Here’s an invented scenario to frame the problem. A distributor with 400 Salesforce users signs a managed services contract on a Thursday. The sales sheet promised “24/7 support” and a “response within 2 hours.” On Friday at 4:30 pm, order entry breaks. An automated acknowledgment lands at 4:31 pm, a human reply arrives Monday at 10:30 am, and the provider calls the contract met, because the 2-hour target counted only business hours. Nobody lied. The contract simply never said which clock applied or who assigned the severity, and the buyer had no remedy worth the name.
This guide lists what a managed services agreement for Salesforce should spell out, using a small set of public sources that all have a stake in the answer. Salesforce’s Success Plans datasheet is vendor material. A blog from Vantage Point, a consulting partner that sells managed services, is marketing. Valiantys publishes a service level page written for Atlassian cloud products, and Silverline publishes another for its CalendarAnything app, so neither describes managed services for Salesforce. A 2008 TechTarget article is dated but general. I found no neutral benchmark of what partners deliver and no public Salesforce platform uptime terms I could verify. Treat every number as a reference point, and confirm with counsel and the provider before signing.
TL;DR
- Define the clock first. Response time, workaround time and resolution time are different promises, and a contract that names only the first protects the provider. State which hours count, who assigns severity and what pauses the timer. A Friday 4:30 pm outage can run 66 hours before a “2 business hour” target expires.
- Price the remedy honestly. Credits are usually a percentage of a single month’s fee. On an invented $5,850 monthly fee, a 10% credit is $585, set against an invented $72,000 loss from a 6-hour outage. Ask for escalation paths, root cause deadlines and termination rights for repeated misses. Those figures show why credits can’t be the main protection.
- Match the contract to the platform. Salesforce ships 3 releases a year, and its own support plans set the response terms your partner depends on. Cover sandboxes, release testing, exclusions and exit terms, then confirm everything with counsel and the provider. Name the staff who’ll do the work, too.
Start with the services the agreement covers
Scope of services. Start by listing exactly what the provider supports. TechTarget’s 2008 piece says an agreement should name what’s covered and what’s excluded, from systems to vendors, plus the minimum environment standards a client has to meet. For Salesforce, that translates into a list of production orgs, sandboxes, clouds, managed packages and integrations. A clause that promises “support for your Salesforce org” invites a dispute the first time a middleware connection breaks, because nobody wrote down whose problem it was.
Request types. Split incident work from request work, and give each its own rules. Incidents get severity targets. Service requests, such as adding a user or a field, need a size ceiling. Valiantys’s published terms say requests estimated at more than 8 hours of work must be ordered as professional services, which is a useful pattern because it stops a small-request queue from quietly absorbing project work. Vantage Point’s sample treats health checks and release readiness as separate line items too.
Shared responsibilities. Valiantys requires a problem description and an incident start time on each ticket, plus prompt notice of faults. Silverline excludes delays in customer information. Those terms are reasonable, but they only work when they’re written down: name the buyer contacts and the access the provider needs, and list what a ticket has to include. Otherwise a provider can pause the clock for missing details that the buyer never knew were required.
Severity levels that people can apply the same way
Severity definitions decide every later number, so write them before the response targets. The table below pairs Valiantys’s definitions, paraphrased from a page written for Atlassian products, with the sample targets in Vantage Point’s blog, which that firm calls “an illustrative starting point.” Mixing vendors’ wording is a drafting shortcut, with no endorsement from either vendor. Use it to see how the pieces fit, then replace every cell with language your own team can apply at 2 am without a phone call.
| Level | Definition (Valiantys, paraphrased) | First response | Workaround | Resolution |
| P1 | Application down or unreachable for all users | 1 to 2 business hours | Same business day | Root cause fixed and RCA within 5 business days |
| P2 | Main features degraded for a large group of users | 4 business hours | 1 business day | 3 to 5 business days |
| P3 | Features unavailable or abnormal for a small group or a single user | 1 business day | 3 business days | Next scheduled release |
| P4 | Feature partly broken, workaround exists, users aren’t blocked | 2 business days | No target | Monthly backlog review |
Who assigns the level matters as much as the definition. Valiantys’s page says severity follows the supplier’s “reasonable assessment,” which leaves the buyer without a lever. A better clause lets the buyer set the initial level, lets the provider propose a change with a written reason, and sends any disagreement to a named escalation contact. Silverline’s app page lists Sev 1 at 1 to 2 hours and Sev 2 at 2 to 4 hours, so ranges like these are common, though these sources serve different products.
