Backlog Conversion

When a company has signed orders it has not yet delivered, does that backlog turn into actual sales on schedule?

The planes that could not be delivered

At the start of 2019, Boeing looked like one of the safest businesses in the world. It had delivered 806 airliners in 2018, and its revenue had passed $100 billion.[^1] Airlines had already signed up for years of future planes: by the end of 2019, Boeing's commercial backlog held more than 5,400 airplanes worth about $377 billion.[^2] At 2018's pace, that was almost seven years of work already on the books.

Then the orders stopped turning into sales. After two crashes of the new 737 MAX killed 346 people, regulators grounded the plane in March 2019. The U.S. ban lasted 20 months.[^3] Boeing turns a plane order into revenue mainly when the plane is delivered, and it could not deliver. Deliveries fell from 806 in 2018 to 380 in 2019, then to 157 in 2020. Revenue fell from $101.1 billion to $58.2 billion.[^1] In the commercial airplanes business alone, revenue fell from $57.5 billion in 2018 to $16.2 billion in 2020, a drop of more than 70%.

Waiting was expensive. Airlines had planned their schedules around planes that did not arrive, so Boeing owed them compensation. In 2019 it took a charge of $8.3 billion for these "customer considerations," money paid or credited to customers for the delays.[^1] It kept building planes it could not hand over, so cash flowed out while little flowed in. Boeing's operating activities had brought in $15.3 billion of cash in 2018; in 2020 they used up $18.4 billion. To stay afloat, the company borrowed heavily. Its debt rose from $27.3 billion at the end of 2019 to $63.6 billion a year later.[^1] An investor who had looked only at the seven years of orders on the books would have missed all of this. The orders were still there. The money they promised had simply moved years into the future.

The backlog itself shrank. In 2020, with the pandemic emptying airports, customers cancelled orders worth $34.6 billion, mostly for the MAX, and Boeing removed another $54.5 billion of orders it no longer believed customers would complete.[^1] Together that was almost a quarter of the commercial backlog it started the year with. Boeing took 184 new orders that year, but its net orders came to about minus 1,026 planes.[^4] The commercial backlog fell from $376.6 billion to $281.6 billion in a single year.[^1]

The twist is that most of the backlog was real. By the end of 2025, Boeing's total backlog had reached a record $682 billion, including more than 6,100 commercial airplanes.[^5] Airlines still wanted the planes; they just had to wait. A backlog is a promise of future sales, but its size tells you little about when the promise will be kept, or whether all of it will be. That is the question this chapter is about.

What backlog is

A backlog (or order book) is the value of orders customers have signed that the company has not yet delivered. Imagine a bakery that makes wedding cakes. If couples have booked $100,000 of cakes for weddings later this year and next, the bakery has a $100,000 backlog. It earns that money, and records it as revenue, only as each cake is delivered.

A few related terms:

• Bookings are the new orders signed during a period. They add to the backlog.
• Book-to-bill is bookings divided by revenue. Above 1, the company is signing new work faster than it finishes old work, so the backlog grows. Below 1, it shrinks.
• Remaining performance obligations (RPO) is the accounting version of backlog. Under the U.S. revenue standard ASC 606, companies must disclose the value of contracted work not yet performed and roughly when they expect to recognize it as revenue. They may leave out contracts lasting a year or less and some usage-based fees, so RPO is not a complete list of future revenue.[^6]
• Backlog conversion is how quickly and reliably the backlog turns into revenue.

Since 2020, U.S. rules no longer require a specific backlog figure in annual reports; companies disclose it only if it is material.[^7] So "backlog" is defined by each company, while RPO follows an audited accounting rule. The two often differ.

Orders also come in different strengths, like promises between friends. A firm order is a signed contract with a price and a delivery date. An option is a customer's right to buy more later, which it may never use. A letter of intent is a statement of interest, not a contract. In government work, an order is funded once the lawmakers have approved the money and unfunded until then. Boeing, for example, splits its backlog into "contractual" orders and "unobligated" amounts, and it leaves out options and announced orders without signed contracts.[^1] A careful reader always asks which of these layers a company has included.

