Skip to content

Leadership and readiness

Dataset delivery backlog management: which signed orders will convert

By SourceX Editorial · Reviewed by Noah Loul ·

10 sections

Short answer

Dataset delivery backlog management treats signed, scoped, undelivered work as a ledger with four flows: enforceable signings in, accepted deliveries out, voids out, and scope changes. Age each order against its promised date and code every wait to supplier rights, preparation capacity or buyer review. Public filings show why the definition matters: Booz Allen's headline backlog was 3.6 times its remaining performance obligations, and Palantir's deal value 3.2 times.

Record timeline

Dataset delivery backlog

This page covers dataset delivery backlog management: which signed orders will convert.Illustrative

Key takeaways

  • Count only signed, scoped orders that neither side can cancel without compensation; letters of intent, options and unsigned renewals belong in pipeline.
  • In two public 10-Ks the headline figure ran 3.2 to 3.6 times the enforceable layer, so ask which number a startup's backlog slide shows.
  • Age every order on two clocks, months since signing and months past its promised date, and code each waiting day to supplier rights, preparation capacity or buyer review.
  • Log voids by reason: in this article's illustrative model, a backlog that empties faster than plan can be losing orders rather than delivering them.
  • Fund preparation capacity only when the oldest orders wait on the company's own specialists; if they wait on buyer review, more staff adds cost without acceptances.

What belongs in a data startup's delivery backlog?

A delivery backlog is the signed work a data company owes but has not yet had accepted: orders under an executed contract that name the records, the format and the acceptance test, where neither side can walk away without paying. Anything short of that is pipeline. A letter of intent, an unexercised option, a renewal the buyer has not signed or an 'up to' volume ceiling may matter commercially, yet none of them obliges a buyer to take a delivery.

The accounting standard draws a similar line. Topic 606 requires public companies to disclose the transaction price allocated to performance obligations still unsatisfied at period end and to explain when they expect to recognize it, with an optional exemption for contracts whose original expected duration is one year or less (FASB, ASU 2016-20). FASB's implementation questions and answers add that when each party can end a contract at any time without compensating the other, the contract's duration stops at the goods or services already transferred, so the cancellable remainder is not an obligation at all.

Backlog has no such discipline. In ASU 2016-20 the FASB describes it as an informal measure that companies disclose voluntarily, one that can fold in anticipated contracts, wholly unperformed contracts, renewal options and constrained variable consideration, and that may not compare across companies or even across years for one company. A data startup's backlog slide inherits that looseness unless the investor writes the definition first.

How far can a headline backlog sit from the enforceable core?

In the two public filings examined here, the headline figure was more than three times the enforceable layer. That gap is the first thing a backlog model has to explain, and companies that disclose both numbers make it measurable.

Booz Allen Hamilton's 10-K for the fiscal year ended March 31, 2026 splits total backlog of $38.19 billion into funded orders, unfunded orders and priced options, and reports remaining performance obligations of $10.7 billion, about 28 percent of the total. Federal contract clauses explain why the layers behave differently. In incrementally funded contracts, FAR 52.232-22, Limitation of Funds, says the contractor is not obligated to keep performing beyond the amount actually allotted, and FAR 52.249-2 lets the government terminate a fixed-price contract, in whole or in part, when the contracting officer decides that serves its interest.

Palantir's fiscal 2024 10-K shows the same pattern across commercial and government customers. Remaining performance obligations were $1.7 billion at December 31, 2024, against total remaining deal value of $5.4 billion. The company's filings define deal value as presuming that every contract option is exercised and no contract is terminated, while noting that most of its contracts carry termination provisions, including for convenience; its RPO note leaves cancelable contracted revenue out.

Neither company sells datasets, which is the point: both print a careful accounting number beside a broader management number, and a startup's backlog slide usually shows only the broader one. The table computes each layer's share from the filings' own figures, with the rounding rule given in the methods note.

