Heartwood

Guide

How to Choose a Software Development Agency

Last updated

Choosing a software development agency comes down to four questions: do they understand your business, can they tell you the truth about scope, will their work survive after launch, and do you actually want to work with them for the next 12 months? Most agencies pitch process and pricing. The right partner shows you their judgment. This guide walks through the eight-step evaluation process we'd use ourselves, including the red flags that predict failure, the questions that predict fit, and the pricing models that predict pain.

The data on software project outcomes is brutal. Only 31% of software projects succeed; 50% are challenged (over budget, late, or missing scope); 19% fail outright[1]. Two-thirds of large-scale tech programs miss their targets[2]. Average IT projects over $15M run 45% over budget, 7% over time, and deliver 56% less value than predicted[3]. Roughly 70% of all software projects exceed their initial budget, with average overrun of 27%[4].

A non-trivial share of those failures trace back to one decision: which partner the buyer chose. The good news is that the agency-selection signal is mostly hidden in plain sight. Once you know what to look for, and what to walk away from, the choice gets clearer.

This guide is the framework we wish every non-technical founder, COO, and product lead had before they signed their first SOW. We didn’t invent it. We learned it from being on the buyer’s side of bad agency relationships before we built Heartwood. Use it whether you hire us or not.

When you actually need an external partner

You probably need a custom software partner if:

  • Your business runs on a workflow no SaaS tool fully covers, and you’re stitching three or more tools together to compensate.
  • You have an idea or product spec, no in-house engineering leadership, and your runway makes a lengthy developer hiring cycle feel painful.
  • You have an in-house team but no senior engineer or engineering leader to architect what’s coming.
  • You’ve outgrown your current platform and the rebuild feels too big for your current team.

You probably don’t need one if you’re early-stage, your product fits an off-the-shelf tool, and the work could be done by a single contractor in a quarter.

Three types of agencies (and which one fits you)

Most software shops fall into one of three structures. The differences matter more than the websites suggest.

Staff augmentation / dev shop

You manage the project; they supply the engineers. Cheaper hourly rate, but you own scope, process, and outcomes. Good if you have engineering leadership in-house and just need hands. Bad if you’re a non-technical founder.

Heartwood does this too, as an Embedded Team: engineers who come with review, coaching, and a bench behind them, so a non-technical team isn’t left managing them alone.

Boutique studio

A small senior team that owns projects end to end. Higher hourly cost, but they bring product judgment, design, and engineering together. Good if you need a partner, not just builders. Heartwood works both ways: as a boutique studio for Custom Builds, and as an Embedded Team when you need people on your team. We work with mission-driven nonprofits, faith-rooted organizations, and scaleups who need senior engineering judgment without enterprise overhead.

Full-service / enterprise

Hundreds or thousands of staff, multiple delivery centers, layered project management. Good for large, complex, multi-quarter programs. Bad if you’re under $500K, you’ll pay for overhead you don’t need.

TypeHourly rate (US)[5]Best forRisk
Staff augmentation$50–$120Companies with in-house tech leadershipYou own all the risk
Boutique studio$120–$250Founders, scaleups, mission-driven orgsCapacity ceiling
Enterprise$250–$900+Fortune 1000, regulated industriesCost, slow delivery
Offshore$25–$82Cost-sensitive, well-defined projectsCultural fit failure rate

Heartwood doesn’t bill hourly. Every Starter has one fixed price.

The 8-step evaluation process

1. Define the problem before the solution

Most agencies will happily build whatever you describe. The good ones make sure you’ve described the right thing. Walk into evaluation knowing the business outcome you’re buying: revenue, retention, efficiency, compliance, not the feature set. The agency’s first question should be about that outcome.

2. Look at their work, not their brand

Pull two or three case studies. Ask what was hard, what they got wrong, what they’d do differently. An agency that can’t tell you about a project that went sideways doesn’t have enough scar tissue to be useful. The interesting case studies aren’t the wins; they’re the recoveries.

3. Run a paid discovery

Free pitches incentivize fast, generic proposals. A paid discovery, even a $2,500 one-week engagement (ours is the Idea Check), produces a real plan you can compare across vendors and use as a hiring artifact even if you walk away. This is the single highest-leverage move in the entire process. The agencies that resist it are usually the ones you don’t want.