Software companies often report a slice called current RPO (cRPO): the part of RPO expected to become revenue within 12 months. It plays the same role as the near-term backlog of a manufacturer. Some of that money may already be in the bank. When a customer pays in advance, the company records deferred revenue (also called a contract liability), a debt of work it still owes. Deferred revenue is one piece of RPO; the rest is work under contract that has not yet been billed.

How to measure it

The backlog behaves like a bathtub. New orders pour in, deliveries drain out, and cancellations leak out:

Ending backlog = Starting backlog + New bookings − Revenue delivered − Cancellations

Conversion rate = Revenue delivered from the starting backlog ÷ Starting backlog

Book-to-bill = New bookings ÷ Revenue

Back to the bakery. It starts the year with a $100,000 backlog. It delivers $60,000 of those booked cakes, books $90,000 of new orders, delivers $15,000 of cakes ordered and made within the year, and loses $5,000 when two weddings are called off.

• Revenue is $60,000 + $15,000 = $75,000.
• Ending backlog is $100,000 + $90,000 − $75,000 − $5,000 = $110,000.
• Conversion rate is $60,000 ÷ $100,000 = 60%.
• Book-to-bill is $90,000 ÷ $75,000 = 1.2.

Investors also divide backlog by annual revenue to get coverage: here, about 1.5 years of revenue already booked. Many companies state how much of their backlog they expect to recognize within 12 months. That near-term slice is often the most useful number of all, because it ties the backlog directly to next year's revenue.

Two more measures help. The first is the cancellation rate:

Cancellation rate = Orders cancelled in the period ÷ Gross orders in the period

Gross orders are all new orders signed, before taking out cancellations; net orders are what is left after. If the bakery signs 20 new cake orders in a quarter and 3 other couples cancel, its cancellation rate is 3 ÷ 20 = 15%, and its net orders are 17.

The second is a backward test, which checks whether last year's promise came true. Suppose that a year ago the bakery said $80,000 of its backlog would turn into revenue within 12 months. This year it reports revenue of $75,000. Even if every dollar of that revenue had come from the old backlog, the bakery fell short of its own schedule, so some cakes slipped into the next year. If instead it had reported $120,000 of revenue, the old backlog could have been delivered in full, with $40,000 of revenue coming from orders signed and filled within the year. Companies rarely tell you exactly which dollars of revenue came from which orders, so this test is rough. But repeated year after year, it shows whether a company's schedules can be trusted. The worked example later in this chapter runs it on real numbers.

Why it matters

Revenue tells you what happened; backlog tells you what customers have already promised to buy. Research shows it carries real information, and that investors often misjudge it.

Shivaram Rajgopal, Terry Shevlin, and Mohan Venkatachalam studied 21,891 company-years of backlog disclosures from 1981 to 1999. Backlog did help predict future earnings, but the stock market behaved as if it predicted more than it really did: investors overpaid for big backlogs. Buying firms with low backlogs and selling short those with high backlogs earned abnormal returns of about 6% a year in their main test.[^8] (An abnormal return is the part of a stock's return that cannot be explained by the overall market and the usual risk factors.) Interestingly, the authors found the mistake was not the analysts': professional forecasters did build backlog into their earnings estimates. The overreaction came from the share price itself.[^8] A big order book is good news, just not as good as it looks.

Later work shows when backlog tells you the most. Rajiv Banker, Russell Barber, Dana Hollie, and Han-Up Park found in 2024 that backlog is an even better guide to future earnings when a company's sales have just fallen, when its cash takes longer to cycle through the business, and when its assets are growing fast. In those cases the order book shows that the dip may be temporary. But they also found a warning: a very high ratio of backlog to sales predicted lower future earnings.[^9] An order book that keeps swelling relative to what the company can deliver may be a sign that work is piling up, not that business is booming.

