Software & Platforms Revenue Models: Subscription, Consumption, Transaction, Advertising, Services, and Hardware

Hello. Last lesson classified businesses by what they provide and the exchanges they govern. Now we will ask a different question: what event causes each business to earn revenue? A cloud provider and an AI application may both charge for consumption; a marketplace may earn subscriptions as well as transaction fees. The company archetype tells you where to look, but it does not determine the revenue mix.

This lesson gives you a method for decomposing a company’s revenue into subscription, consumption, transaction, advertising, services, and hardware models. We will use disclosures from two hybrid companies to distinguish what the numbers establish from what remains an estimate. Allow about 45 minutes, including the video and readings.

1. Classify the charge, not the product

A useful revenue decomposition starts with the unit that triggers payment. Consider a cloud-hosted application. Its technology might be identical under two contracts, yet one customer pays a fixed annual fee for access while another pays according to computing resources used. Those contracts have different revenue drivers even though both involve cloud software.

These are analytical buckets, not a claim that every company reports them separately. A recurring managed-services contract is recurring revenue, but if the customer chiefly pays for operating work, services may be more informative than subscription. A cloud contract can combine a committed base fee with metered usage. A marketplace can charge a commission and sell the same merchant a subscription and sponsored placement. The task is to identify each distinct charge, rather than assign the entire customer relationship to one bucket.

Keep a residual or mixed category. A perpetual software license, for example, is neither hardware nor necessarily a subscription. Direct sales of retail goods should not be relabeled “hardware” merely to fit the six buckets. A credible decomposition can say “not separately disclosed”; an apparently complete one can conceal unsupported assumptions.

2. Read disclosures without confusing reporting labels and economic models

Financial statements often organize revenue by segment, product family, or product versus service. None of these necessarily answers how customers are charged. Microsoft’s 2025 Annual Report makes the distinction unusually clear: it discusses subscription and consumption within cloud services, while its product-and-service offering table groups several models together.

The report distinguishes several mechanisms inside one corporation. Hosted cloud software may be sold by subscription or consumption. Consulting revenue is recognized as work is provided. Hardware is generally recognized when ownership transfers; an on-premises software license can be recognized upfront. Recognition timing is not itself a revenue model, but the policy helps reveal what the customer has bought and when the company has delivered it.

Now inspect the offering table. “Server products and cloud services” combines offerings that need not share a charging model; “Windows and Devices” likewise does not isolate hardware revenue. The report’s Microsoft Cloud figure cuts across several offering lines already in the table. Adding it to those lines would count revenue twice.

This yields the core discipline for company analysis: reported categories are evidence to map, not ready-made answers. If a filing reports a blended line, retain that line as blended unless another disclosure supports a finer split.

3. Distinguish transaction volume from transaction revenue

Transaction businesses create a particularly consequential measurement trap. A marketplace may facilitate a large value of purchases while recognizing only its fees from those exchanges. The money paid to a seller is not automatically the platform’s revenue.

Watch the selected portions of Sharetribe’s Step Two: Choose your marketplace revenue model. The video is framed around marketplace design; use it here to distinguish the value exchanged, the fee charged, and additional ways to monetize the same participants.

Suppose buyers spend $\$1,000$ with independent sellers through a marketplace and the operator earns a $\$100$ commission. The transaction value is $\$1,000$; the commission is $\$100$. Whether accounting presents the operator’s revenue on a net fee basis or a gross basis depends on its role and applicable accounting judgments, so check the company’s disclosure rather than assuming either treatment.

The distinction matters beyond marketplaces. For a payments provider, payment volume is not the same as the provider’s fee revenue. For an advertiser-funded platform, audience activity creates opportunities to sell ads, but activity itself is not advertising revenue. Each model has a different link between activity and recognized revenue:

• A subscription links revenue to paid access over time, although customer count and price can change.
• Consumption links revenue more directly to usage; a customer can remain active while spending less.
• A transaction fee links revenue to the number, value, or mix of exchanges, subject to the fee arrangement.
• Advertising links revenue to sellable inventory, advertiser demand, and the agreed basis for charging.
• Services link revenue to obligations fulfilled; hardware links it to physical products sold.

This is why “recurring customer” and “recurring revenue” deserve separate scrutiny. A customer may buy cloud capacity every month, but the amount can fluctuate substantially with utilization.

4. Build a disclosed-revenue bridge: Amazon’s Q4 2024 example

Amazon’s quarterly disclosure provides a useful contrast to Microsoft: it reports several commercial categories explicitly, but those categories still do not divide neatly into our six models.

Amazon reports $\$187,792$ million of net sales for Q4 2024. The disclosed categories reconcile to that total:

For example, advertising’s reported category share is $\$17,288$ million divided by $\$187,792$ million, or approximately 9.2%. That is a supportable statement. By contrast, labeling all $\$47,485$ million of third-party seller services “transaction commissions” would contradict the footnote: the line also contains fulfillment, shipping, and other fees.

Likewise, it would be tempting to classify all AWS revenue as consumption because much cloud computing is metered. The table does not establish that split. And neither online-store revenue nor a broad “product sales” figure tells us how much is specifically hardware.

Notice what this bridge does achieve. It identifies disclosed subscription and advertising categories, locates transaction-related and cloud revenue within mixed lines, preserves product sales and other revenue rather than forcing them into unsuitable buckets, and reconciles the amounts to a common period and total. It does not manufacture a precise six-model breakdown that the source does not provide.

5. Turn the bridge into an industry-intelligence view

For a new company, work in the following order:

1. Fix scope and period. State whether you are analyzing the corporation or one business, and use one reporting period and currency for the bridge.
2. Record reported lines first. Capture their values, definitions, and total before applying your own labels.
3. Map charges to models. For each material offering, ask who pays and whether payment is triggered by access, usage, exchange, advertising delivery, work performed, or physical delivery.
4. Mark mixed lines and overlaps. Split them only when disclosed evidence permits it. Keep an explicit residual for licenses, retail products, or other items outside the six models.
5. Reconcile and qualify. Separate amounts you can calculate from hypotheses about amounts hidden inside blended lines.

A short analyst conclusion might therefore read: “Advertising and subscriptions are separately disclosed. Seller-services revenue contains commissions alongside other services; cloud revenue is disclosed without a full subscription-versus-consumption split. A precise transaction or hardware share cannot be derived from these figures.” That is more decision-useful than a complete-looking pie chart built on guesses.

Revenue mix is a starting point for strategy, not a verdict on business quality. Two firms can report similar growth but rely on different drivers: paid access, higher workload utilization, greater transaction value, advertiser spending, or more services delivered. Those drivers face different competitive pressures and can change at different speeds. The decomposition tells you which questions to ask next.

Conclusion

Decompose revenue at the level of the charge, while reconciling back to what the company actually reports. Subscriptions sell access; consumption bills usage; transactions monetize exchanges; advertising sells promotional opportunities; services compensate work or service obligations; and hardware sells physical products. Keep blended lines blended when the evidence does not support a finer allocation, and never confuse marketplace or payment volume with revenue.

The next lesson builds on this revenue map to locate profit pools. Revenue identifies where money enters the value chain; margins, growth, and bargaining power help establish where economic value is retained.