How far can a headline backlog sit from the enforceable core?
Filing lineAmount as reportedShare of the headline figureEnforceable today?
Booz Allen funded backlog, March 31, 2026$4,319 million11.3% of total backlogYes: funding appropriated or authorized
Booz Allen unfunded backlog$10,183 million26.7%Ordered, but funding not yet authorized
Booz Allen priced options$23,685 million62.0%No: exercised only at the client's option
Booz Allen remaining performance obligations$10.7 billionAbout 28% (27.9-28.2% given filer rounding)The accounting measure of unsatisfied obligations
Palantir total remaining deal value, December 31, 2024$5.4 billionHeadline figureAssumes all options exercised and no terminations
Palantir remaining performance obligations$1.7 billionAbout 31% (30.3-32.7% given filer rounding)Noncancelable contracted revenue only

Two clocks for every signed order: age since signing and slip against the promise

Age each open order on two clocks, months since signing and months past its promised acceptance date, because they answer different questions. Age since signing shows how long the company has carried an obligation it has not been able to turn into an accepted delivery. Slip against the promise shows whether the date sold to the buyer was ever achievable.

Reconcile every promised date with a feasible one. The feasible date is the month the preparation queue reaches the order at current specialist capacity, plus the buyer's documented review window. An order whose feasible date already falls after its promise is late on the day it is signed, and reporting it as on time until the deadline passes hides a sales problem inside an operations metric.

Then give every waiting day one cause code. Supplier-side waits cover rights confirmation and extraction from the source system; preparation-side waits cover the specialist queue and rework after a failed check; buyer-side waits cover a review that has not started, an acceptance test that changed after signing and buyer-side access problems. Buyer-side waits are the ones a hiring plan cannot shorten, which is why they must not be blended into one 'in progress' status.

Use aging buckets the company cannot quietly redraw. The buckets below are editable defaults; what matters is that they are fixed in writing before the first report:

  • On or before the promised acceptance date.
  • 1 to 60 days past the promised date.
  • 61 to 180 days past the promised date.
  • More than 180 days past the promised date.
  • No promised date recorded, reported as a data defect rather than as on time.

What converts a signed order into recognized revenue, and what voids it?

An order converts when the buyer accepts a delivery that meets the contract's specification, and it voids when the obligation ends without one. A backlog model needs both exits, each with a reason, because a shrinking backlog can mean either.

Steps

What converts a signed order into recognized revenue, and what voids it

Acceptance is the usual conversion event for data deliveries because quality is judged against the buyer's own test. Topic 606 lists a customer's acceptance of an asset among the indicators that control has passed and directs companies to its customer acceptance guidance to judge what a contractual acceptance clause does to timing (FASB, ASU 2014-09). An upload or a shipped drive is therefore a delivery, not a conversion; keep it in backlog as delivered and awaiting acceptance. How accepted work then flows into reported revenue is a separate diligence question that this article leaves to a revenue recognition review.

Voids come from a short list, and each should carry a code, the order's age and the document that ended it:

  • Termination under a contract right: the buyer ends the order under a cancellation or convenience clause, with FAR 52.249-2 as the federal model, or under a right either side could use without compensation.
  • Rights failure: the supplying company declines to sign, narrows the scope after review, or withdraws records before delivery.
  • Failed acceptance after cure: the delivery misses the test and the contract's correction path runs out.
  • Specification lapse: the buyer's task changes, the scoped records no longer fit, and the order is cancelled rather than amended.
  • Window expiry: a delivery deadline passes with no extension signed.
  • Not a void: an amendment that adds or removes volume is a scope change, recorded as its own flow and never netted against new signings.

The backlog ledger an investor should ask to see

The asset is a roll-forward ledger: opening backlog plus enforceable signings, minus accepted deliveries and voids, plus or minus scope changes, equals closing backlog. Each line below carries a definition, the field that proves it and the flattering version a diligence team should refuse.

Count orders as well as contract value. A value-weighted backlog hides the case where one large order is aging while many small ones convert, and an order count can be checked against the contract file one row at a time. Ask for the ledger by quarter, with each quarter's closing balance equal to the next quarter's opening balance.