Backlog also creates temptations. Barber and Hollie found that companies with large backlogs, or large changes in their backlogs, were more likely to later restate (correct) the revenue they had reported. They suggested that auditors and investors should read such companies' accounts with extra care.[^10] A 2025 study by Thomas Gilliam, Frank Heflin, and Jeffrey Paterson found something subtler. When revenue was about to fall short of last year's level or of analysts' forecasts, managers appeared to speed up deliveries, filling orders early so they would count as revenue this quarter. The sign was an unusually small backlog left over at the end of the period, seen mainly in firms that beat their targets by a hair.[^11] Converting the backlog faster is not always good news. Sometimes it means next quarter's sales were pulled into this one.

New orders move prices too. Chipmaker shares moved significantly on the days a semiconductor trade group published industry-wide orders and shipments, and analysts used the data mainly to judge whether sales changes would last.[^12] But managers shape what you see: a 2024 study found they tend to give a precise book-to-bill number when the news is good and a vague description when it is bad.[^13]

Put together, the research gives a balanced picture. Backlog really does carry information about future sales and profits. Investors tend to get too excited about large backlogs. And because managers choose what to disclose and when to deliver, the figure needs checking against what actually happens. That checking is what this chapter calls backlog conversion.

The key questions investors ask

1. Pace: is the backlog converting on schedule?

Compare actual revenue with what the company said its backlog would produce, with its own history, and with analysts' forecasts. Conversion is improving when revenue arrives faster or more predictably without new bookings weakening; stable when timing matches the disclosed schedule; and deteriorating when deliveries slip, as Boeing's did in 2019 and 2020. Slippage often shows up first as a shrinking share of backlog due within 12 months, or as missed delivery targets.

Three checks make this concrete. First, run the backward test: did last year's 12-month slice show up in this year's revenue? Second, compare actual deliveries or revenue with the company's own guidance, which is the main yardstick for this thesis. Third, compare with what analysts expected, since the share price already reflects their forecasts. Remember the research above: unusually fast conversion right at the edge of a target can mean sales were pulled forward, so "improving" is only good news if it lasts into the next quarter.

2. Replenishment: are new orders refilling the tub?

A company can convert its backlog perfectly and still slow down if it stops winning orders. A book-to-bill steadily below 1 means the backlog is being used up. A ratio well above 1 is encouraging, but only if the company can actually deliver the extra work. Watch several quarters, because big contracts make bookings lumpy.

The two questions interact. A company with a book-to-bill of 0.8 and two years of backlog can keep its revenue steady for a while, so the slowdown in orders does not show up in sales until later. That delay is exactly why investors watch bookings: they are the first sign of next year's revenue. Over four quarters, a trailing book-to-bill that drifts from 1.2 toward 0.9 tells you more than any single quarter's number.

3. Quality: will the orders really be paid for?

Not every order is equally firm. Who are the customers, can they afford what they ordered, and how easily can they cancel? Boeing's 2020 removals were orders where it could "no longer assert that the customer is committed to perform."[^1] A backlog dominated by one customer is only as strong as that customer (see Revenue Concentration). Rising cancellations are the clearest sign that quality is weakening.

The weakest "backlog" of all is one that was never a set of real orders. The electric-truck start-up Lordstown Motors told investors it had more than 100,000 "pre-orders" for its Endurance pickup, mostly from commercial fleets. The pre-orders were nonbinding: no one had to buy anything. In 2024 the SEC found that most of them came from companies that did not run fleets or did not plan to buy the truck for their own use, and that Lordstown had also misstated its delivery timeline. The company had filed for bankruptcy in 2023. Without admitting or denying the findings, it agreed to an order that included $25.5 million of disgorgement (giving up money gained from the misconduct).[^14] The lesson is simple: before you measure how fast a backlog converts, check whether anyone is actually bound to pay for it.

