Custom Software vs. SaaS Sprawl: When the Stack Becomes the Bottleneck

By MercConsulting · Published 2026-08-13 · Updated 2026-08-30

Off-the-shelf software fits about 70% of how your business runs. The other 30% you cover with workarounds and more subscriptions. Here is how to know when the stack is the bottleneck and when building pays.

Off-the-shelf software fits roughly 70 percent of how any business actually runs. The remaining 30 percent — the way you quote, the way you schedule crews, the approval step your biggest customer insists on — gets covered by workarounds: duplicate data entry, exported spreadsheets, and one more subscription. Custom software becomes the right answer when the cost of those workarounds — labor hours, errors, per-seat fees, and slow decisions — exceeds the cost of building the missing layer. And that crossover point has moved: AI-assisted development has cut typical build costs to a fraction of what they were five years ago, which means businesses that were priced out of custom software in 2020 often are not today.

That does not mean every business should build. If your process still changes every quarter, or a workflow only runs a handful of times a month, a $40-per-month tool and a little tolerance is the better trade. The real decision is not custom versus off-the-shelf as a philosophy. It is a line-by-line question: which parts of your operation are commodity, and which parts are the business?

Here is how to tell when the stack itself has become the bottleneck, what sprawl actually costs, when building is the wrong move, and how the build-versus-buy math has changed.

The 70 percent problem: why every tool almost fits

Every SaaS product is built for the average version of a business like yours. The CRM assumes a sales process. The project management tool assumes work arrives as tasks. The field-service app assumes jobs get dispatched a certain way. For the generic 70 percent of your operation — contacts, invoices, calendars, files — that assumption holds, and the tool works.

The trouble is the other 30 percent, because that is usually the part that makes you money. Your pricing logic. The two-step approval your commercial accounts require. The way a job moves from estimate to crew to invoice to warranty. No vendor built for that, because no vendor knows your business exists.

Owners respond in one of two ways. They bend the business to fit the tool, which quietly sands off the edge that made them different. Or they pay people to bridge the gap, which is how a company ends up with a full-time employee whose actual job is moving data between screens. Neither cost appears as a line item. Both show up as payroll, errors, and slow weeks.

Six signs the stack is the bottleneck

You do not need a formal audit to know. If three or more of these describe your operation, the stack is costing more than it saves:

• Duplicate data entry. The same customer gets typed into the CRM, the accounting system, and the scheduling tool. Three chances to typo a phone number; zero systems you fully trust.
• Swivel-chair work. Someone spends hours a day copying from one screen into another. That is a human API, usually earning $18 to $28 an hour to do what an integration does for pennies.
• Per-seat fee creep. You buy seats not for people who work in the tool but for people who occasionally need to look something up. Ten lookup-only seats at $45 each is $5,400 a year for read access to your own data.
• The workflow bends to fit the tool. You changed how you quote, or dropped a service variation, because the software could not model it.
• Spreadsheet glue. The real system of record is an export someone massages every Monday. If that person is out sick, reporting stops.
• No straight answers. A simple question — what did we actually make on that job? — takes a day and three tools to answer, so it mostly goes unasked.

What sprawl actually costs

The subscription line is the number owners see, and it is the small one. A typical 10-to-50-person business now runs 20 to 40 SaaS products; $2,000 to $8,000 a month is common once you count the tools bought by individual departments on company cards. Annoying, but survivable.

The bigger costs hide in operations:

• Bridging labor. One employee spending two hours a day re-keying and reconciling, at a loaded $25 an hour, runs roughly $13,000 a year. Most sprawled stacks demand that from three or four people, not one.
• Error cost. Re-keyed data drifts. A quote built on a stale price list, a missed follow-up because the lead never reached the CRM, an invoice that never went out — one lost $8,000 job covers a lot of software.
• Decision lag. When margin reports take a week to assemble, you find out a service line went underwater a month after it happened. Slow data is expensive in ways that never get attributed to the stack.

Run the numbers on a real shape: a Houston services firm with 30 seats spread across 14 tools was paying about $4,600 a month in subscriptions — and burning roughly 60 hours a month of staff time bridging them. All-in, the stack cost north of $10,000 a month and still could not answer job-level profitability without a spreadsheet.

"We were paying for eleven different tools, and the business still ran out of one spreadsheet only two people understood."

