I usually get the call about six months too late, and the trouble almost always traces back to choosing a software development partner in a hurry.
A founder or an operations director reaches out because the app is half built, the budget is mostly spent, and the agency that promised a spring launch has gone strangely quiet. They want me to look at the code and tell them whether it can be saved. That’s often the classic rewrite vs. refactor legacy code question. Sometimes it can be saved. More often, the honest answer is that the problem started long before anyone wrote a line of code. It started in the sales meeting, when nobody asked the questions that would have exposed the mismatch.
As a fractional CTO, I sit on the buyer’s side of the table for companies that don’t have a full time technology executive. A big part of that work is choosing a software development partner, or cleaning up after the wrong one was chosen. I have sat through more vendor pitches than I can count. Every one of them has polished case studies. All of them say they are agile. Most of them have a slide with a handshake on it.
Why These Questions Matter When Choosing a Software Development Partner
What separates the good firms from the risky ones only shows up when you ask uncomfortable questions and watch closely how they answer. Not just what they say, but whether they hesitate, deflect, or pass you back to someone in sales.
The stakes are not small. Research by McKinsey and the University of Oxford found that large IT projects run 45 percent over budget and 7 percent over time on average, while delivering 56 percent less value than predicted. Software projects carried the highest risk of cost and schedule overruns, and 17 percent of IT projects went so badly they threatened the very existence of the company. The original Standish Group CHAOS study painted an even bleaker picture back in the nineties, with a success rate of only 16.2 percent and nearly a third of projects cancelled outright. The industry has improved since then, but the underlying pattern has never fully gone away.
Below are the seven questions I ask every vendor before I let a client sign anything. I’ve put them in the order I ask them, because the early answers change how I read the later ones.
Before Choosing a Software Development Partner, Get Your Own House in Order
Choosing a software development partner starts with your own homework. A vendor can only be as clear as the brief you hand them. Before the first call, I make my clients write three things down.
First, the business problem in one plain paragraph. Not the feature list. The problem. “Our field technicians lose two hours a day on paperwork” is a problem. “We need a mobile app with offline sync” is a solution someone already decided on, and it might be the wrong one. Not every problem justifies custom software development for small business in the first place.
Second, who on your side makes decisions. Every project I’ve seen stall had a moment where the development team needed an answer and nobody on the client side felt empowered to give one. Name the person.
Third, a realistic budget range. You don’t have to reveal it on the first call, but you need to know it yourself, or you will judge every proposal on price alone.
That last point matters more than most people realize. Daniel Haiem, who runs an app development firm, wrote in Fast Company that failed projects rarely come down to bad engineering. Instead, they trace back to misaligned expectations, vague contracts, and vendors chosen on price alone. He also points out that founders often overestimate what a vendor will handle on its own. You still need to supply clear requirements, review deliverables, and push back when work drifts. A good partner reduces that burden. No partner removes it.
Once that groundwork is done, here are the questions.
Question 1: Who, by Name, Will Actually Build This?
This is the first question because it exposes the oldest trick in the services business. The people in the pitch meeting are rarely the people who write your code. You meet a sharp principal engineer and a charming account director, you sign, and two weeks later your project is being run by a junior developer you’ve never heard of who is splitting time across four other clients.
So I ask for names. Then I ask follow up questions:
What is each person’s role and seniority? How long have they been with the company? What percentage of their week is committed to our project? Who is the technical lead, and can I speak with that person before we sign? If someone leaves halfway through, how quickly is a replacement onboarded, and who pays for the ramp up time?
Haiem makes the same point bluntly. Any good vendor should let you speak with the engineers before you commit, and deflection is a red flag.
What a Good Answer Sounds Like
A good answer sounds like this: “Your tech lead would be Priya, she’s been with us four years and led two projects similar to yours. She’d be at about 80 percent on your account. Here’s her calendar link.” A bad answer sounds like this: “We assign the team after kickoff based on availability.” That isn’t necessarily dishonest, but it tells me the team I’m evaluating doesn’t exist yet. I’m buying a promise, not a team.
