Day91 Brief — Issue 007

A · read — 10-Q opened. Promotes Issue 006 entry 2 from B to A and corrects one derived figure in it. Bucket values and the year-end comparison are ours; the 12% year-end share is from the Q4 call.

FACT. Oracle filed its 10-Q for the quarter ended 31 August 2026. Remaining performance obligations were $664 billion, of which about 13% is expected to be recognised in the next twelve months, 37% in months 13 to 36, 34% in months 37 to 60 and the remainder thereafter. Derived: roughly $86 billion in the next year, $246 billion in years two and three, $226 billion in years four and five, and $106 billion after year five. Management gave 12% for the next-twelve-month share at year-end in June; derived, the near-term portion rose by about $10 billion in the quarter. The filing states Oracle received $11.4 billion of customer prepayments carrying a significant financing component in the quarter, against none a year earlier; total deferred revenue doubled to $30.8 billion from $15.4 billion. Operating cash flow was $23.1 billion and capital expenditure $28.5 billion. Oracle also fully used its at-the-market programme, issuing 141 million shares for $19.9 billion. Off-balance-sheet lease commitments not yet commenced were $288 billion, against $260 billion at 31 May, substantially all data centres, with terms of fifteen to nineteen years. The same filing describes Oracle Cloud arrangements as generally one to five years.

READ. Three things in one document. First, the backlog is mostly a five-year stock: the next twelve months equals about 1.1 times the current quarterly revenue run-rate, not eight. Second, about half the quarter's operating cash came from customers paying before delivery; without the $11.4 billion, operating cash flow would have been about $11.7 billion and free cash flow about negative $16.8 billion. The gap was filled with a $19.9 billion share sale. Third, the durations do not match: contracts generally run up to five years, and 84% of the backlog lands inside five, while the leases committed to serve them run three to four times longer. Oracle has not said how the lease tail is covered after the contracts end, and nothing in the filing says it is a problem. It is a question the filing raises and does not answer.

SO WHAT. For any vendor quoting backlog, ask for the next-twelve-month share before the total. For a buyer offered capacity on prepayment, the 10-Q confirms Oracle now accounts for that prepayment as financing it receives from you. Negotiate it like a loan: delivery dates, remedies and what happens to the cash if capacity slips.

→ Oracle 10-Q, quarter ended 31 August 2026 — SEC → Oracle 10-K, fiscal 2026 — lease commitments at 31 May — SEC → Oracle Q4 FY26 call — 12% next-twelve-month share

A · read — 10-Q opened. Ratios are ours.

FACT. In the same 10-Q: sales and marketing expense was $1.81 billion, against $2.06 billion a year earlier, down 12%, while total revenue rose 30% to $19.35 billion. Within the cloud and software segment, sales and marketing fell 13% to $1.56 billion. Derived: sales and marketing fell from 13.8% of revenue to 9.4%. The segment's direct margin nonetheless fell from 59.6% to 54.5%, as cloud and software delivery costs rose to $6.2 billion from $3.4 billion. Software licence revenue fell 14.5% to $655 million; software support fell 1.2% to $4.9 billion. The filing describes the fiscal 2026 restructuring plan as including the adoption and integration of AI technologies across certain functions, with estimated costs of up to $2.1 billion, and states that after quarter-end management supplemented the plan by about $700 million.

READ. This is not a like-for-like efficiency story. Infrastructure revenue arrives in very large contracts with a handful of buyers and does not need a sales force in proportion to its dollars; the licence line, which does, shrank. What the filing does show is the direction: selling cost is falling in absolute terms, the restructuring budget is rising, and the stated reason includes AI in functions. For a revenue team, that is the vendor telling you where the next cut lands.

SO WHAT. When a vendor's selling cost falls while revenue grows, check the mix before crediting productivity. If you sell alongside Oracle's applications or licence business, expect fewer field resources on your joint accounts; plan coverage accordingly.

→ Oracle 10-Q — income statement, segment note, restructuring note — SEC

A · read — RPO figures from three SEC documents. The Q2 ARR figure is from a filing summary and marked secondary. Adobe's Q3 10-Q, due in September, was not located.

FACT. Adobe's remaining performance obligations were $22.52 billion at fiscal year-end on 28 November 2025, $22.27 billion at 29 May 2026 and $22.16 billion at 28 August 2026. Derived: down $0.36 billion, or 1.6%, year to date, and down $0.11 billion in the third quarter. Year over year, RPO grew 8.4% from $20.44 billion. Over the same nine months, total Adobe ARR rose from $25.66 billion to $27.50 billion, up 7.2%. Secondary: Q2 ARR was reported at $27.10 billion, implying Q3 net new ARR of about $0.40 billion. Last year, RPO rose $2.08 billion in the fourth quarter. Derived: to hold 8% year-over-year RPO growth at this year-end, Q4 needs to add about $2.16 billion.

