources
8. UiPath added $37 million of net new ARR. Its full-year guide needs $73 million in Q4.
Evidence tier A — Company guidance from the release and presentation; the quarterly split is ours.
UiPath's Q2 FY27, quarter ended 31 July 2026, reported 3 September: revenue of $410 million, up 13%. ARR of $1.938 billion, up 12%. Net new ARR of $37 million. Dollar-based net retention 109%. Cloud ARR of $1.3 billion, up 19%. GAAP operating income of $32 million. Q3 guidance: ARR of $1.992–1.997 billion. Full-year ARR guidance: $2.065–2.070 billion. Derived: second-half net new ARR of $127–132 million; Q3 of $54–59 million; and at guidance midpoints, Q4 of about $73 million — roughly twice the quarter just reported.
Missed in Issue 005's window and included here because the shape matters. A back-half-weighted guide is normal for an enterprise vendor with a January year-end. A doubling from Q2 to Q4 is aggressive for a business growing ARR at 12%. Management framed the outlook as prudent. The quarterly split is the test of that word.
Do the same back-solve on your own plan: if Q4 carries more than a third of the year's net new ARR, write down what has to be true in pipeline today for it to land, and check it in October.
Sources
9. Microsoft's recast shows its new big segment earns about 51 cents of operating profit per dollar. The small one's profit fell faster than its revenue.
Evidence tier A — 8-K presentation opened for the recast segment tables. Promotes Issue 005 entry 5 from B to A for the segment facts. The $29.42 billion Azure quarter is not in the section read and stays as reported.
Microsoft's 2 September 8-K furnishes two years of recast results under the new structure. Agents and Infra: fiscal 2026 revenue of $268.1 billion and operating income of $136.4 billion, a margin of 50.9%; June-quarter margin of 49.3%, against 49.7% a year earlier. Devices and Consumer: fiscal 2026 revenue of $63.7 billion and operating income of $18.9 billion; June-quarter operating income of $3.87 billion, down 5.6%, on revenue of $15.43 billion, down 1.5%. The two segments reconcile to June-quarter revenue of $90.0 billion and fiscal-year revenue of $331.8 billion. Separately, on its fiscal Q4 call Microsoft said GitHub Copilot moved to usage-based pricing in June and that segment margins improved through the quarter after the change.
Issue 005 noted that segment cost and margin disclosure for the three old segments disappears. The recast replaces it with one very large blended margin: Azure, Microsoft 365, server products and GitHub in a single line. Investors gain an Azure revenue number and lose the ability to see whether infrastructure or applications carry the profit. The Copilot remark is the more operator-relevant sentence: a seat product was moved to usage pricing because usage was eroding margin.
Log the Copilot change as a reference case. If a product's cost scales with usage and its price scales with seats, the margin tells you before the customer does. Watch your own gross margin by heavy-usage cohort.
Sources
10. Two days after Oracle reported, a syndicated summary had its backlog at $64 billion. The real figure is ten times that.
Evidence tier C — No document supports these figures. Carried to show what the wrong version looks like, not as a signal.
A news summary published after Oracle's results reported cloud infrastructure revenue of $4.4 billion, more than $10 billion of new AI contracts, and total remaining performance obligations of $64 billion for Q1 FY27. Every other account located — the call transcript and three independent reports — gives $7.4 billion, more than $30 billion, and $664 billion. The $4.4 billion figure is close to Oracle's prior-year quarter; the $64 billion figure is a dropped digit.
This is the failure mode the method exists for. The page is well formatted, dated, and states its numbers with confidence. A retrieval system that finds it first will repeat it. Nothing in the text signals that it is wrong except disagreement with the primary record.
When a number is surprising, check it against a second source that did not derive from the first — and prefer the one closest to the filing. If your team uses an AI research tool for competitive briefs, spot-check one figure per brief against the document it claims to summarise.
Sources
Every claim links to its source. Spotted something wrong? It will be fixed and logged on the corrections page. Also published on Beehiiv.