Response time and resolution time are different promises
A response target says when a person engages with the ticket. It doesn’t say when the problem ends. Valiantys states it plainly: “The targets relate to responses only, not fixes.” It also defines the response as an active human response, which matters because an automated acknowledgment can otherwise start and satisfy a clock in the same minute. If the contract says “response,” ask the provider to define it in a sentence, and require a named engineer working the ticket.
Resolution targets are where providers differ. Vantage Point’s sample sets separate targets for a workaround and for a fix, with the fix for a P1 expected within 5 business days alongside a root cause analysis. Silverline’s wording for Sev 1 is that work continues until the issue is resolved or a workaround is given, which is a commitment to effort, not to an outcome. Both approaches can be fair, but the buyer should know which model is being sold and price it accordingly.
Platform defects complicate resolution because the fix may sit with Salesforce. Its Success Plans datasheet lists a 2-day response during local business hours on the Standard plan and a 1-hour response for business-stopping issues on Premier, with 24/7/365 case submission. A partner can’t promise a faster platform fix than the platform vendor gives. What the partner can commit to is how quickly it opens and escalates the Salesforce case, how often it reports status, and what workaround it offers meanwhile.
Coverage hours, time zones and holidays
- Write the business day out. Valiantys defines business hours as Monday to Friday with 10-hour days, running 8am to 6pm CET in Europe and 8am to 8pm EST in North America, and it excludes Christmas Day and New Year’s Day. That level of detail is the right model. A contract that says “business hours” without a time zone, a start time and a holiday calendar leaves every Friday afternoon and every regional holiday open to argument.
- Reserve 24/7 for the top severity. Vantage Point says it recommends round-the-clock coverage only where justified, and notes that it doesn’t offer 24/7 coverage itself. Valiantys’s Starter tier covers S1 around the clock and nothing else, while its higher tiers add 24/5 for the other severities. Pick the level that matches the cost of downtime for your business, then price after-hours work explicitly. Ask what the after-hours team can really fix overnight.
- Know how little of the week is covered. A 40-hour business week covers about 23.8% of the 168 hours in a week, and Valiantys’s 10-hour, 5-day model covers about 29.8%. Everything outside those windows depends on the pause rules, the after-hours policy and the escalation phone number, so those terms deserve more attention than the headline response figure. That’s why a headline target of 2 hours says little about what happens on a Saturday morning.
- Spell out the after-hours policy. TechTarget’s article says an agreement should state what happens to service requested outside normal hours. I found no public benchmark for after-hours surcharges in Salesforce managed services, so ask for the rate, the minimum charge and the approval method in writing, and confirm with the provider whether urgent after-hours work draws down a prepaid hours pack. Without those details, an urgent weekend call becomes a negotiation at the worst moment.
Clock pauses and what stops the timer
Outside coverage hours. Business-hour clocks stop at close of business and restart at the next opening. In the invented scenario, a P1 reported Friday at 4:30 pm has 30 minutes of business time left that day. A 2-business-hour target therefore expires Monday at 10:30 am, which is 66 elapsed hours later. The same Friday report under a 4-business-hour P2 target runs to Monday at 12:30 pm, or 68 hours, and a 1-business-day P3 target runs 72 hours.
Waiting on the customer. Vantage Point notes that clocks may pause while the provider waits on the customer, and Silverline excludes delays in customer information. Pausing is fair when the provider has asked a specific question. Ask the contract to require a written request for the pause, to restart the clock the moment the buyer replies, and to record the pause in the ticket so the monthly report can show it. Open-ended pauses make a response target unfalsifiable.
