Pull the original Salesforce business case out of the drawer and read it again. The promise was something like this: a unified view of every customer, productivity gains for sales reps, predictable forecasting, and a platform you would never outgrow. The total cost of ownership over three years would pay back inside 18 months. The implementation partner had a slide that proved it.
Three years later, the picture looks different. The platform is in place. The reps use it because they have to. Forecasting is still inaccurate. The cost has roughly tripled compared to the original quote. And the unified view of the customer turned out to be a unified view of pipeline, with all the operational data still living somewhere else.
This is not a Salesforce hit piece. It is the most honest accounting we can give of why ROI rarely shows up the way the original deck predicted, and what the path forward actually looks like.
Reason 1: The Implementation Partner Was Not Incentivized to Finish
The first economic reality is the partner relationship. The implementation partner is a separate business from Salesforce, and they bill by the hour. Their economics depend on hours billed, not workflows completed. When the project plan calls for three sprints to go live, those sprints will fill the time available. When you call them back to add a new field, that is also billable. When the field breaks the report, fixing the report is billable.
This is not malice. It is structure. The partner is doing what their business model rewards, which is staying engaged with you for as many quarters as possible. The original ROI model assumed the partner would finish and leave. They almost never finish. They graduate from build to support to enhancement, and the meter keeps running.
The total cost of ownership in the first business case rarely included the partner's ongoing work after go-live. It was modeled as a one-time line. In practice, partner spend continues at 30 to 60 percent of the original implementation cost every year, indefinitely.
Reason 2: The Customization Treadmill
The second reason is the customization treadmill. Every workflow change costs $250 per hour, sometimes more, sometimes routed through a Salesforce admin who reports to the CFO and is overloaded with backlog. The result is that small, sensible adjustments get queued up for weeks. The team works around them with shadow spreadsheets. The shadow spreadsheets become the actual source of truth. The Salesforce data drifts.
Compounding this: the partner who built the original implementation often baked in custom objects and triggers that future changes have to navigate. Removing a field is not a click. It is an analysis, a regression test, and a release window. By year three, the customizations you no longer use are still there, still costing maintenance time, and still slowing every new build.
The customization treadmill is the reason most Salesforce orgs accumulate technical debt faster than they pay it down. The platform is flexible, which means it bends to whatever the previous admin wanted. The next admin spends six months understanding what they inherited.
Reason 3: The AppExchange Tax
The third reason is what we call the AppExchange tax. Salesforce out of the box does not do everything mid-market companies need. To close the gaps, the typical org installs five to ten add-on apps from the AppExchange. Each one costs between $25 and $150 per user per month. Each one has its own license model, its own admin overhead, and its own integration risk.
For a 50-user organization, that math gets ugly fast. Seven add-ons at an average of $75 per user per month is $31,500 per year. Add ten more for premium tools and you are well past $45,000 just in AppExchange spend, on top of the base license fee. Each app you remove breaks something. Each app you keep costs you.
Reason 4: Einstein and the Generic AI Problem
The fourth reason is the AI layer. Einstein is a credible product, but it is generic by design. It learns patterns across all Salesforce customers, which means the predictions it makes for your business are based on aggregated patterns rather than the specifics of your operation. The lead score it gives you is not wrong. It is just not as useful as a model trained on your historical data.
The other issue is access. Einstein operates within the Salesforce data model, which means the operational data that lives outside Salesforce, in your accounting system, your project tool, your customer support platform, never enters the prediction. You get a forecast that is informed by pipeline activity but not by actual project margins, billing realities, or service health.
For a generic insurance use case, this is fine. For a structural engineering firm, a fresh produce distributor, or an estate planning practice, the generic patterns are not the patterns that matter. The AI value lives in the operational data that Salesforce does not see.
Where Salesforce Still Wins
We have to be honest about this. Salesforce remains an excellent fit for organizations with very large, complex, multi-region deployments where the ecosystem of integrations matters more than the platform itself. If you have 1,000 reps in 30 countries and you need every conceivable integration partner to work with your CRM, Salesforce earns its keep. The 200 plus third-party integrations that exist in the Salesforce ecosystem are unmatched.
Salesforce also wins for organizations with a strong internal admin team that can rein in customization requests and force discipline on the platform. If you have a CRM director with veto power, the customization treadmill is much shorter.
For mid-market firms with 20 to 200 employees, none of those conditions usually hold. The integrations needed are five or six, not 200. The internal admin team is one person, often shared with another role. The economics of the platform stop making sense around the third year.
The Wrap and Replace Pattern
The path forward is not to rip out Salesforce and replace it with another mega-platform. We have written about that elsewhere. See our walkaway story for the full case study, where a 50-employee structural engineering firm with offices across California, the largest hillside specialist in the state, was 85 percent migrated to Salesforce and chose to walk away. The replacement was not a different CRM. It was a wrap layer.
The wrap layer reads from your real systems of record (accounting, project tracking, document storage) and exposes a unified view to the team. The CRM-equivalent functions sit in the wrap layer. The cost is a fraction of Salesforce plus AppExchange plus partner fees. The customization is yours, owned in code, not rented from a partner.
If this sounds like the path your firm should consider, we walk through the full economics in our Salesforce alternative analysis. The piece that surprises most CFOs is how quickly the math turns around once partner fees and AppExchange spend roll off.
The Real Question
The question to ask is not whether Salesforce is good or bad. It is whether the value Salesforce provides at your scale, in your industry, with your team, justifies the total cost of ownership including all four of the categories above. For some companies, the answer is still yes. For most mid-market firms we talk to, the answer changed somewhere between year two and year three, and nobody re-ran the math.
If your Salesforce ROI never showed up the way the slide deck predicted, you are in the majority. The fix is not another consultant. The fix is to look at what you actually need, what your data actually shows, and whether a wrap layer or a different architecture would deliver the value at a fraction of the cost.