How to Choose a Systems Integration Partner
The proposal arrives on a Thursday. It is eleven pages, well designed, and confident. It promises to connect your accounting system, your CRM, and your scheduling tool in six weeks. It does not say what happens when one of those vendors changes its interface. It does not say who owns the code. It does not say what you would need to do if you wanted to leave.
Most bad integration engagements are not bad because the integrator was dishonest or unskilled. They are bad because the buyer never asked the questions that expose how the work would actually be run, and the proposal never volunteered the answers. This guide is for the person who signs. It covers what to ask, what to read in a scope document, how to check references, and which terms about ownership and exit you want in writing before work starts.
Decide what kind of partner you are hiring
Before you compare firms, be clear about the job. A systems integration partner connects systems you already own or buy, so that data and workflow move between them reliably. That is different from three other things people confuse it with.
- A reseller sells you licenses and may offer setup. Their incentive is the sale.
- A development shop builds custom software. Their incentive is the build.
- A managed service provider runs your technology day to day. Their incentive is steady operation.
An integrator can overlap with any of these, and that is fine. What you want to know is which incentive sits behind the advice. A reseller will tend to recommend the product it resells. A development shop will tend to recommend building. Ask directly, and listen for whether the answer is "it depends on your situation" followed by specifics.
Ask for discovery before a price
A trustworthy partner wants to understand your process before quoting. If the first call ends with a fixed price and you have shown nobody a single record, the price is a guess dressed as a number.
A good discovery phase should produce a short written document. It should list the systems involved, what data moves between them, in which direction, how often, and who owns each piece. It should name the unknowns. It should also state what the integrator needs from you, such as access, sample data, and a decision maker.
You can pay for discovery separately, and I recommend it for anything beyond a simple connection. It turns a vague engagement into something both sides can estimate, and it leaves you holding a useful document even if you choose a different partner afterward.
The questions that reveal how they work
These are the questions I would ask, grouped by what they expose.
How they handle failure
- What happens when a sync fails halfway through a batch?
- How will I find out, and how fast?
- Can the same message be delivered twice, and what stops a duplicate record from being created? A good answer mentions idempotency, which means that repeating an operation produces the same result as doing it once.
How they handle change
- When a vendor changes its interface, who notices and who pays for the fix?
- How do you track announced deprecations for the systems you connect?
- How are changes to the integration tested before they reach production?
How they handle security
- Where are credentials stored, and who on your team can read them?
- Do you ask for the narrowest permissions the job needs, or for administrator access?
- How do you verify that incoming webhooks are genuine? A strong answer mentions signature checking, where the sender signs each message with a shared secret and the receiver confirms the signature before trusting it.
How they handle people
- Who exactly will do the work, and who will be my contact?
- What happens if that person leaves your firm mid-project?
- How do you hand over knowledge to my staff?
How they handle money
- Which costs are yours and which are the vendors'? License and platform fees usually belong to the software vendors and are billed to you separately from the integrator's services.
- What is excluded from this estimate?
- How are changes priced once the work has started?
Notice that none of these questions is technical trivia. You do not need to understand the implementation to judge the answer. You are listening for specifics, for willingness to say "I don't know yet," and for whether the answer changes depending on your constraints.
Red flags worth taking seriously
No single item here disqualifies a firm, but several together should slow you down.
| Red flag | Why it deserves a second look |
|---|---|
| A fixed price with no discovery | The estimate cannot be grounded in your real systems |
| Reluctance to name who will do the work | The people you meet may not be the people who build |
| "We never have failures" | Every integration fails sometimes. Maturity shows in how failures are handled |
| Requests for administrator access to everything | Suggests they have not scoped the permissions needed |
| No written handover or documentation in the scope | You will depend on them indefinitely |
| Vague language about who owns the code | You may pay for something you cannot take with you |
| Pressure to sign quickly | Legitimate partners can wait a week |
| References that are all the same kind of business, or none at all | Ask why |
| A proprietary platform that only they can operate | A lock-in decision wearing a technical costume |
Also watch for what is missing. A proposal with no section on monitoring, no section on testing, and no section on handover is describing a build, not an operation. Integrations live for years after the build.
Read the scope document like a contract
The scope document, sometimes called a statement of work, is where the real agreement lives. The sales conversation is not the agreement. Read it for these items.
- Systems and versions. Each system is named, with the plan tier you hold and the interface that will be used.
- Data objects and direction. Which records move, which fields, and which direction. "Customers" is not specific enough. "Customer name, billing address, and tax status from the CRM to the accounting system, one way" is.
- Triggers and frequency. Real time, hourly, nightly.
- Error handling. What is retried, how many times, and where failures go for a human to review.