4. Pressure-test their judgment

Ask: “What should we not build?” Ask: “Where is our scope unrealistic?” Ask: “What would you do with this budget if you were us?” An agency that gives you back a smaller, sharper plan than the one you walked in with is showing you their judgment in real time. An agency that just nods is selling you the project they want, not the project you need.

5. Investigate ownership and handoff

Who owns the code? (You.) Who owns the IP? (You.) What does documentation look like at the end? (Real, written, in your repo.) If the answers are vague, the lock-in is the business model.

6. Verify the team you’ll actually work with

Ask which engineers, designers, and PMs will work on your project. Get names. Ask whether any of the work will be subcontracted, and to whom. The bait-and-switch (senior engineer in the pitch, junior engineer in the build) is the #1 staffing complaint across software-buyer surveys.

7. Check references, but ask the right questions

“Were you happy?” gets you nothing. Ask: What surprised you? What did they push back on? Did they ever tell you no? Did they stay after launch? Would you hire them again knowing what you know now? The most useful reference is the one that ended.

8. Test the working relationship

Run a paid pilot, a fixed-price engagement, a discovery week, a small first milestone. You will learn more in a paid two-week pilot than in three rounds of sales calls. This is also when you find out whether they communicate well at the speed your business needs.

Ours is the Team Trial: two weeks, $5,000, one of your tickets taken to a merged PR, and a code review your team sits in on.

Red flags

Eight signals that predict trouble:

  • Fixed-price quote for a whole build before any discovery. A fixed price for a short, defined piece of work is a good sign; a fixed price for a system that doesn’t yet exist on paper means they’re either underestimating (and you’ll pay in change orders) or padding (and you’ll overpay).
  • No discovery process or a discovery that’s actually a sales pitch in disguise.
  • No questions about your business, only questions about your features.
  • Subcontracting they don’t disclose.
  • A team you’ll never meet until kickoff.
  • Lack of senior engineering leadership in the room. If a salesperson is the most technical person you’ve talked to, walk.
  • No conversation about what happens after launch.
  • Pressure tactics, urgency framing, or “expiring” pricing on a six-figure decision.

How to compare pricing models

ModelWhat it isBest forWatch out for
Fixed priceSet scope, set costTight, well-defined projectsChange orders eat the savings
Time and materialsYou pay for hours workedProjects with evolving scopeNo incentive to ship fast
Dedicated teamMonthly retainer for a fixed teamLong-term partnershipsCapacity locked in
Outcome-basedPayment tied to business resultsMature buyers, mature vendorsHard to define metrics

The 2024 Deloitte Global Outsourcing Survey found cost has fallen from the #1 outsourcing driver (70% in 2020) to just 34% in 2024, replaced by access to specialized talent (42%) and meeting customer demand (35%)[6]. The smartest buyers in 2026 select for fit and durability, not price.

Ten questions to ask in your sales call

  1. What kinds of projects do you say no to, and why?
  2. Walk me through a project that went badly. What did you learn?
  3. Who specifically will work on this project, and at what staffing level?
  4. What’s the smallest version of this we could ship, and why might that be the right answer?
  5. What does discovery cost, and what do I walk away with?
  6. Who owns the code, IP, and documentation at the end?
  7. What happens after launch, what does ongoing support look like?
  8. How do you handle scope changes mid-project?
  9. Can I talk to a client whose project ended without a renewal?
  10. What would you do with our budget if you were us?

The answers will tell you more than any pitch deck.

Why durability matters more than speed

Hiring the wrong agency creates technical debt that follows you for years. Companies estimate technical debt at 20–40% of their entire technology estate, and 86% of IT executives report tech debt impacted their company in the past year[7]. Per Stripe’s Developer Coefficient, 42% of professional developer time goes to managing technical debt[7]. Bad code today is bad code your future team is paying for.

This is why partnership posture beats project posture. The 2024 Deloitte survey found 40% of firms are expanding outsourcing contracts and shifting to outcome-based models: buyers are voting with renewals, not RFPs[6]. The agencies worth working with are usually the ones still answering your texts a year after launch.