Waiting on the vendor. Vantage Point also lists vendor escalation as a pause condition. That’s the riskiest of these pauses, because a P1 involving a Salesforce case could then sit paused for days while the partner’s own duties go undone. Limit the pause to the portion of the work that only Salesforce can perform, keep the partner’s status updates and workaround efforts on a running clock, and ask for a named owner who reports at set intervals until the Salesforce case closes.
Service credits and what they actually remedy
Credits are the part of an agreement buyers read last and rely on most. Vantage Point says they’re usually a percentage of a single month’s fee and rarely compensate for real disruption. An invented example shows the gap. On a $5,850 monthly fee, a 10% credit is $585, a 25% credit is $1,462.50 and the full month is $5,850. If a 6-hour order-entry outage costs an invented $12,000 an hour, the loss is $72,000, and even a full month’s credit covers about 8.1% of it.
Valiantys publishes a fairly detailed credit schedule among those I found, again for Atlassian cloud products. Uptime credits step through 10%, 25% and 100% of monthly cloud fees, a missed recovery objective earns 60% of the month’s fees (which would be $3,510 on the invented fee), and total credits can’t exceed that month’s fees. The buyer must claim within 30 days with failure dates and redacted logs or forfeit the credit, and the supplier’s downtime and credit decisions are final and binding. Read those as the provider-friendly end of the range.
Silverline’s app page takes a different approach: if the service is unusable for more than 4 weeks consistently within a month, a credit is added to the term, not refunded, and the customer must request it in writing within 5 days. Compare the models on a few points. Ask whether credits apply automatically or only on claim, whether they’re cash or term extensions, whether they survive termination, and whether they’re the sole remedy. A credit that’s the only remedy for a serious failure deserves a negotiation, not a signature.
Reporting metrics worth asking for
Vantage Point lists the monthly measures it recommends, and the list is a sensible starting menu for a buyer. The table below restates them with the question each answers. Buyers rarely need all of them. Pick the handful that map to the risks in the contract, require the provider to deliver them on a fixed date each month, and make the raw ticket data available so the numbers can be checked instead of taken on trust.
| Metric | What it shows | Why a buyer cares |
| First response by severity | Time to the first human reply for each level | Tests the headline promise against actual tickets |
| Time to restore for P1 and P2 | Elapsed time until service returns | Shows whether resolution targets are real |
| Backlog age at 30, 60 and 90 days | How long open work has waited | Reveals queues that quietly stall |
| Change success rate | Share of deployments that worked without rollback | Signals the quality of release and change control |
| Release readiness | Progress on testing before each Salesforce upgrade | Prevents surprises on upgrade weekends |
| Health-check findings closed | Share of identified issues actually fixed | Shows whether advice turns into work |
| Adoption and CSAT | User activity and satisfaction scores | Gives a view beyond ticket counts |
Vantage Point also recommends a quarterly governance review in addition to the monthly report. A report only helps if someone acts on it, so write down who attends, what decisions the meeting can make and how a missed target turns into a documented follow-up. I found no public source that benchmarks which of these metrics buyers use most, so treat the list as a single vendor’s view and trim it to fit your own priorities.
Release management across the 3 yearly upgrades
- Put the release calendar in the schedule. Vantage Point says Salesforce ships 3 releases a year. For Winter ’27, Salesforce’s admin blog lists a sandbox preview beginning August 28 and production arrivals on September 4, October 2 and October 9, which leaves 7 to 42 days between preview and production depending on the instance. The same page contains a conflicting January 8 reference for a sandbox deadline beside an August 27 date, so confirm the dates with Salesforce’s Trust site before relying on them.
- Name the tests the provider will run. The Salesforce blog advises testing features in a sandbox against your customizations and reviewing Release Updates in Setup. A good agreement turns that advice into duties: a regression script list, a sign-off from a named buyer contact, and a written readiness report delivered a set number of days before the first production date. I found no public benchmark for that lead time. Each duty should appear as its own numbered item.
- Decide who pays for release work. Some providers treat release testing as part of the monthly fee, and others bill it against hours. The sources I read don’t say which is typical, so ask directly, and confirm with the provider whether a failed upgrade that needs emergency fixes counts as an incident, a change or billable project work. Release weekends can produce urgent unplanned work, so the answer decides who bears that cost.