The figure draws the same ledger as a stock with flows: what enters, the aging buckets inside, and the two exits:

The backlog ledger an investor should ask to see
Ledger lineDefinitionField that proves itFlattering version to refuse
Opening backlogSigned, scoped orders not yet accepted at the start of the periodContract ID, signing date, scoped record set, acceptance test referenceAn opening figure that differs from last quarter's close
Enforceable signingsOrders executed this period that neither side can cancel without compensationCountersigned contract or order form with its dateLetters of intent, options and unsigned renewals
Accepted and convertedDeliveries the buyer accepted against the written testAcceptance notice or contractual deemed-acceptance event, datedUploaded or shipped work counted as done
Voided, by reasonOrders that ended without an accepted deliveryTermination notice, rights withdrawal, failed-cure record or lapse dateVoids netted against new signings
Scope changesAmendments that add or remove volume on an existing orderSigned amendment linked to the original orderAmendments booked as fresh signings
Closing backlogOpening plus signings minus conversions and voids, plus or minus scope changesA roll-forward that ties to the order listA closing number with no order list behind it
Aging against promiseEach open order's time past its promised acceptance datePromised date as written in the contract or orderPromised dates reset whenever they slip
Waits by causeWaiting days coded to supplier rights, preparation capacity or buyer reviewStatus history with one cause code per waiting dayA single undifferentiated 'in progress' status
A stock-and-flow diagram of a delivery backlog: enforceable signings flow into four aging buckets, which drain through accepted deliveries or voids with reasons, while options and letters of intent stay outside the stock.
The backlog ledger as a stock with two exits. Illustrative structure; no company's figures.

Six months of coverage that is not six months: an illustrative model

Coverage months, backlog divided by monthly delivery capacity, overstates how quickly a backlog converts and says nothing about voids. A small deterministic model makes the gap visible; every input is an editable assumption, not a measurement of any company.

The model starts with 24 signed orders and capacity to prepare 4 a month, so naive coverage is six months. Prepared orders spend a set time in buyer review, 85 percent pass first review and the rest return to the front of the queue for rework. An order still waiting for preparation after its promised acceptance month voids at 15 percent a month, and orders are worked first in, first out.

Three readings follow. Even when promises match capacity (scenario A), only 71 percent of orders are accepted by month six, because review time and rework sit outside the coverage arithmetic. When buyer review takes three months instead of one (B), acceptance by month six falls to 43 percent and the backlog runs to month 14, yet extra specialists would change neither number. When sales set dates as if capacity were double (C), month six looks identical to A, but 11 percent of orders void by month 24 and the backlog empties sooner, in month 9, because voided orders leave the queue.

Scenario C is the trap for a board that reads one number. A backlog that shrinks faster than plan can be losing orders, and only a ledger with void reasons tells the two apart. The results:

Six months of coverage that is not six months: an illustrative model
Scenario (Illustrative)What changes from AAccepted by month 6Voided by month 24Under half an order open from
A: promises match capacityNothing: review takes 1 month, dates assume 4 orders a month71%0%Month 10
B: buyer review runs longReview takes 3 months43%0%Month 14
C: promises assume double capacityDates assume 8 orders a month71%11%Month 9
D: B and C togetherReview takes 3 months and dates assume 8 a month43%9%Month 13

Fund specialists, buyer-side review or neither? A decision tree and milestone table

Release money for delivery capacity only after the ledger shows that the oldest orders are waiting on the company's own specialists. If they wait on supplier rights or buyer review, spend on that dependency, or on nothing until it moves.

The strongest objection to a capacity raise is scenario B: hiring and automation scale cost while conversions follow the buyer's review clock. Ask for the share of waiting days coded to buyer review in the oldest bucket. If it dominates, the remedy sits in the contract and the buyer relationship, such as an acceptance test agreed before signing, a named reviewer on the buyer's side or agreed acceptance mechanics, rather than in headcount.