READ. Issue 006 reported management's view that RPO steps up in Q4. Last year's $2.08 billion Q4 increase supports that. It does not explain why ARR and contracted backlog moved in opposite directions for nine months. The likely reading — ours, not Adobe's — is the freemium and monthly-plan push: it adds annualised revenue without adding contracted duration. That is a softer kind of ARR. The Q4 number is now a clean test with a stated threshold.

SO WHAT. When a competitor's ARR rises while its RPO does not, ask what share of its new ARR sits on monthly or cancellable terms. Track the $2.16 billion threshold when Adobe reports in December.

→ Adobe 8-K Ex. 99.1, Q3 FY26 — SEC → Adobe 10-Q, quarter ended 29 May 2026 — RPO — SEC → Adobe 8-K Ex. 99.1, Q4 FY25 — year-end RPO — SEC → Adobe 8-K Ex. 99.1, Q3 FY25 — prior-year RPO — SEC → Adobe Q2 FY26 ARR — filing summary (secondary)

A · read — Q1 10-Q opened. Closes the open item from Issues 005 and 006. The quarterly billings split is ours.

FACT. Sprinklr's 10-Q for the quarter ended 30 April 2026, filed 4 June, gives first-quarter revenue of $219.5 million and a $7.2 million decrease in deferred revenue, against $205.5 million and $1.9 million a year earlier. Derived: Q1 calculated billings of $212.3 million, up 4.3%. Combined with the six-month figures from the Q2 10-Q read last week: Q2 calculated billings of $187.3 million against $200.7 million, down 6.7%, on revenue up 0.8%. Current RPO was $618.8 million at 31 January, $627.1 million at 30 April and $614.1 million at 31 July. The Q1 10-Q also discloses that a large multinational customer was 10.4% of consolidated revenue in the quarter, attributed to the timing of revenue recognition, and that amortisation of contract acquisition costs was $18.0 million against $12.0 million.

READ. The direction in Issue 005 was right and the size was not. On the standard definition — revenue plus the change in deferred revenue — the decline is 6.7%. The third-party figure may use a different definition; we could not reproduce it, so it comes out of the record. The more useful new fact is the 10.4% customer: a single account crossed the disclosure threshold in the quarter before billings fell. The filing does not say whether it is the same customer, and we do not assume it.

SO WHAT. Check whether any competitor has a customer above 10% of revenue; it is disclosed and rarely read. At a vendor of Sprinklr's size, one renewal moves the quarter.

→ Sprinklr 10-Q, quarter ended 30 April 2026 — SEC → Sprinklr 10-Q, quarter ended 31 July 2026 — SEC

B · unpulled — The London Stock Exchange announcement was not opened. Figures are consistent across a reproduction of the release, the July trading update and two independent reports. The expansion arithmetic is ours.

FACT. Trustpilot reported first-half 2026 results on 15 September. Bookings of $171.2 million, up 22% reported and 18% at constant currency. Revenue of $151.4 million, up 23%, or 19% at constant currency. ARR of $313 million, up 15%, or 17% at constant currency. Last-twelve-month gross retention of 87%, against 86%; net dollar retention of 101%, against 103%. The company attributed the retention movement to completing the annualisation of previously flagged plan migrations from 2024 and 2025. ARR from customers paying more than $20,000 a year rose 36%; North American bookings rose 27% at constant currency, UK 11%. Full-year guidance was reiterated at high-teens revenue growth and mid-teens bookings growth at constant currency. Results included a $2.6 million restatement of a prior period for historical US sales tax. Shares fell more than 13% on the day.

READ. Net retention minus gross retention is the expansion the base delivered: 14 points this year against 17 last year. Gross churn improved, so the fall is all upsell. The company's own explanation is the tell: price increases from plan migrations lifted retention for a year and then rolled out of the twelve-month window. Separately, a mid-teens full-year bookings guide after 18% in the first half means the second half is guided slower. The growth is also concentrating — enterprise and North America — which is the same pattern Braze showed last week.

SO WHAT. If you migrated customers to new plans in the last year, strip the migration uplift out of net retention before you plan on it recurring. It will leave the metric on its anniversary whether you model it or not. Separately: if you sell into US states, the sales-tax restatement is a reminder to check your own nexus position.