Environments, sandboxes and change control
Silverline’s app page excludes sandbox and other non-production environments from its service levels, which is a common pattern and a quiet risk. If the provider’s team does its testing in a sandbox, then a broken sandbox delays every release even though no production user is affected. Ask whether sandbox support has any target at all, even a loose goal, and whether sandbox refreshes are scheduled in the contract. A refresh before a release preview is worth a calendar entry.
Change control belongs in the agreement because most production incidents follow a deployment. Vantage Point tracks change success rate as a monthly measure, and Valiantys sets change review targets of within 16 hours on its Standard support level and within 8 hours on Advanced. The buyer should know who can approve a production deployment, whether emergency changes follow a shorter path, and how the provider records a rollback. A change log the buyer can read without asking is a low-cost safeguard.
Backup and security were both missing from every source I read. TechTarget’s 2008 article doesn’t cover security terms, and the vendor pages focus on response times. That doesn’t mean the topics are unimportant. Ask the provider who holds production access, how credentials are removed when staff leave the account, whether data is exported or backed up on a schedule, and where those records live. Then ask counsel to confirm that the answers appear in the contract instead of a sales deck.
Root cause analysis and problem management
Deadline and owner. Vantage Point’s sample ties the P1 fix and a root cause analysis to 5 business days. A deadline matters more than the format, because without a deadline the analysis arrives after attention has moved on. Name the author, the delivery date and the reader. If Salesforce caused the incident, the contract should say the provider will request the vendor’s explanation and pass it along, while recognizing that the vendor may not supply an explanation on the same schedule.
Required content. A useful analysis states the timeline of events, the confirmed cause, the fix that was applied and the change that prevents a repeat, each with an owner and a date. Ask for it in writing and keep it in a shared location. A verbal summary on a call can’t be reviewed later, and it can’t support a termination argument when the same failure appears again next quarter. Keep each entry factual and brief.
Repeat incidents. Problem management means tracking incidents that recur and fixing the cause instead of restoring service each time. Vantage Point recommends termination rights for repeated P1 misses, which gives problem records a legal consequence. Define repeated in numbers, such as 2 missed P1 targets in 90 days (an invented threshold, to be negotiated), and require that the provider records every incident against a problem entry so that the count is auditable.
Exclusions that deserve a close read
Every service level agreement has exclusions, and the list tells the buyer where the provider’s responsibility ends. The table below collects exclusions from the Silverline and Valiantys pages, which describe different products, so treat it as a catalog of common wording, since it says nothing about what any Salesforce partner offers. Read each row as a prompt to ask what the exclusion does to the buyer’s real risk. Most of the wording is standard, but the details vary.
| Exclusion | Where it appears | What to ask |
| Sandbox and non-production environments | Silverline | Is there any target for sandbox support? |
| Customer-made modifications | Silverline | Who decides whether a change caused the issue? |
| Third-party integrations | Silverline | Are named integrations covered separately? |
| Delays in customer information | Silverline, Valiantys | What counts as a delay, and is it logged? |
| Force majeure | Silverline, Valiantys | Does the definition include vendor outages? |
| Salesforce planned maintenance | Silverline | How much notice does the buyer get? |
| Actions of the client or third parties | Valiantys | Does this include the buyer’s other consultants? |
| Suspension for non-payment | Valiantys | How long is the cure period? |
The platform vendor’s own terms sit underneath all of this. Vantage Point notes that Salesforce’s commitment has exceptions for planned downtime and events outside its control, and I couldn’t find the underlying uptime terms in a source I could verify. Salesforce’s Success Plans datasheet adds a second layer: Premier is priced at 30% of net license spend (some editions are listed as including it), and Signature is custom priced. On an invented $400,000 of annual net spend, Premier would cost $120,000 a year, or $10,000 a month, so check that your partner’s service levels and the Salesforce plan don’t duplicate or contradict each other.