One more detail I always ask about is contractors. Plenty of excellent firms use subcontractors, and that’s fine. What isn’t fine is finding out after the fact that your code is being written by a third company in a different country under a contract you’ve never seen. Ask directly, and get the answer in writing.
Question 2: What Would You Change About Our Plan?
This is my favorite question, and it catches people off guard. I hand the vendor our brief and ask them to tell me what’s wrong with it.
A vendor who agrees with everything is not being polite. They’re telling you they will build whatever you ask for, whether or not it makes sense, because that’s how they bill. That’s an order taker, not a partner. When I’m choosing a software development partner for a client, I want a team with enough experience and enough backbone to say “that feature will cost you six weeks and almost nobody will use it.”
DevSquad’s guide to vetting partners puts it well: a partner shouldn’t simply build whatever you ask for, and should have a proper discovery process in place. Stack Builders makes a similar case, noting that a strong partner digs into your strategy, competitive pressures, and user needs before any code gets written.
The best answers I’ve heard to this question include things like:
“Your first release has eleven features. I’d cut it to four and get it in front of real users in ten weeks.”
“An existing tool would cover 90 percent of that custom admin dashboard for a fraction of the cost.” That’s a build vs. buy software call, and a good partner will raise it unprompted.
“Nobody has mentioned how this connects to your accounting system. That integration is usually where the surprises live.”
If your project touches older systems, ask how the vendor has handled legacy application modernization before. Also ask whether they’ve helped clients migrate legacy Windows applications without downtime. Their answer will tell you whether they understand the old system or just the new one.
Why a Paid Discovery Phase Helps When Choosing a Software Development Partner
If the vendor proposes a paid discovery phase before committing to a full build, don’t treat that as a sales tactic. In my experience, a short, paid discovery engagement of two to four weeks is the single best investment a nontechnical buyer can make. It gives you a working sample of the relationship before you’re locked in. If the discovery work is sloppy, you walk away having spent a small amount instead of a large one.
Question 3: How Did You Arrive at That Number, and What Happens When It’s Wrong?
Every estimate in software is wrong. That isn’t cynicism, it’s the nature of building something that hasn’t existed before. The question isn’t whether the estimate is accurate. The real question is whether the vendor is honest about how uncertain it is and has a plan for when reality diverges.
I ask to see the breakdown behind the number. Not just a total, but the major pieces, the assumptions behind each one, and how much contingency they’ve built in. A credible estimate comes with a list of assumptions like “the client provides final designs by week three” or “the payment provider’s API works as documented.” Those assumptions are where the risk lives, and a vendor who lists them is a vendor who has done this before. If you’re new to this, learn how to read a software development cost estimate before the proposals arrive.
Atomic Object, a firm that has written thoughtfully about this from the vendor side, suggests buyers ask how the partner will determine whether the budget actually funds the project. I love that framing because it turns the conversation from “give me a number” into “help me understand whether this is realistic.”
Ask About the Last Project That Ran Over
Then I ask the harder follow up: “Tell me about the last project that went over budget. What happened, and what did you do?” Every honest firm has an answer. A firm claiming it has never gone over is either very new or not telling the truth.
Pricing Models to Weigh When Choosing a Software Development Partner
On pricing models, here’s my practical view. Fixed price contracts feel safe, but vendors protect themselves by padding the estimate and fighting every change request. Time and materials contracts are more honest about uncertainty but require you to stay engaged and watch the burn rate. For most of my clients, I prefer time and materials with a cap per phase, short phases, and a clear review point at the end of each one.
Short phases matter for a reason that’s backed by data. The McKinsey and Oxford research found that the longer a project is scheduled to last, the more likely it is to run over, with every additional year adding roughly 15 percent to cost overruns. Break big projects into small, shippable pieces, and the math works in your favor.