Nuances

Industries convert at very different speeds, and long backlogs carry more risk because more can change before delivery. Lockheed Martin, the defense contractor, ended 2025 with a $193.6 billion backlog against $75.0 billion of sales, about 2.6 years of work, and expected about 37% of it to become revenue within 12 months. About $120 billion was "funded," meaning Congress had already appropriated the money.[^15] By the bathtub formula, its 2025 bookings were about $92.6 billion, a book-to-bill of roughly 1.2. Airbus delivered 793 jets in 2025 against a record backlog of 8,754, about 11 years of production at that pace.[^16]

Here is how the main industries differ:

• Defense. Contracts run for years and the customer is a government, so cancellations are rare but timing depends on budgets. The funded share is the firmest part. A long political fight over spending can delay new awards without touching the work already funded.
• Commercial aircraft. Backlogs stretch a decade, and revenue arrives mainly at delivery. Customers usually pay deposits, which makes them think twice before cancelling, but an airline in trouble can still defer or walk away. The limit on conversion is usually the factory and its suppliers, not demand.
• Homebuilding. Here backlog means homes sold but not yet handed over, usually within months. Buyers put down modest deposits, and their mortgage depends on interest rates, so cancellations can swing wildly (see the KB Home example below).
• Software and cloud. RPO and cRPO capture subscription contracts, often one to three years long. Conversion is smooth because revenue is recognized month by month as the service runs, but usage-based fees and short contracts may sit outside RPO entirely.
• Engineering and construction. Large projects are often fixed-price: the builder promises a total price up front and bears the risk if costs rise. Such a backlog can convert right on schedule and still lose money. For these firms, ask about the expected profit margin in the backlog, not only its size.

Cancellations can spike and then fade: KB Home. Homebuilders show cancellation risk in fast motion. In the quarter ended November 30, 2022, as mortgage rates jumped, KB Home signed 2,169 gross orders, but cancellations equaled 68% of them, compared with 13% a year earlier. Net orders fell to 692 from 3,529. Its backlog shrank 25% to 7,662 homes, worth about $3.69 billion. Management said it had "prioritized delivering our large backlog and protecting our high margins," and that it was "getting more aggressive with our pricing" to win new orders.[^17] What happened next is the conversion test. In fiscal 2023, KB Home delivered 13,236 homes, only 4% fewer than the year before, and revenue fell 7% to $6.41 billion. By the fourth quarter of 2023, the cancellation rate was back down to 28%. But the backlog kept shrinking, to 5,510 homes worth $2.67 billion, and the average selling price fell from $500,800 to $481,300.[^18] The homes already in backlog mostly converted. The weakness showed up instead in replenishment and price, which is where the next year's revenue comes from.

Price and acquisitions hide inside backlog. Long contracts often include price escalation, so backlog value can rise with no new units ordered, which links it to Price and Mix Growth rather than Volume Growth. Buying a company brings its order book along, so strip that out as you would for Organic Growth.

Backlog is not all future revenue. Renewals and usage often sit outside it; that growth shows up in Existing-Customer Expansion and Customer Base Growth. All of it feeds the headline number in Revenue Growth.

A contrasting case: Airbus in the same storm

The opening story showed a backlog that converted far behind schedule. Boeing's great rival lived through the same pandemic and the same airline crisis, with a very different result. In 2019 Airbus delivered 863 commercial aircraft and ended the year with a backlog of 7,482.[^19] In 2020 it delivered 566, a drop of about a third, in line with a reduced production plan it had announced in April. It booked 383 gross orders, lost 115 to cancellations, and ended with 268 net orders and a backlog of 7,184 aircraft, only about 4% smaller than a year before.[^20]

Compare Boeing: 157 deliveries, net orders of about minus 1,026, and a commercial backlog that lost a quarter of its value.[^1][^4] (The two companies measure backlog differently, Airbus in planes and Boeing in dollars, so the percentages are a rough comparison, not an exact one.) Both sold to the same kinds of airlines, facing the same empty airports. The difference was that Airbus could keep handing over planes, so its backlog kept converting, at a slower but predictable pace. When customers can see their planes arriving, they have fewer reasons to cancel. When deliveries stop, every month of waiting is another chance to walk away. Conversion and quality are linked: a backlog that converts on time tends to stay firm.