Governance, escalation and named roles
- Write the escalation ladder with names and times. Silverline’s page sends issues to a customer success manager and then to executives, and Vantage Point recommends escalation paths in the contract. Valiantys’s page describes no escalation process at all. A good schedule lists each rung by role, gives a time after which an unresolved severity level moves up, and gives the buyer a phone number that reaches a person outside the ticket queue.
- Fix the review cadence. Vantage Point proposes monthly metrics and a quarterly governance review. Put the dates in the calendar when the contract is signed, and record decisions in writing. Meetings without minutes tend to repeat the same discussion, and a buyer who wants to terminate for repeated misses needs the documented history to show the provider had a chance to fix them. Without a written record, the same disagreements will keep returning.
- Name the people who do the work. Salesforce’s datasheet describes a customer success manager as a single point of contact, which is the type of role a buyer should ask a partner to staff by name. Ask for a lead, a named backup and a notice period for staffing changes. Vantage Point’s Essentials package lists senior hours, so ask how seniority is defined and whether a junior can bill against the same pack.
- Agree how disputes get settled. Severe disagreements and credit claims will happen. Decide in advance who hears the first appeal, how many days they get to answer and what the buyer can do if the answer is late. Valiantys makes its downtime decisions final and binding, which is exactly the kind of clause to negotiate down to a documented review with an outside option. A written appeal path keeps small disagreements from becoming contract breaches.
Pricing models and how hours packs interact with service levels
Vantage Point’s Essentials package gives a concrete example of an hours-based model, which pairs 30 senior hours a month with a $5,850 fee on Salesforce or multi-platform work, or $4,800 for HubSpot, on a 12-month term with licenses excluded. That works out to $195 an hour and $160 an hour, or $70,200 over the year for 360 hours on the Salesforce price. It’s a single vendor’s package and it’s marketing, so use it as an illustration of structure, not as a market rate. I found no neutral price benchmark.
The structure raises a question the SLA has to answer: do incidents draw down the pack? Suppose, in an invented month, 8 hours go to incident work. That leaves 22 hours for enhancements, and at $195 an hour the incident share is worth $1,560 and the enhancement share $4,290. If response targets are strict and incidents are common, the buyer pays for the same hours twice in effect: once as support and once as lost build capacity. Ask for incident work to be priced separately or capped.
Fixed hours also shape behavior. A provider with a hard monthly cap has little reason to take on an unplanned P1 at 3 am unless the contract names the extra rate, and a buyer with a pack of unused hours has little reason to open tickets early. The remedy is a written rule for overages, rollover and after-hours work. Confirm with the provider whether unused hours expire monthly or carry forward, and whether the 12-month term allows a change in package size after a review.
Termination rights, onboarding and exit
Termination for repeated misses. Vantage Point recommends termination rights for repeated P1 misses, and TechTarget’s 2008 article lists the termination clause and the limitation of liability among the elements an agreement should contain. Write both: the number of misses that triggers a right to exit without penalty, the notice period, and whether prepaid fees come back. A 12-month term, like the term Vantage Point lists for its package, makes this clause the buyer’s main protection against a poor fit.
Onboarding period. I found no public source that says how long a provider typically needs before its targets apply. That’s a gap worth closing in the contract. Ask for a stated start date for each target, a handover plan covering documentation and access, and a written list of known issues in the org on the first day, so the provider isn’t blamed for old defects and the buyer isn’t charged to discover them.
Exit assistance. The way out deserves as much detail as the way in. Ask for documentation of changes made during the term, return or deletion of credentials, a transition period at the agreed rate, and a statement of what happens to unused hours. Valiantys forfeits credits on account termination, and Silverline adds credits to the term, so a buyer leaving the relationship should know whether earned credits survive. TechTarget advises that attorneys for both sides review the final documents.