It’s also worth asking whether the vendor will tie any payments to outcomes rather than hours. That approach is becoming more common. In Deloitte’s most recent global outsourcing study, 67 percent of respondents said they are adopting outcome based outsourcing relationships. Not every project suits it, but a vendor who is willing to discuss it is a vendor who believes in their own delivery.
Question 4: Who Owns the Code, the Accounts, and the Keys From Day One?
This one sounds like a legal formality. It is not. Of everything involved in choosing a software development partner, this is the question that has saved my clients the most money and the most grief.
I once stepped into a situation where a company had paid a development shop for nearly two years of work. When the relationship soured, they discovered that the code repository sat under the agency’s account, the cloud hosting was billed to the agency’s credit card, and the app store listing was registered to the agency’s developer profile. Technically the client had paid for everything. Practically, they owned nothing they could touch. Untangling it took months and a lawyer.
The Ownership Checklist I Use When Choosing a Software Development Partner
So I’m specific. I want written confirmation of all of the following before the first sprint begins:
- The code repository lives in the client’s organization account, with the vendor added as collaborators.
- Cloud infrastructure accounts are opened in the client’s name and billed to the client.
- Domain names, app store developer accounts, and analytics accounts are registered to the client.
- Every third party service, from email delivery to payment processing, is set up under the client’s ownership.
- Admin credentials are stored in a password manager the client controls.
- The contract includes a clear assignment of intellectual property to the client upon payment.
Haiem flags the same issue, noting that the contract should clearly spell out code and repository ownership, because unclear terms can cause legal trouble later, especially during a fundraise or exit. Investors and acquirers will ask about this during due diligence, and “the agency has it” is not an answer they want to hear.
How a Good Vendor Responds
A good vendor won’t push back on any of this. In fact, the good ones set it up this way by default, because it signals trust and makes the relationship cleaner for both sides. If a vendor resists putting your code in your account, find out why. The answer will tell you a lot. And of course, have your own attorney review the contract terms. I’m a technology advisor, not a lawyer, and intellectual property language deserves a proper legal read.
Question 5: Show Me What “Done” Looks Like on Your Team
Every vendor says they write clean, tested code. Very few can show you what that means in practice. So I stop asking them to describe their process and start asking them to show it.
Specifically, I want to know their definition of done. When a developer says a feature is finished, what has actually happened? Has another engineer reviewed the code? Are there automated tests, and roughly how much of the code do they cover? Has the feature been deployed to a staging environment where I can click through it myself? Has someone outside the development team checked it against the original requirement? And before launch, will they welcome an independent application security assessment, or treat it as an insult?
Visibility Into Real Progress
I also ask about visibility. Will I have read access to the code repository and the task board from day one? How often will I see working software? My rule is simple: if I can’t see a live demo of real progress at least every two weeks, I assume nothing is progressing. Slide decks and status emails are not software.
Where AI Tools Fit When Choosing a Software Development Partner
Then there’s the question that barely existed a few years ago: how does your team use AI tools? This is no longer hypothetical. Deloitte found that 83 percent of surveyed executives are already using AI as part of their outsourced services. Yet the same research notes that real gains in productivity or cost have been limited, largely because of weak governance and contracting around AI.
I have no problem with developers using AI coding assistants. Used well, they speed up routine work. But I want to know three things. Who reviews code suggested by those tools before it’s merged? Is any of my proprietary code or customer data being sent to outside AI services, and under what terms? And if the tools make the team faster, does that efficiency show up in my invoice, or only in the vendor’s margin? A thoughtful vendor will have clear answers. A vague vendor will say “we use the latest tools” and change the subject.
Question 6: When Did You Last Deliver Bad News to a Client, and How?
Every software project hits trouble. A key library turns out to be unmaintained, one of the quieter end-of-support software risks. An integration partner changes their API without notice. A developer quits. Sometimes the estimate was simply wrong. None of that worries me. What worries me is a vendor who hides it.