Where Heartwood fits in

We’re a boutique studio. We’re partial to long relationships, paid discoveries, and small first builds that prove the path forward. We say no to a lot, not because we don’t want to work hard, but because we want the right work. We’d rather tell you the hard truth than sell you something you don’t need.

We have a name for this: Irrational Care. We care more than makes sense on paper, because the best work happens when someone is invested in the outcome rather than contracted to it.

If software can help you set your corner of the world right, we’d love to hear about it. See how we partner, or read the founder’s letter.

If you want to test that, start with a Starter: one to four weeks, one fixed price, and worth having even if you never hire us again. We’d rather earn the next phase than assume it.

Something new to build
Idea Check · $2,500 · about one week
You want to see us work before you commit
Team Trial · $5,000 · two weeks
Code you inherited, and don’t trust
Code Audit · $2,500 · two to three weeks

After a Starter, Custom Builds start at $20,000, and an Embedded Team is $6,500/month per quarter lane, 3-month minimum.

Find the Starter that fits your problem

Frequently Asked Questions

How much does a custom software project cost in 2026?

Custom software projects on Clutch typically start around $25,000 and exceed $500,000 for complex platforms, with most software shops charging $25–$49/hour at the median price band, $120–$250/hour for boutique studios, and $250–$900+ for enterprise firms[5]. The right number depends on scope, complexity, and team mix. A focused MVP usually lands $30K–$150K; a full custom platform $150K–$1M+. Anyone giving you a number before discovery is guessing. An Idea Check is $2,500 and about one week, and ends in a real one.

How long does it take to build custom software?

A focused MVP typically ships in 8–12 weeks; a moderate custom platform in 4–9 months; an enterprise system in 12–24+ months. Each year of project duration adds roughly 15% to cost overruns, per the McKinsey-Oxford analysis of 5,400+ IT projects[3]. The shortest viable scope wins. Iterate from there.

What’s the difference between staff augmentation and a full-service agency?

Staff augmentation gives you engineers who work under your management, cheaper hourly, but you own outcomes. A full-service or boutique agency owns the project end to end, including discovery, design, engineering, and delivery, higher hourly, but you’re buying judgment, not just hours. If you don’t have senior engineering leadership in-house, staff aug usually fails. If you do, an agency may be overkill.

How do I avoid getting burned by a software agency?

Run a paid discovery before any full engagement. Get the names of the people doing the work. Ask for references whose projects ended without a renewal. Verify code and IP ownership in writing. Watch for a fixed-price quote on a whole build before any discovery: it’s a leading indicator of either underestimation or padding. And check whether the agency is willing to tell you what not to build. The ones who can are the ones worth hiring.

Should I hire onshore, nearshore, or offshore?

Onshore (US, UK, Australia) costs $100–$250+/hour and offers the lowest communication friction. Nearshore (Latin America, Eastern Europe) runs $30–$100/hour with strong time-zone overlap. Offshore (South Asia, Philippines) runs $20–$45/hour but cultural and time-zone misalignment is a top-five reason offshore projects fail[8]. The right answer depends on project complexity, your in-house management capacity, and how much ambiguity the work involves. Highly defined projects can offshore well; ambiguous, fast-moving products usually shouldn’t.

Sources

  1. Standish Group CHAOS Report 2020, via ACM Queue
  2. Boston Consulting Group, "Most Large-Scale Tech Programs Fail" (2024)
  3. McKinsey & University of Oxford, "Delivering large-scale IT projects on time, on budget, and on value."
  4. Forecast.app analysis citing McKinsey & Oxford
  5. Clutch Software Development Pricing Guide (April 2026)
  6. Deloitte 2024 Global Outsourcing Survey
  7. McKinsey Digital, "Tech debt: Reclaiming tech equity."
  8. SQ Magazine 2026 outsourcing statistics

Start here

What are you hiring an agency for?

We’ll point you to the Starter that fits.

Which is closest? Check any.

We reply within one business day. Then a 30-minute call. If it fits, you get a fixed-price agreement.

or email us at hello@heartwood.agency

Or find your Starter in the problem grid