A worked example: Oracle

Oracle, the database and cloud company, has the most dramatic backlog of recent years. On September 9, 2025, it reported that its RPO had jumped 359% in a year to $455 billion, against analyst expectations of about $180 billion. The next day the stock rose almost 36%, its best day since 1992.[^21] The jump came from a few very large cloud contracts for AI computing; OpenAI alone agreed to spend more than $300 billion.[^22]

Oracle's fiscal year ends May 31. Here are its reported figures:[^23]

Quarter ended	RPO	Share due within 12 months	Implied 12-month amount	Revenue in quarter
May 31, 2025 (Q4 FY25)	$137.8 billion	33%	≈ $45 billion	$15.9 billion
Aug 31, 2025 (Q1 FY26)	$455.3 billion	10%	≈ $46 billion	$14.9 billion
Nov 30, 2025 (Q2 FY26)	$523.3 billion	10%	≈ $52 billion	$16.1 billion
Feb 28, 2026 (Q3 FY26)	$552.6 billion	12%	≈ $66 billion	$17.2 billion
May 31, 2026 (Q4 FY26)	$638 billion	12%	≈ $77 billion	$19.2 billion
Aug 31, 2026 (Q1 FY27)	$664 billion	13%	≈ $86 billion	$19.3 billion

The first two rows hold the key lesson. RPO more than tripled in one quarter, yet the amount due within 12 months barely moved, from about $45 billion to $46 billion. Almost all the new backlog was scheduled years ahead. The headline said a lot about Oracle in 2029 and little about 2026.

To judge whether Oracle's schedules can be trusted, step back and look at whole fiscal years, before and after the AI contracts:[^23]

Fiscal year (ends May 31)	Revenue	RPO at year end	Share due in 12 months (amount)	Capital spending	Free cash flow
2023	$50.0 billion	$67.9 billion	49% (≈ $33 billion)	$8.7 billion	$8.5 billion
2024	$53.0 billion	$97.9 billion	39% (≈ $38 billion)	$6.9 billion	$11.8 billion
2025	$57.4 billion	$137.8 billion	33% (≈ $45 billion)	$21.2 billion	−$0.4 billion
2026	$67.4 billion	$638 billion	12% (≈ $77 billion)	$55.7 billion	−$23.7 billion

Free cash flow is the cash a business brings in from operations minus what it spends on buildings and equipment (capital spending). Negative free cash flow means the company spent more than it generated.

Now run the backward test from earlier in this chapter. At the end of fiscal 2023, Oracle expected about $33 billion of its RPO to become revenue within 12 months. Fiscal 2024 revenue was $53.0 billion, so the old 12-month slice could account for at most about 63% of it, and the other $20 billion or so came from contracts signed during the year, renewals, usage, licenses, hardware, and services. A year later the slice was about $38 billion against fiscal 2025 revenue of $57.4 billion (67%); then about $45 billion against fiscal 2026 revenue of $67.4 billion (67%). In each year, revenue comfortably exceeded the near-term slice, which is what you would expect if the scheduled work was delivered and new business kept arriving. The test cannot prove that each contracted dollar arrived on time, because Oracle does not report which revenue came from which contract. But no year shows the warning sign of revenue falling short of the slice.

The latest row changes the picture. The slice due in fiscal 2027 is about $77 billion, and management's revenue guidance for that year is about $90 billion.[^24] That means about 85% of the expected revenue is already under contract, compared with roughly two-thirds in earlier years. On one hand, that makes the guidance better covered: Oracle needs only about $13 billion of revenue from outside the existing 12-month slice. On the other, it makes the year depend more on delivery than on selling. If a data center opens late, the revenue tied to it moves into the next year, much as Boeing's planes did.

The capital spending column shows why. Oracle's cloud contracts cannot be fulfilled with software alone; they need buildings, power, and chips. Capital spending rose from $6.9 billion in fiscal 2024 to $55.7 billion in fiscal 2026, and free cash flow went from $11.8 billion positive to $23.7 billion negative.[^23] Here the spending comes before the revenue, so a delay hurts twice: the cash has been spent, but the money promised by customers arrives later.

Pace. So far Oracle has delivered roughly on schedule. In September 2025 it forecast $18 billion of cloud infrastructure revenue for fiscal 2026;[^21] it reported $18.1 billion.[^25] In the first quarter of fiscal 2027, revenue grew 30% and cloud infrastructure revenue grew 121%.[^24] The 12-month slice has nearly doubled to about $86 billion, and about half of the $664 billion is due within 36 months.[^23] Against the company's targets, conversion looks stable to improving.