A review checklist before you sign
Close with a pass through the contract using the questions below. The list is a synthesis of the points in this guide and isn’t taken from any single source. Each row names a clause, the question that tests it and the signal that should slow the signature. Adapt it to your risk, hand it to the person who negotiates the agreement, and ask counsel to confirm the final wording. Print it and mark each row.
| Clause | Question to ask | Red flag |
| Scope | Does it list orgs, clouds, packages and integrations? | “Support for your Salesforce org” |
| Severity | Who assigns it, and who can dispute it? | Supplier decides alone |
| Targets | Are response, workaround and fix separate? | Response only |
| Hours | Which time zone, hours and holidays apply? | “Business hours” with no detail |
| Pauses | What stops the clock, and for how long? | Open-ended vendor pause |
| Credits | Are they automatic, and are they the sole remedy? | Claim window of a few days |
| Releases | Who tests each upgrade, and by when? | No named duty |
| Exit | What triggers termination, and what’s handed over? | No right to leave |
The checklist works best when a business reader and a technical reader complete it independently and then compare. Disagreements usually point to a clause that means different things to different readers. Fix those before signing, because after signing the same ambiguity will surface during an outage, when resolving it costs the most. Keep the completed list with the contract. Comparing the completed copies in a short meeting should be enough to find the gaps.
FAQs
What should a Salesforce managed services SLA cover at minimum?
At minimum it should cover the scope of services, severity definitions, response and resolution targets, coverage hours with a time zone, the rules that pause the clock, credits and remedies, reporting, escalation, exclusions and termination rights. The sources I read agree on most of this, but none of them is neutral, so use the list as a prompt and confirm with counsel. Backup and security terms were missing from every source, so add them yourself.
How fast should a provider respond to a critical Salesforce issue?
Published figures range from 1 to 2 business hours in Vantage Point’s sample and Silverline’s app page to 1 hour for the higher Valiantys tiers, while Salesforce’s Premier plan lists a 1-hour response for business-stopping issues. Those pages describe different products and coverage models, so the numbers aren’t directly comparable. Ask the provider what counts as a response, whether the clock runs outside business hours and what happens if the target is missed.
Should the SLA include 24/7 coverage?
Only if downtime costs enough to pay for it. Vantage Point recommends 24/7 coverage only where justified, and Valiantys limits round-the-clock coverage to its top severity on the entry tier. A common compromise is 24/7 for the top severity and business hours for everything else. Price the after-hours work explicitly, and check whether it draws from a prepaid hours pack, since that determines whether the coverage is real or just a phone number.
What’s a reasonable service credit?
I found no neutral benchmark. The examples I read run from 10% to 100% of a month’s fees in Valiantys’s schedule, with 60% for a missed recovery objective, while Silverline adds time to the term. Vantage Point says credits rarely compensate for real disruption. Treat credits as a nudge, and ask instead for escalation, root cause deadlines and the right to terminate after repeated misses. Confirm the claim process and caps with the provider.
Does the partner’s SLA replace Salesforce Premier or Signature support?
No. They sit on different layers. The partner’s agreement covers its own response and work on your org, while a Salesforce plan covers Salesforce’s response on platform cases. Per the datasheet, Standard lists a 2-day response during local business hours, Premier a 1-hour response for business-stopping issues, and Signature adds 15- and 30-minute updates for those issues. Confirm with both parties how a platform case moves between them, and who owns it meanwhile.
How do hours-based packages work with an SLA?
Usually the pack buys time and the SLA sets the speed. Vantage Point’s Essentials package, for example, lists 30 senior hours a month. The open question is whether incident work consumes those hours, which can leave little for planned improvements. In an invented month with 8 incident hours, 22 hours remain. Ask the provider in writing how incidents, overages, rollover and after-hours work are billed before you sign. Get that answer in the contract.
Who decides how severe an incident is?
The contract should say. Valiantys’s page assigns severity by the supplier’s reasonable assessment, which favors the provider. A fairer clause lets the buyer set the initial level and lets the provider propose a change with a written reason, with a named escalation contact settling disagreements. Because severity drives every response target and credit, a vague clause here weakens everything else in the agreement, so settle it before you negotiate the numbers.
What should I check before signing?
Check that every target has a defined clock, a named owner and a remedy. Read the pause rules and the exclusions twice, because that’s where most disputes begin. Ask how releases are tested, what reports arrive monthly, who staffs the account and what you get if the provider repeatedly misses targets. Then ask counsel to review the contract, including limitation of liability, since I’m not a lawyer and the sources here are commercial.
