A software contract is signed, the sales commission is earned and the customer begins deployment. Those events can start three different clocks: payment to the salesperson, revenue from the customer and the expense shown in the software vendor’s accounts. Reading only this quarter’s sales-and-marketing line can miss part of that sequence.
The useful question is how much selling effort must be paid for before a customer relationship produces its return. Capitalised contract costs help explain the timing. They do not, on their own, reveal whether the product is economical to sell.
Fact: an asset can represent an earlier selling cost
Salesforce (NYSE: CRM) says it capitalises certain costs of obtaining customer contracts, including commissions, and spreads them into expense on a straight-line basis. Its filing explicitly distinguishes that expense timing from the associated cash payment. At 31 July 2026, its balance sheet showed $2.074 billion of current and $2.870 billion of non-current net costs capitalised to obtain revenue contracts: $4.944 billion combined. These are asset balances, not the amount spent that quarter. Source: Salesforce’s July 2026 Form 10-Q.
Adobe (NASDAQ: ADBE) describes capitalising incremental contract-acquisition costs when their expected benefit exceeds one year, primarily sales commissions. The balance was $815 million at 28 August 2026, compared with $721 million at 28 November 2025. The increase is $94 million. It is not a measure of new cash commissions: additions, expense recognition and other movements have to be distinguished. Source: Adobe’s August 2026 Form 10-Q, contract balances.
ServiceNow (NYSE: NOW) provides another dated example. At 30 June 2026, current deferred commissions of $594 million and a non-current portion of $1.136 billion totalled $1.730 billion. Its cash-flow reconciliation for the six months then ended separately showed $340 million of deferred-commission amortisation and a negative $376 million deferred-commissions movement. The reporting periods and labels matter. Source: ServiceNow’s June 2026 Form 10-Q.
Fact · ServiceNow · H1 2026
Two entries, different jobs
Amortisation
Cash-flow add-back
Deferred commissions
Operating-asset movement
Adding back amortisation when reconciling profit to operating cash does not make selling free. That expense is non-cash in the period of recognition; the separate asset movement helps account for the different timing. Neither line is a complete customer-acquisition-cost measure.
Interpretation: follow a customer, then follow the accounting
Imagine a hypothetical vendor paying a £1,200 qualifying commission at the start of a customer relationship. Assume, solely for this example, a three-year benefit period and equal annual amortisation. With no impairment or other changes, £400 enters expense in each year, even though the £1,200 cash payment happens at the start. The payment and the expense describe the same cost on different dates.
A longer assumed benefit period would reduce the annual charge in that example. It would not return any cash. A shorter period would do the reverse. This is why a reader needs the policy and its supporting business assumptions before treating a low current expense as proof of improved sales productivity. The example is arithmetic, not any of these companies’ commission policy.
The customer’s workflow supplies the commercial test. Does deployment become routine use? Does the customer renew, expand or leave? A sales payment made today can be sensible if the relationship produces enough future contribution after hosting, support and further selling costs. The asset alone cannot establish that outcome.
Useful software still has a cost. Winning the customer is one cost; keeping the software useful is another. A vendor can improve one while the other deteriorates. Looking at contract costs alongside retention evidence and delivery economics gives a fuller picture than treating an accounting balance as a customer-success score.
Who could benefit?
Interpretation: Salesforce and Adobe could gain economically when existing relationships generate further paid business without an equivalent repeat of the original acquisition effort. The mechanism is lower selling cash required per pound or dollar of future customer contribution, over renewals and subsequent years. That would require durable use, acceptable pricing and manageable renewal incentives; capitalisation itself creates none of those advantages.
ServiceNow has the same conditional opportunity if deployment becomes embedded in recurring workflows and customers remain commercially attractive after ongoing support and selling effort. Its disclosed commission asset establishes financial exposure to the timing issue. It does not disclose a return for each acquired customer. None of these balances establishes which company will benefit most, or how material a change would be to group earnings.
Who’s at risk?
Interpretation: all three vendors face the opposite business mechanism if acquiring or retaining customers requires more cash incentives while future contribution weakens. That pressure can emerge across several contract cycles. Current reported selling expense may reflect earlier acquisition activity as well as today’s work, making a sudden commercial change harder to diagnose from a single quarter.
Readers also risk confusing an asset with available liquidity. A deferred commission is not a pot of cash that management can spend again. Nor does its presence establish improper accounting. The relevant questions are whether the expected benefit remains supportable and whether future customer economics justify the resources committed.
Speculation: a cheaper route to the next contract
Self-service buying or better sales tools could reduce the effort needed to secure some contracts. More complicated procurement and deployment could increase it. These are possible mechanisms, not measured outcomes for the named companies. A claim that automation has improved acquisition economics needs evidence connecting changed selling work to cash costs and durable customer contribution.
This article uses three historical reporting dates to explain a timing problem, not a league table. The companies differ in scale, products, contract mix and accounting estimates; the balances should not be compared as efficiency ratios. In particular, none isolates the cost of selling AI products. Checked on 23 October 2026, the examples do not forecast subsequent results.
The evidence that would strengthen the business case is specific: consistent cost definitions, a clear bridge between additions and amortisation, and customer relationships that keep generating contribution after the original selling bill. A larger capitalised balance by itself settles none of those questions.
Photo by Kaboompics on Pexels. Hands signing a document.