Replenishment. RPO grew by about $318 billion in the quarter ended August 2025, then by $68 billion, $29 billion, $85 billion, and $26 billion. By the bathtub formula, latest-quarter bookings were about $26 billion + $19.3 billion ≈ $45 billion, a book-to-bill near 2.3. That is a rough estimate, since some revenue never passes through RPO, but the tub is still filling faster than it drains.

Quality. Here lie the doubts. One customer, OpenAI, accounts for an estimated $300 billion of the backlog, and in April 2026 Oracle shares fell after a report that OpenAI had missed its own growth targets.[^22][^26] Delivering the backlog also needs huge spending first: free cash flow was negative $23.7 billion in fiscal 2026, and debt reached about $125 billion.[^25][^27] Oracle says the prepaid and customer-supplied-hardware portions of its large AI contracts total $75 billion, which reduces how much it must borrow.[^25] By late December 2025, the stock had fallen 43% from its September peak.[^22]

The thesis. An investor could state: "Oracle's AI backlog will convert to revenue on the schedule management has set." The resolution rule: "Oracle reports fiscal 2027 revenue (the year ending May 31, 2027) of at least $90 billion, and at least $100 billion of RPO due within the following 12 months." The first part is Oracle's own guidance, slightly above the analyst consensus of about $89.8 billion.[^24][^27] The second follows from the schedule: about $246 billion (37%) of today's RPO is due in months 13 to 36, so the near-term slice should keep rising if deliveries stay on time.

Each quarterly report updates the thesis. A rising near-term slice makes it more likely; a falling 12-month share, a restructured contract, a delayed data center, or trouble raising money make it less likely, even if headline RPO keeps growing.

In practice, an investor would keep a simple scorecard each quarter. Did revenue match the guidance given three months earlier? Did the 12-month amount rise by roughly as much as the revenue just recognized, showing that work is moving forward on schedule from the later years into the near term? Did management repeat or cut its full-year target? Did any large customer change its contract? If the 12-month amount stalls while total RPO keeps rising, the backlog is being pushed further out, and the thesis is weakening even though the headline looks stronger. The resolution comes in mid-2027, when Oracle reports its full fiscal year, which fits the typical horizon for this kind of thesis.

Common traps

• Mistaking size for speed. A huge backlog due in five years says little about next year. Find the near-term slice.
• Assuming orders are firm. Options, letters of intent, and unfunded awards can disappear. Read the company's definition.
• Ignoring the customer. A backlog is only as good as the customers' ability to pay.
• Overreacting to one lumpy quarter. One giant contract can make book-to-bill look spectacular, and its absence a collapse.
• Missing the silence. When a company stops giving a precise backlog figure, ask why.
• Counting nonbinding interest as orders. Pre-orders, reservations, and memorandums of understanding are marketing until someone signs a contract, as Lordstown's 100,000 "pre-orders" showed.
• Cheering every speed-up. Deliveries rushed out just before quarter-end can lift revenue today and leave a hole tomorrow. Check whether the next quarter holds up.
• Forgetting the margin. A fixed-price backlog can convert on schedule and still lose money if costs rise. Size and timing are not the same as profit.
• Comparing backlog with RPO. A company's own backlog figure and its audited RPO can differ widely. Compare like with like, across years and across companies.
• Ignoring the cost of delivering. When a company must spend heavily before it can deliver, as Oracle must on data centers, a delay also strains its cash and borrowing.

Checklist

When you study a company's backlog, ask:

1. What exactly does the company count as backlog or RPO?
2. How much is due within 12 months, and is that share rising or falling?
3. Did last year's backlog convert on the schedule the company described?
4. Is book-to-bill above or below 1, and what is the trend?
5. Are cancellations and removals growing?
6. How concentrated is the backlog, and how strong are the main customers?
7. Does the company have the capacity, suppliers, and money to deliver on time?
8. Which parts of the backlog are firm, funded, or binding, and which are options, letters of intent, or pre-orders?
9. Did revenue in the past year at least equal the 12-month slice reported a year earlier (the backward test)?
10. Was a target beaten by a very small margin while the ending backlog fell unusually fast, a possible sign of pulled-forward deliveries?
11. Is the backlog priced to earn a profit, especially on fixed-price contracts?