Run the tree in order and stop at the first 'no'; the milestone table then sets what evidence releases each kind of spending:

  • 1. Does every order in the backlog trace to an executed, scoped contract that neither side can cancel without compensation? If no, restate the backlog before any capital decision.
  • 2. Is each order's feasible date, from its queue position at current capacity plus the buyer's review window, on or before its promised date? If no, fix how dates are promised before adding capacity.
  • 3. In the oldest aging bucket, are most waiting days coded to the company's own preparation work? If no, fund acceptance-test work with buyers when review dominates, or rights operations when supplier waits dominate.
  • 4. Have void rates, by reason, held flat or fallen over the last two quarters? If no, find and fix the void reason before funding throughput.
  • 5. Does the roll-forward tie order by order across quarters? If no, stop: the ledger is not yet evidence. If yes, release capacity spending in steps sized to the oldest bucket and rerun the ledger before each step.
Fund specialists, buyer-side review or neither? A decision tree and milestone table
MilestoneEvidence that shows itSpending it can releaseStop or reassess if
Backlog restated to enforceable ordersOrder list tied to executed contracts, with excluded items listed separatelyNone yet; it is the prerequisite for every later stepAny order lacks an executed, scoped contract
Feasible dates computed for every orderQueue position, capacity assumption and buyer review window per orderA small fix to how sales sets datesFeasible dates trail promises for most new signings
Wait causes coded for two quartersStatus history with a cause code on each waiting daySpending aimed at the dominant cause onlyCodes are missing or collapse into one status
Void log by reasonTermination, rights, failed-cure and lapse records linked to ordersThroughput spending once voids are flat or fallingVoids rise in the oldest bucket
Roll-forward ties across quartersEach closing balance equals the next opening, order by orderCapacity hires or tooling in stepsRestatements arrive without a written reason

What could make a backlog model mislead?

Five things can make even a carefully kept backlog ledger tell the wrong story, and each has a check an investor can run:

  • Definitions drift. The FASB notes that backlog lacks a common definition and can change over time for one company (ASU 2016-20); fix the definition in writing and require a restatement note whenever it changes.
  • One buyer dominates. Orders from a single buyer convert or void together, so report the largest buyer's share of open orders and of open value alongside the total.
  • Termination rights sit in the fine print. Palantir's filings say most of its contracts carry termination provisions, including for convenience; read the cancellation clause of every large order before counting it.
  • Acceptance can be partial. A delivery accepted with credits or a reduced scope converts less than its signed value; record the difference as a scope change with a reason.
  • The illustrative model is simple by design. It holds capacity, review time and void rates fixed, treats orders as equal in size and ignores seasonality; a real forecast needs the company's own status history.

Backlog questions for any data company, SourceX included

Put the same questions to every data licensing company and expect the answers to come from the order list, not a slide.

SourceX's published process sets where its backlog would begin. The supplying company approves scope and terms and signs before any records reach SourceX, and each deal's signed agreement fixes exclusivity and term. Because SourceX itself handles sourcing, rights review, contracting, preparation, delivery and payment, all three wait causes in this article's ledger, supplier rights, preparation and buyer review, sit inside one operating company and should be visible in one status history.

SourceX has not published a backlog, delivery timing or order metrics. Investors who want company and commercial figures can ask for them through the materials request at sourcex.si/investors and bring these questions with them:

  • Which open orders could a buyer cancel without paying, and are any of them counted in backlog?
  • What promised acceptance date did each order carry when it was signed, and has any date been reset?
  • What share of waiting days in the oldest bucket is coded to buyer review?
  • What voided in the last four quarters, by reason, and how old was each order when it voided?
  • How much of the backlog, by count and by value, belongs to the largest buyer?
  • Does each quarter's closing backlog equal the next quarter's opening, order by order?