→ Trustpilot H1 2026 results — release reproduction with KPI definitions → Trustpilot H1 2026 — revenue, ARR, retention, share reaction → Trustpilot H1 trading update, July 2026 — retention and migration note → Trustpilot H1 2026 call highlights — sales-tax restatement

B · unpulled — Atlassian's future-pricing page was not located in search. Consistent across six partner summaries; partners compared Atlassian's own current and future price tables. Stays at B until the page is read.

FACT. Partner-reported: Atlassian notified customers on 28 August of Cloud list-price increases effective 13 October 2026, applying to new purchases, renewals and upgrades. Increases range from about 3% to 10% by product, plan and seat count. For Jira, Confluence and Teamwork Collection on Standard and Premium, subscriptions below 5,000 seats rise 7% and those at 5,000 seats and above rise 10%. Service Collection and Jira Service Management rise about 7.5% on Standard and Premium and 10% on Enterprise; Jira Product Discovery about 5%; Atlassian Guard 3–5%. Quotes generated before 13 October are honoured until they expire. Existing monthly and auto-renewing annual subscriptions move to the new price on the first bill dated on or after 13 October. This is separate from the usage meters reported in Issue 006, which begin billing on 3 December.

READ. Most price books discount with volume. This one does the opposite at the 5,000-seat line: the largest customers take the larger increase. It fits the meter announced three days later — large organisations are where AI and automation usage concentrates, and the seat is becoming the allowance multiplier. Set beside Navan last week, where the largest customers pushed the unit rate down, the two show the choice every repricing vendor faces: concede to scale, or charge it.

SO WHAT. If you renew Atlassian above 5,000 seats, generate a quote before 13 October; it is honoured to expiry. If you are designing your own increase, a tiered rise that falls on your largest accounts needs a reason those accounts will accept. Atlassian's is the AI bundle; have yours ready before the renewal conversation, not during it.

→ Partner summary — tiers, 5,000-seat threshold, quote protection → Partner summary — per-product increases from Atlassian's tables → Partner summary — notification date and future-pricing page

B · unpulled — Market data from news coverage. No company document changed. Day91 takes no view on the substance of the essay.

FACT. On Monday 14 September, software stocks rose while semiconductor and AI-infrastructure shares fell, after a weekend essay by Anthropic's chief executive urged frontier AI labs to slow the rate at which model capabilities improve, and other lab leaders publicly agreed. The Nasdaq Composite closed down 0.56% after falling as much as 1.3%; the S&P 500 closed down 0.48%. Coverage reported ServiceNow up about 5% and Adobe up about 4% on the day. Going into the week, Meritech put the median public software multiple at 4.2 times implied ARR as of 11 September, down from 4.7 times at its previous update.

READ. Nothing in any software company's disclosures changed between Friday and Monday. The move was a repricing of a risk — how fast AI substitutes for software — not new information about any vendor's customers, pricing or retention. It connects to entry 1 in one place: Oracle's $288 billion of fifteen-to-nineteen-year leases are the commitments most exposed to how fast demand for frontier capacity grows. That exposure was in the 10-Q before the essay and is unchanged by it.

SO WHAT. Do not update competitive assessments on a price move with no document behind it. If your board asks what changed this week, the documents say: very little. Keep the question of pace on your list and watch for it in the next round of guidance, where it would have to show up in numbers.

→ Market close and software rally, 14 September — NBC News → Single-stock moves, 14 September — 24/7 Wall St. → Software multiples as of 11 September — Meritech

C · no document — The calendar listing is carried to show what the wrong version looks like. The correct date is from FactSet's own release.

FACT. A widely indexed earnings calendar listed FactSet for Thursday 17 September before the market opened, with an EPS estimate of $4.34. FactSet's own release of 2 September states that it will report fourth-quarter fiscal 2026 results on Wednesday 30 September, with the call at 9:00 a.m. Eastern. Last year FactSet reported on 18 September 2025. Our inference, not confirmed: the calendar carried forward last year's timing.

READ. A small error, and exactly the shape of the large ones. The listing is specific, well formatted and attached to a real estimate. A retrieval system asked when FactSet reports has two confident answers, and nothing in the text of either marks which one came from the company. Only one of them is the company.

SO WHAT. Take reporting dates from the company's own announcement, not an aggregator. If an AI tool feeds your competitive calendar, have it return the source link with each date.

→ FactSet — Q4 FY26 results date, 30 September → The calendar listing, for reference → FactSet Q4 FY25 release, 18 September 2025 — SEC

Every claim above carries the state of its own sourcing. Derived figures state the calculation and the inputs. If a link is wrong or a number is misread, reply and it is corrected in the next issue with the correction stated plainly, and logged publicly.

Day91 · Document-sourced revenue signals. AI-drafted, human-approved.