Key takeaways

• Backlog is signed work not yet delivered; RPO is its audited accounting cousin.
• Conversion is about timing and reliability, not size. Check the near-term slice.
• Judge it on pace (on schedule?), replenishment (book-to-bill), and quality (cancellations, concentration, customer strength).
• Backlog helps predict earnings, but investors tend to overvalue big backlogs.
• A backlog can be real and still late, as Boeing showed; a rival that kept delivering, Airbus, kept its backlog nearly intact through the same crisis.
• Cancellations can spike quickly when customers can walk away cheaply, as KB Home's 68% cancellation rate in late 2022 showed, and nonbinding "orders" may never have been orders at all.
• Unusually fast conversion near a target can mean sales were pulled forward, so judge conversion over several periods.

Data needed to support this thesis

To build and keep updating a backlog conversion thesis, you need the following data.

Required

Data	What it tells you	Where to find it
Backlog disclosures	The size of signed but undelivered work, and how much is due within the next 12 months	Remaining performance obligations in the revenue note of annual and quarterly reports (Forms 10-K and 10-Q); backlog tables in results releases
Revenue history	How fast past backlog turned into sales, to compare with the schedule	Income statements in the same filings; at least three comparable periods

Helpful

Data	What it tells you	Where to find it
Contract data	Contract size, length, and timing of delivery	The revenue note in filings; contract announcements and press releases
Company guidance	What management expects to deliver and when	Results releases and earnings calls
Customer concentration	How much of the backlog depends on one or a few customers	The concentration note in filings; news reports and earnings calls

The comparisons must be like-for-like. Companies define backlog differently. Some include options or unfunded orders and some do not. Some change the definition or remove cancelled orders. Some count backlog in units, as Airbus does with planes and KB Home with homes, and some in dollars, as Boeing and Oracle do. Use one definition, and check what the company counts before comparing periods. When comparing a company's cancellation rate over time, make sure it divides by gross orders the same way each period.

The thesis is usually updated by three kinds of events: earnings releases (new backlog and revenue figures), contract awards (large new orders), and guidance updates (a new delivery schedule). A typical horizon for checking the thesis is about a year and a half.

In the Oracle example, the required data was its reported remaining performance obligations and the share due within 12 months, plus quarterly and annual revenue, capital spending, and free cash flow, from its annual and quarterly reports for fiscal 2023 through the first quarter of fiscal 2027.[^23] The helpful data was the company's own forecasts and results releases, and a news-based estimate of how much of the backlog one customer represents.[^21][^22][^25][^24] The example had no contract-level data, since Oracle does not name its customer or publish contract terms.[^22]