Methods note: filings, model and what was not measured

The backlog ledger, wait codes, decision tree and milestone table were built for this article on 2026-10-27. Public evidence: FASB ASU 2014-09, ASU 2016-20 and FASB's revenue recognition implementation questions and answers; Booz Allen's fiscal 2026 10-K; Palantir's fiscal 2022 and 2024 10-Ks; and FAR 52.232-22 and 52.249-2 as published in GSA's FAR repository (commit da52ccb, March 30, 2026). Filing figures were checked on sec.gov on 2026-10-27.

Shares in the filings table are exact fractions of the reported figures, rounded once to one decimal place, half up. Where a filer had already rounded a figure to the nearest $0.1 billion, the table gives the range that rounding allows. The illustrative model has no randomness: 24 orders, preparation capacity of 4 a month, first-in first-out with rework served first, 85 percent first-review acceptance, a 15 percent monthly void rate for orders still waiting after their promised month, review of one or three months, and dates promised at 4 or 8 orders a month, assuming one month of review in every scenario; prepared orders and orders in review do not void. Shares are rounded once to whole percent. Both scripts are kept with the article's working files for review.

Not measured: the backlog, aging, void rates or conversion of SourceX or of any private data company; how accepted work becomes reported revenue; and whether the two public companies' ratios carry over to data licensing, where contracts are typically smaller and buyers fewer.

Frequently asked questions

Is a backlog the same as remaining performance obligations?

No. Remaining performance obligations are an accounting disclosure of the price allocated to obligations not yet satisfied, while backlog is a management measure without a common definition. In Booz Allen's fiscal 2026 10-K, remaining performance obligations were about 28 percent of total backlog.

Should letters of intent count in a data startup's backlog?

No. A letter of intent does not oblige the buyer to accept a delivery, so it belongs in pipeline with its own conversion history. It enters backlog only when an executed order names the records, the format and the acceptance test.

How should a dataset delivery backlog be aged?

Twice: by months since signing and by time past the promised acceptance date. Code every waiting day to supplier rights, preparation capacity or buyer review, so the aging report shows whose clock is running.

What usually voids a signed dataset order?

A cancellation right being used, a supplier that declines to sign or withdraws records, a delivery that fails acceptance after the correction path, a buyer whose task changes, or a deadline that lapses. Log each with its reason and the order's age.

Does a growing backlog mean a data company should hire?

Only if the oldest orders are waiting on the company's own specialists. If buyer review or supplier rights hold them, more staff raises cost without raising acceptances, as the long-review scenario in this article's illustrative model shows.

Sources

  • Topic 606 requires disclosure of the transaction price allocated to unsatisfied performance obligations, with an exemption for contracts of one year or less; FASB describes backlog as an informal, voluntary measure that can include anticipated contracts and renewal options and lacks a common definition. Source
  • If each party can terminate a contract at any time without compensating the other, its duration does not extend beyond the goods or services already transferred. Source
  • Topic 606 treats a customer's acceptance of an asset as an indicator that control has transferred and refers contractual acceptance clauses to its customer acceptance guidance. Source
  • Booz Allen's fiscal 2026 10-K reports total backlog of $38,187 million at March 31, 2026 (funded $4,319 million, unfunded $10,183 million, priced options $23,685 million) and remaining performance obligations of $10.7 billion. Source
  • Palantir's fiscal 2024 10-K reports remaining performance obligations of $1.7 billion and total remaining deal value of $5.4 billion at December 31, 2024, and excludes cancelable contracted revenue from remaining performance obligations. Source
  • Palantir defines total remaining deal value as presuming exercise of all contract options and no terminations, and says most of its contracts are subject to termination provisions, including for convenience. Source
  • FAR 52.232-22, Limitation of Funds, provides that the contractor is not obligated to continue performance or incur costs beyond the amount allotted to the contract. Source
  • FAR 52.249-2 allows the government to terminate a fixed-price contract in whole or in part when the contracting officer determines termination is in its interest. Source

Request SourceX investor materials

SourceX shares company and commercial metrics with investors directly. Any offering is made only to verified accredited investors.

See if your company qualifies