The most expensive projects I’ve cleaned up weren’t the ones with the biggest technical problems. They were the ones where the vendor knew about a problem for weeks and kept reporting “on track” because they hoped to fix it quietly. By the time the client found out, the options had narrowed from “adjust the scope” to “start over.”
What a Strong Answer Includes
So I ask for a story. “Tell me about the last time you had to tell a client something they didn’t want to hear. What was it, when did you tell them, and how did it play out?” I’m listening for specifics, timing, and ownership. A strong answer sounds like: “We realized in week five that the reporting module would take twice as long as planned. We flagged it in the next weekly call, gave the client three options with costs attached, and they chose to push it to phase two.”
DesignRush’s guide on vetting agencies lists the vendor’s approach to risk management and escalation among its core questions, and I agree that escalation paths belong in the conversation early. Find out who you call when the project manager isn’t responding. Check whether you’ll have a direct line to someone senior at the firm. It also helps to request a sample weekly status report and see whether it ever shows anything other than green.
If they can’t tell you a single story about delivering bad news, they’ve either never had a hard project or they’ve never been honest about one. Neither is reassuring.
Question 7: If We Part Ways in a Year, What Do I Walk Away With?
Nobody likes to talk about the breakup during the first date. But in software, every engagement ends eventually. Your product matures, you hire your own engineering team, the budget shifts, or the relationship simply runs its course. The way a vendor answers this question tells you whether they’re building your capability or building your dependence. It’s also the most overlooked part of choosing a software development partner.
This happens more than people think. Deloitte’s research found that 70 percent of executives have brought work back in house from a third party at some point over the last five years. Planning for transition isn’t pessimism. It’s realism.
Planning the Exit When Choosing a Software Development Partner
I ask what documentation they produce as a matter of course. Is there a written architecture overview? Are deployment steps documented so someone new could release the software without calling the vendor? Do they write down the reasons behind major technical decisions? I also ask about the handover itself. How much notice does either side need to give? Will they commit to a transition period where their engineers pair with my incoming team? What does that cost?
DesignRush’s author, who has managed and audited dozens of agencies, says he has watched companies double their costs because the agency walked away with the knowledge. I’ve seen the same thing. The code was technically handed over, but nobody could run it, nobody understood why it was built the way it was, and the new team spent their first three months reverse engineering instead of building. Undocumented code like that turns into a legacy problem almost overnight, and securing legacy applications is much harder when nobody understands them.
A good partner wants you to be able to leave. That sounds counterintuitive, but the firms confident in their work don’t need to trap clients. They keep them by being worth keeping.
How I Score the Answers When Choosing a Software Development Partner
After the meetings, I don’t just go on gut feel. Each vendor gets scored on four things.
Specificity. Did they give real names, real numbers, real stories? Or did they speak in generalities? Specific answers are harder to fake.
Consistency. Did the salesperson and the engineer tell me the same story? I make a point of asking the same question to different people at the firm. When the answers don’t line up, I pay attention.
Willingness to say no. Did they push back on anything in our brief? Were they willing to turn down any part of the work as outside their strengths? A vendor who can do everything usually does nothing particularly well.
References that hold up. I always call at least two past clients, and I ask questions the vendor didn’t prepare them for. “What went wrong on your project, and how did they handle it?” “Would you hire them again for something more complex?” “Is there anything you wish you’d negotiated differently?” DevSquad is right that case studies aren’t enough. A five minute phone call with a real former client is worth more than a fifty page capabilities deck.
Red Flags to Watch for When Choosing a Software Development Partner
A few answers make me stop the evaluation right there, no matter how strong the rest of the pitch is.
Some vendors won’t let me meet the developers before signing. Others insist the repository and hosting stay under their own accounts. A firm fixed price after one thirty minute call, with no questions about integrations, users, or scale, is another warning sign. So is a firm that can’t name a single project that went wrong. Pressure to sign quickly because a discount expires on Friday ends the conversation too. Finally, I walk away from anyone who answers every technical question with “we use whatever is best for the client” without ever naming what that actually is.