---

[^1]: Boeing, 2020 Form 10-K: deliveries, revenue, 2020 cancellations ($34,618 million) and ASC 606 removals ($54,450 million) against a starting commercial backlog of $376,593 million ($281,588 million at year end); commercial airplanes revenue of $57,499 million (2018) and $16,162 million (2020); 2019 customer-consideration charges of $8,259 million; operating cash flow of $15,322 million (2018) and −$18,410 million (2020); debt of $27.3 billion and $63.6 billion at the ends of 2019 and 2020; contractual and unobligated backlog definitions; revenue recognized mainly at delivery. SEC. [^2]: Boeing, fourth-quarter 2019 results (January 29, 2020). Boeing. [^3]: Federal Aviation Administration, "FAA Updates on Boeing 737 MAX" (grounding of March 2019 rescinded November 18, 2020). FAA. [^4]: Gross orders: Al Jazeera (January 12, 2021). Net orders after cancellations and ASC 606 adjustments: AirInsight (January 12, 2021). [^5]: Boeing, fourth-quarter 2025 results (January 27, 2026). Boeing. [^6]: BYU School of Accountancy, RevenueHub, "Disclosures in ASC 606." RevenueHub. Oracle's 10-Qs note its use of the variable-consideration exemption (see footnote 23). [^7]: SEC, "Modernization of Regulation S-K Items 101, 103, and 105," Release No. 33-10825 (August 26, 2020). SEC. [^8]: Shivaram Rajgopal, Terry Shevlin, and Mohan Venkatachalam, "Does the Stock Market Fully Appreciate the Implications of Leading Indicators for Future Earnings? Evidence from Order Backlog," Review of Accounting Studies 8 (2003). Springer; sample size and 5.8% hedge return from the working paper. [^9]: Rajiv D. Banker, Russell Barber, Dana Hollie, and Han-Up Park, "Order Backlog and Its Association with Fundamental Analysis Metrics and Future Earnings," Review of Quantitative Finance and Accounting 62 (2024): 1733–1753. RePEc. [^10]: Russell Barber and Dana Hollie, "Does Order Backlog Matter for Financial Reporting Quality? Evidence from Revenue Restatements," Advances in Accounting 53 (2021). RePEc. [^11]: Thomas A. Gilliam, Frank Heflin, and Jeffrey S. Paterson, "Revenue Management through Order Backlog Manipulations," Review of Quantitative Finance and Accounting 65 (2025): 1067–1096. RePEc. [^12]: Uday Chandra, Andrew Procassini, and Gregory Waymire, "The Use of Trade Association Disclosures by Investors and Analysts: Evidence from the Semiconductor Industry," Contemporary Accounting Research 16 (1999). Wiley. [^13]: Kimball Chapman, Zachary Kaplan, and Chase Potter, "Disclosure Specificity: Evidence from Book-to-Bill Ratios," Journal of Business Finance & Accounting 51 (2024). RePEc. [^14]: SEC, "SEC Charges Lordstown Motors with Misleading Investors about Company's Flagship Electric Vehicle," Press Release 2024-29 (February 29, 2024). SEC. [^15]: Lockheed Martin, 2025 Form 10-K: backlog of $193.6 billion ($176.0 billion a year earlier), funded backlog of $120.2 billion, sales of $75,048 million. SEC. Bookings and book-to-bill are our calculation. [^16]: Airbus, 2025 commercial aircraft deliveries (January 2026). Airbus; Runway Girl Network. [^17]: KB Home, fourth-quarter and full-year 2022 results (January 11, 2023), Exhibit 99.1 to Form 8-K: quarter ended November 30, 2022. SEC. [^18]: KB Home, fourth-quarter and full-year 2023 results (January 10, 2024), Exhibit 99.1 to Form 8-K: fiscal year ended November 30, 2023. SEC. The 7% revenue decline is our calculation. [^19]: Airbus, "Airbus delivers strong 2019 commercial aircraft performance" (January 2020). Airbus. [^20]: Airbus, "Airbus 2020 deliveries demonstrate resilience" (January 8, 2021). Airbus. Percentage changes are our calculations. [^21]: CNBC, "Oracle stock gains 36% to post best day since 1992" (September 10, 2025). CNBC. [^22]: CNBC, "Oracle stock on pace for worst quarter since 2001" (December 26, 2025). CNBC. Oracle does not name the customer; OpenAI's share is estimated from the reported contract value. [^23]: Oracle, Forms 10-K for fiscal 2023, 2024, 2025 and 2026; Forms 10-Q for quarters ended August 2025, November 2025, February 2026, and August 2026. RPO, 12-month shares, revenue, capital expenditures, and free cash flow as reported; fiscal 2023 capital spending and free cash flow from the fiscal 2024 10-K. Implied 12-month amounts, fourth-quarter revenue, and backward-test ratios are our calculations. [^24]: Oracle, Q1 fiscal 2027 results (September 10, 2026). SEC. [^25]: Oracle, Q4 and fiscal 2026 results (June 10, 2026). SEC. [^26]: CNBC, "OpenAI reportedly missed revenue targets" (April 28, 2026). CNBC. [^27]: CNBC, Oracle Q1 fiscal 2027 earnings (September 10, 2026). CNBC.