- Out of scope. An explicit list. This protects both sides.
- Acceptance criteria. How you will decide the work is done. Ideally these are tests you can run, using real examples.
- Deliverables. Code, configuration, documentation, a runbook, and training.
- Assumptions. Everything the price depends on, such as data quality or the availability of your staff for testing.
- Change process. How new requests are estimated and approved.
If an item is missing, ask for it in writing. Verbal promises evaporate the first time staff change.
Check references the way you would check a contractor
References are easy to fake and easy to waste. Ask for two or three clients with a similar size and a similar stack, and ask for one whose project did not go smoothly. A partner who says every project went perfectly is either inexperienced or not being candid.
When you call, ask questions that produce stories rather than ratings.
- What did they build, and is it still running today?
- Tell me about something that went wrong. How was it handled?
- Did the final cost differ from the first estimate, and why?
- When you needed a change a year later, what was that like?
- Would you hire them again for the same job?
I will add two cautions. First, do not assume a reference list means any particular client agreed to be named or relied on the work, and do not trust logos on a website without a conversation. Second, if the firm cannot offer any reference, ask what they can offer instead, such as a sample deliverable with client details removed, or a small paid pilot.
Settle ownership of code and data before you sign
This is the section most owners skip and later regret. Three things need clear answers in the contract.
Your data is yours. The agreement should state that all of your business data, including copies held in any system the integrator operates, belongs to you, and that you can get it back in a usable format on request. Ask what the format is and how long it takes.
You know who owns the code. There are several legitimate arrangements. You may own custom code outright. You may receive a license to use it. The integrator may retain ownership of reusable components and give you a license to the specific configuration built for you. Each is workable. What is not workable is silence. Ask for the intellectual property clause to be explicit, and have counsel read it. Ownership rules depend on jurisdiction and on the exact language, and this is general information, not legal advice.
You hold the keys. The accounts for your software vendors, your cloud hosting, and your credentials should be registered in your company's name, with your people as administrators. An integrator should operate inside your accounts with delegated access, not hold your accounts in theirs. If they host it in theirs, you need an explicit transfer plan.
Put exit terms in writing
You are not planning a breakup. You are making the relationship safe enough that either side can walk away without damage.
Look for these exit terms.
- Termination notice. A reasonable period, in both directions.
- Transition assistance. A defined number of hours or days of handover help after termination, with a stated rate.
- Documentation on demand. Architecture notes, a runbook, a list of credentials and where they live, and a list of every scheduled job.
- Return and deletion of data. Your data returned in a standard format, and deleted from their systems on your request, with confirmation.
- No penalty for leaving. Be cautious about long minimum terms combined with early termination fees.
- Source access. If you own the code, you have it. If you license it, understand what happens if the integrator closes or is acquired.
A clean test is this. Ask yourself whether a competent stranger could take over the work using only what the contract guarantees you. If the answer is no, the contract has a gap.
A short worked scenario
This scenario is hypothetical. Suppose you run a 20-person property services company and you need your scheduling tool to create invoices in your accounting system when jobs are completed.
You get two proposals. Firm A offers a fixed price, a four-week timeline, and a proprietary connector it hosts. Firm B proposes a paid discovery week, a written scope, a build inside your own accounts, and a short runbook.
Firm A looks cheaper and faster. But ask the five questions. What happens when an invoice is created twice? Where does it run? Who holds the credentials? What happens if you leave in two years? If the answers are vague, the low price probably reflects work that has been left out, such as monitoring and handover, or costs that arrive later in the form of a monthly hosting fee you cannot escape.
Firm B costs more at the start but leaves you with something you control. That does not mean Firm B is always right. A small, stable connection might be perfectly well served by a hosted connector, if the price is fair and the exit terms are real. The point is to make the trade-off visible before you sign.
A one page checklist
Before you sign, confirm the following.
- A discovery document exists and names the unknowns.
- The scope names systems, data objects, direction, frequency, and error handling.
- Acceptance tests are defined using your real examples.
- Named people are doing the work, with a backup.
- Accounts and credentials are in your name.
- Code and data ownership are stated explicitly and reviewed by counsel.
- Monitoring, alerting, and a runbook are deliverables.
- Pricing separates their services from vendor and license charges.
- Exit terms include transition help and data return.
- At least two references have been called, one of them about a difficult project.
What to do Monday morning
Write down, in one page, the three processes you want connected and why. Send it to two or three prospective partners and ask each of them to respond with their discovery questions rather than a price. The quality of those questions will tell you more than any proposal.
If it would help to have someone sanity check a proposal you have already received, we are glad to read it with you. We would rather you choose well, even if it is not us, than sign something with gaps you cannot see.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.