None of these automatically means a firm is dishonest. But each one tells me the risk is sitting on my client’s side of the table, and my job is to move it back to the middle.
Final Thoughts on Choosing a Software Development Partner
Choosing a software development partner is less about finding the most impressive firm and more about finding the one whose habits match yours. You want a team that tells you the truth early, shows you real progress often, leaves your ownership beyond doubt, and makes it easy for you to stand on your own when the time comes.
The seven questions above won’t guarantee success. Nothing does in software. But in my experience, they filter out most of the vendors who would have led to that phone call six months too late. Take your time with the answers when choosing a software development partner. Ask the follow ups. Call the references. The few extra weeks you spend choosing carefully are almost always cheaper than the months you’d spend recovering from the wrong choice.
And if you don’t have someone senior and technical on your side of the table, consider bringing in that expertise before choosing a software development partner, even part time. Haiem suggests that companies without a technical cofounder or CTO should consider hiring part time senior technical leadership first. I may be biased, but I think he’s right.
Frequently Asked Questions About Choosing a Software Development Partner
How long should choosing a software development partner take?
For most small and midsized projects, plan on three to six weeks from first conversation to signed contract. That allows time for initial calls, written proposals, reference checks, and ideally a short paid discovery phase. Rushing this step is one of the most common reasons projects fail, as the McKinsey research on large IT projects makes clear.
Should I choose a fixed price or time and materials contract?
Fixed price works when the scope is small, well defined, and unlikely to change. For anything more complex, time and materials with a spending cap per phase usually gives you more flexibility and more honest estimates. Atomic Object’s guide to engagement management offers useful questions for testing whether a budget is realistic.
What should a software development contract include about ownership?
At minimum, it should assign intellectual property to you upon payment and confirm that the code repository, cloud accounts, domains, and app store accounts are held in your name. Fast Company’s guidance for founders highlights code ownership as one of the most important terms to settle before signing. Always have an attorney review the final language.
Is it a red flag if a vendor uses AI coding tools?
Not on its own. Most firms now use them in some form, according to Deloitte’s Global Outsourcing Survey. What matters is how they’re governed: who reviews the output, whether your code or data is shared with outside services, and whether any efficiency gains are reflected in your pricing.
How many vendors should I evaluate?
Three is usually the right number. Fewer than that and you have no basis for comparison. More than five and the process becomes exhausting, and the quality of your evaluation drops. DesignRush’s agency vetting guide offers a helpful framework for narrowing a long list.
Do I need a fractional CTO before choosing a software development partner?
If nobody on your team can read a technical proposal critically, judge an estimate, or spot a weak architecture decision, then yes, it’s worth bringing in senior technical guidance first. Even a few hours of independent review during vendor selection can prevent costly mistakes. Stack Builders’ partner selection guide explains why business alignment matters as much as technical skill.
References
- Bloch, M., Blumberg, S., and Laartz, J. Delivering large scale IT projects on time, on budget, and on value. McKinsey & Company.
- The Standish Group. The CHAOS Report. Academic reprint, University of Oslo.
- Aneja, K., Coronado, J., and Stoler, M. Global Outsourcing Survey 2024: Multidimensional Sourcing. Deloitte.
- Haiem, D. 4 Questions to Ask an App Developer Before You Sign. Fast Company.
- Oliveira, S. 10 Questions to Ask a Software Development Agency Before Hiring. DesignRush.
- Atomic Object. Choosing the Right Software Partner: The Engagement Management Process. Atomic Spin.
- DevSquad. How to Choose a Software Development Partner. DevSquad Blog.
- DevSquad. 16 Questions to Ask a Potential Software Development Firm. DevSquad Blog.
- Stack Builders. 10 Questions to Consider When Choosing a Custom Software Development Partner. Stack Builders Insights.
- de Villiers, M. How to Kill a Black Swan. ITWeb.