Investigating Chargeback As An Internal API Use Case Posted inStrategy Kristopher Sandoval October 14, 2021 When we use APIs — or any technology, for that matter — we tend to make the costs of those systems invisible. Yet, every time we use our phone, turn on a light switch, or scan a train pass, we engage in a transaction with a specific cost. Every call has a cost, and every resultant data response has a value.In the business scenario, one of the underlying questions is how that value should be reclaimed and reconciled. One example is a business unit using another business unit’s API internally. Or, it could be more complex, such as a multi-team corporation with hundreds of employees using an API at different intensities and for different purposes. In this scenario, value reclamation becomes a more important topic.How we assign value and then reclaim that value is core to success in many modern business environments. Below, we’ll discuss one approach to value reclamation, chargeback, and how enterprises may utilize such a system for reclaiming value in the era of web APIs.What is Chargeback?“Chargeback” is a relatively simple idea — internal customers utilize an API, and that utilization is given a value assignment. Where this becomes more complicated is the nature of the value assignment, such as whether this assignment is a direct chargeable value charged at use or an un-charged proportional representation of utilization. Additional complication arises around how that value estimation is calculated, and how the actual value is derived from this estimate.In the most simple terms, a chargeback model places the ownership of the API under one group and allows access — for a price. This price may be determined using various pricing methods, including sliding scales, allocated budget, fair use allocation, and more. This raises the question of whether it’s sensible to charge internal users for their use of an internal API. The reality of modern business is that nothing is free, and even if the utilization was uncharged, there’s still a finite amount of resources and a cost assigned to each API call. Those costs don’t go away if the requesting party is internal.Chargeback Process MethodsSo how does a chargeback actually work? Generally, chargeback systems can take two forms — manual chargeback or automatic chargeback.In a manual chargeback system, a type of reconciliation is used to match the cost of API usage with a period of time in which that use was authorized. This is done manually, matching costs with utilization and ensuring that the numbers “add up”. For instance, for a Net30 payment arrangement, an API may be used by a given team over a monthly period. At the end of this month, the recorded API calls will be collated and charged in one form or another to the body responsible for generating those calls. Payment, if applicable, will then be processed.While this flow is simple, much more complex manual chargeback systems exist, and their specific layout and form will entirely depend upon the system they govern. The key thing to remember here is that these processes are exactly what they say they are on the tin — manual. Thus, there are often accountancy requirements and teams that engage in the tracking and reconciliation process. Sometimes, this entire process can occur outside of the API unit itself and instead be a function of the accountancy body.Another approach is an automated chargeback system. In such a system, a user would have some sort of credential that defines who they are and what unit they function under. This credential would note what they can access, for what period, and for what purposes. Most importantly for this process, that credential would also define what the individual user, or in many cases, the collective unit, can be billed for their use of the defined API. When the user calls the API, the account is tracked, metered, and charged automatically based upon business agreements between the units.Again, just like the manual chargeback model noted above, there are many ways to structure this system. At the end of the day, the most important element here is that it is fully automated — perhaps aside from normal budgetary processes at the end of the quarter or year.5 Types of Chargeback Pricing ModelsThere are almost as many chargeback models as there are things to chargeback, but today we’ll look at a few common methods.1. Notional ChargingNotional charging is the idea of having the value derived from the API utilization be noted but not necessarily charged. There are many cases where the total cost will be considered sunk regardless of who is actually using the API. In this case, reclamation of value is not the chief focus, but knowing where those costs live is still vitally important to allocate resources.In such a reality, notional charging, while not a pricing model per se, is nonetheless valuable for setting the expected costs for the next year as well as controlling ongoing expenses in the current. Current use limits can be established, trends in use can be monitored, and overages and extreme use cases can be identified and rectified before the costs cause extreme overrun.2. Tiered UseAnother pricing model is to adopt a tiered use payment system. Such a system allocates costs to specific tiers — some premium, some standard, and perhaps even some free. When the API is consumed, the utilization amount is recorded and charged according to either a pre-purchased tier allocation or a sliding cost scale. In many cases, the more calls are made in a tiered system, the lower the cost, as a large number of calls at a low price may be more profitable than few calls at a normal price.3. SubscriptionA derivative system of a tiered use payment system is the subscription system. In this case, a tiered approach is still used, but with a limit in-band. For instance, a tiered system may equate 1,000 calls with a specific price on a sliding scale, with all calls treated equally. On the other hand, a subscription service may price calls depending on their volume while prioritizing calls from certain subscription classes over others. This can be a good way to derive value from calls while also ensuring that certain business functions are prioritized or deprioritized depending on the business value conversion.4. Metered UsageAnother, more straightforward form of billing is metered usage. This is the most common system, as costs are defined by singular call and are charged regardless of the volume, the class, or usage tier. A call is charged by a specific amount, and once the use is metered, this is reconciled on a 1:1 basis with an established budget. Note that there is typically no deviation in price in this model — a call is expensed per call, so 1,000 calls from one provider should be 1/10th as expensive as 10,000 calls from another provider.5. Fixed Cost ModelFinally, the fixed cost model utilizes a total aggregate cost system to divide the costs of the API by the previous term’s use. For example, if the business unit that runs the API has a set cost per year, the cost is charged by the amount of utilization by each constituent part. For instance, if the business spent one million euros on API calls last year and three org units utilized that API, they would be responsible for the costs in the next year as part of a budget reconciliation.In 2021, if that amount was one million, and the business units utilized 10%, 20%, and 70% of the API, in 2022, their budgets will be charged 10%, 20%, and 70% of that same cost. This assumes fixed costs from year to year, which, in some cases, is appropriate but requires a high level of control over the API and its systems.ConclusionUltimately, the appropriate choice of whether or not to adopt a chargeback system — and more specifically, which chargeback system to adopt — will depend largely on the organizational structure that underpins the API and its users. Specific structures, such as traditional business units with divisions and teams, would benefit from wildly different models than a flat management organization might. Figuring out the API use scenario and the possible economic impacts of each strategy will determine which model, if any, is appropriate.What do you think about chargeback systems? Do you know of any good example where this has been done transparently? Let us know in the comments below!