When custom software is the wrong answer

Custom builds have a bad reputation mostly because people build at the wrong time. Custom is the wrong answer when:

• The process is still moving. If you would design the workflow differently next quarter, you are pouring concrete on wet ground. Stabilize the process on cheap tools first; software freezes whatever it touches. Map the process before anyone writes code.
• Volume is tiny. A workflow that runs five times a month does not justify a build, no matter how annoying it is. A checklist and a template are fine.
• The function is commodity. Never rebuild accounting ledgers, payroll, or email. QuickBooks is not your bottleneck; the swivel-chair work around it is.
• Nobody owns it. Custom software needs one person inside the business accountable for it — deciding what changes, testing, keeping the builder honest. No owner, no build.
• The brief is "get everything in one place." That is not a problem statement; it is a mood. Good builds start from a specific, measured pain — hours, error rates, response times — not from a desire for tidiness.

When building pays

The strongest custom-software cases share one profile. Check your candidate workflow against five rules:

• It is core to how you make money, and you do it differently from competitors on purpose.
• It is stable: you have run it substantially the same way for six months or more.
• It has real volume: daily or hourly, not monthly.
• You can name the number it fixes — hours of re-keying, quote turnaround, error rate, speed to lead.
• You genuinely tried the off-the-shelf options and found yourself fighting them rather than configuring them. If you have not run that comparison honestly, start with off-the-shelf versus custom AI before committing either way.

Notice what is not on the list: company size. A five-person firm with a high-volume, differentiated workflow is a better custom candidate than a fifty-person firm running commodity processes.

How AI-assisted development changed the math

Five years ago, a custom line-of-business application — intake, scheduling, quoting, invoicing hooks — typically ran $100,000 to $250,000 and took six to twelve months, with a maintenance contract behind it. At those prices, only severe stack pain justified building, and most owners were right to live with SaaS sprawl.

AI-assisted development changed the inputs. Experienced teams using modern AI tooling routinely deliver the same scope for a fraction of the old price — often in the $20,000 to $75,000 range depending on integrations and complexity — in weeks rather than quarters. Iteration got cheaper too: the change you would have batched into "phase two" next year now ships next week.

What did not change matters just as much. Discovery is still the hard part; software built on an unmapped process automates the mess. Integration quality still separates systems that work from systems that almost work. And someone still has to maintain what gets built. AI made building cheap. It did not make building the wrong thing cheap. A disciplined shop still spends more time understanding how the pieces of your business connect than writing code.

The shape of what gets built changed as well. The modern pattern is not a monolith that replaces everything. It is a connective layer over a commodity core: keep QuickBooks, keep the calendar, keep the phone system — build the spine that moves a job from intake to crew to invoice without a human re-typing anything, with AI handling the classification and drafting steps in between.

The middle path: buy the commodity, build the difference

Most businesses do not face a binary choice. The sequence that works:

That sequencing is most of what a good AI and custom software partner does before writing any code: separating the commodity 70 percent worth renting from the 30 percent worth owning.

Frequently Asked Questions

How many SaaS tools is too many for a small business?

There is no magic count; a 15-person firm can run 25 tools cleanly or 12 tools badly. The test is overlap and bridging: if the same data lives in three systems, if someone re-keys between them daily, or if you pay for seats people only use to look things up, the stack is oversized for how it is wired, whatever the count.

What does custom software cost for a small business in 2026?

With AI-assisted development, focused line-of-business builds — an intake-to-invoice workflow, a quoting engine, a customer portal — typically land between $20,000 and $75,000 and deliver in weeks to a few months. Complex integrations and regulated data push costs up. The old six-figure, year-long benchmark no longer reflects what disciplined teams charge for scoped builds.

Should I replace QuickBooks or my CRM with custom software?

Almost never. Accounting, payroll, and core CRM are commodity functions with mature, inexpensive, well-supported products; rebuilding them buys you maintenance burden and no advantage. The better pattern keeps those systems as the core and builds the connective layer around them, so data flows between intake, operations, and invoicing without manual re-entry.

When is custom software a bad idea?

When the process it would automate is still changing, when volume is low, when the function is commodity, or when nobody inside the business will own the system. Custom software freezes whatever process it captures — if you would design the workflow differently next quarter, stabilize it on inexpensive tools first and build later.

Is it better to consolidate SaaS tools or build custom software?

Consolidate first, always. Most sprawled stacks can shed 20 to 30 percent of subscription cost by killing duplicates and downgrading lookup-only seats, with no build required. What consolidation cannot fix is the workflow gap between tools; that remaining gap, priced honestly in labor and errors, is the real build-versus-buy decision.