In this article
Do not promise a B2B customer a faster response until you know who will be available, which incidents qualify, how time will be measured, and what happens after a breach. A good service-level agreement turns “fast support” into a product with boundaries and a cost. A vague SLA sells unlimited priority at a fixed price, leaving operations to fund the difference through on-call coverage, escalation and out-of-scope work.
This is an operating and pricing model, not legal drafting. Have the contract reviewed for the customer’s jurisdiction and the nature of the service. The objective is to make sales and operations agree on a promise they can measure and finance before it enters a contract.
Separate four promises that are often confused
Availability measures whether a service functions during a defined window. Response time runs from receipt of a valid incident to acknowledgment and the start of action. Restoration time focuses on returning the affected function, potentially through a workaround. Root-cause resolution may require a longer investigation, change and test. Combining them into one commitment creates disagreement even when the team works hard.
The published Google Compute Engine SLA illustrates why covered service, downtime, the measurement method, credits and exclusions must be defined together. Do not copy its percentages or conditions into your own service. The useful lesson is that an impressive number has little meaning without a boundary and an agreed remedy.
Define the clock before choosing the target
Specify when measurement starts: arrival in an approved channel, receipt of mandatory information, or assignment of severity. State service hours, time zone and holidays. “Within one hour” is a very different obligation during business hours and around the clock. Define when the clock pauses, such as while awaiting customer access or approval, and record the reason and duration instead of stopping it without evidence.
Base severity on impact rather than the requester’s urgency. A critical incident might stop a core function for a meaningful group of users or create a defined transaction or information risk. One person’s urgent request is not automatically critical. Give included and excluded examples for each level, identify who can reclassify an incident, and describe how the customer can challenge the decision.
Convert the promise into a capacity model
Use three to six months of operating records to estimate incidents by severity, arrival by hour, hands-on work and escalation rates. A monthly average is insufficient: ten simultaneous incidents need different capacity from ten spread across a month. Include coverage shifts, on-call time, scarce skills, redundancy, monitoring tools and service management.
If evidence is limited, start with one customer or a narrow scope and a cap on included incidents. Do not set a permanent price from one quiet pilot month. Link repricing to changes in users, locations, integrations or coverage hours rather than waiting until the margin fails and then surprising the customer.
A hypothetical premium-response price
Assume a B2B service company offers a premium tier to ten customers. It promises a 30-minute response to critical incidents around the clock, compared with four business hours on the standard tier. The team expects 40 included premium incidents each month. All figures are hypothetical and are not results attributed to Fahad ALNaimi or any particular company.
- 28 additional monthly hours for coverage and escalation × QAR 120 per hour = QAR 3,360.
- Monthly on-call allowance for the coverage team = QAR 2,000.
- Additional monitoring, reporting and service-management tools = QAR 540.
- An expected service-credit provision based on pilot evidence, not a guaranteed outcome = QAR 600.
Estimated incremental cost is QAR 6,500 a month, or QAR 650 per customer if allocated equally. If the target contribution margin is 30% of price, the team should not simply add 30% to cost. Dividing QAR 650 by 0.70 gives an indicative price of about QAR 929 per customer before account-specific costs. Adding 30% produces QAR 845 and an actual margin of only about 23%.
This is not a recommended market price. Allocate more cost to a customer that generates most incidents or needs a rare skill. Avoid counting rework twice inside both labor and the credit provision. Stress-test simultaneous incidents; three critical events at once may require a second responder or an alternative escalation path.
Use credits as an incentive, not fictional compensation
Define credits against the affected portion of the service fee rather than automatically applying them to the whole commercial relationship. Set a cap, claim window and reasonable evidence. A credit addresses a failed commercial commitment but may not equal the customer’s actual loss, so it cannot replace risk controls or liability terms reviewed by appropriate counsel.
Clearly identify planned maintenance, customer delay and dependencies outside your control when an exclusion is justified and demonstrable. A long list of vague exclusions destroys trust. When your promise relies on a third party, reflect that dependency in the SLA rather than selling a number the supply chain cannot support.
Watch percentiles, not only averages
Report compliance rate, 90th-percentile response, restoration time, reopened incidents, credits as a share of tier revenue, and contribution margin after coverage cost. An average can hide a smaller group of customers waiting much longer. Separate each severity and coverage window.
Connect the report to account cost-to-serve so a contract does not look profitable while escalations consume its margin. If the commitment supports a subscription relationship, review it alongside recurring-revenue design. Renewal quality matters more than hiding support cost inside a monthly fee.
The operating decision
Begin with two tiers, impact-based definitions, one measurement window and a limited two-month pilot. Review capacity, margin and concurrent incidents before expanding. Do not sell a premium tier whose success depends on one hero who can never take leave. The best SLA is not the harshest promise; it is a specific promise with an owner, capacity, price and evidence path for both success and failure.
Contact on WhatsApp
