The demo always works. That's the point of a demo.
The sales engineer has a pristine sandbox, curated data, and a script that avoids every rough edge in the product. What you don't see is the rate-limited API, the export button that only spits out CSV, or the support queue that takes four days to answer a P1.
By the time you discover those, you've migrated your data, trained your team, and wired the tool into three other systems. Switching costs are a moat that works against you.
Here are the seven questions we ask on behalf of clients before signing anything. They cut through the polish fast.
1. What does your API actually let me do?
"We have an API" is meaningless. The real questions: Is every feature in the UI available via the API, or just a subset? What are the rate limits? Is there a webhook system, or do I have to poll?
Ask for the docs before you buy. Public, versioned API docs signal a company that treats integration as a first-class concern. A PDF sent by a sales rep signals the opposite.
A quick smoke test says a lot:
import requests
resp = requests.get(
"https://api.vendor.com/v1/records",
headers={"Authorization": f"Bearer {token}"},
params={"limit": 100, "updated_since": "2024-01-01"},
)
print(resp.status_code)
print(resp.headers.get("X-RateLimit-Remaining"))
print(len(resp.json().get("data", [])))
If you can't get a trial token to run something like this, that's your answer.
2. Who owns the data, and how do I get it out?
Every vendor promises you own your data. Fewer make it easy to leave with it.
Ask specifically:
- Can I export all my data, including relationships and metadata, not just flat tables?
- Is export self-service, or do I have to file a support ticket?
- What format? JSON with full fidelity, or a lossy CSV dump?
If the export path is painful, you're not a customer. You're a hostage.
3. How does this fit my existing technology stack?
No tool lives alone. It needs to talk to your CRM, your data warehouse, your auth provider, and probably an automation layer like n8n or Zapier.
Map the integration surface before you commit:
- Native integrations (maintained by the vendor)
- iPaaS connectors (Zapier, Make, n8n)
- Raw API + webhooks (you build it)
Native sounds best but ages badly if the vendor stops maintaining it. A clean API often beats a fragile native integration you can't debug. We usually prefer webhook-driven flows because they fail loudly instead of silently drifting.
4. What happens when something breaks at 2am?
Support tiers are where marketing meets reality. Get the SLA in writing:
- Response time for P1 vs P3
- Is 24/7 support real, or business hours in one timezone?
- Do I get a named contact, or a shared inbox?
- Is there a public status page with incident history?
Read the status page history. A vendor that publishes honest post-mortems is one you can trust in a crisis. A suspiciously green status page usually means they hide incidents.
5. How do you handle security and compliance?
Skip the logo wall of certifications and ask for artifacts:
- SOC 2 Type II report (not just "SOC 2 compliant")
- Data residency options
- SSO/SAML on your plan, or paywalled as an "enterprise" upsell
- Sub-processor list
The SSO tax is a red flag. If basic security controls are locked behind a 3x price jump, the vendor is optimizing revenue over your risk posture.
6. What does the true cost look like at scale?
The sticker price is the beginning of the conversation. The bill you actually pay depends on:
- Per-seat vs usage-based pricing, and how usage is metered
- Overage charges when you exceed limits
- Feature gating that forces a tier upgrade you didn't plan
- API call costs if the API is metered separately
Model it out for your projected volume in 18 months, not today. Tools that are cheap at 5 users can become brutal at 50.
const monthlyCost = (seats, usage) => {
const base = seats * 29;
const includedCalls = seats * 10000;
const overage = Math.max(0, usage - includedCalls) * 0.001;
return base + overage;
};
console.log(monthlyCost(50, 2_000_000)); // budget shock check
Run the numbers before the CFO does.
7. Is this company going to exist in three years?
You're not just buying software. You're betting on a company's roadmap.
Look for signals: funding stage, customer count, release velocity on the changelog, and whether the founders are still involved. A tool acquired by private equity often gets squeezed for margin, which means price hikes and stalled development.
Ask directly about the roadmap for the features you care about. Vague answers mean it's not happening.
Turn this into a scorecard
Don't hold these questions in your head. Build a simple weighted scorecard, rate each vendor 1-5 per question, and let the numbers argue.
The goal isn't to find a perfect vendor. It's to buy with your eyes open, knowing exactly which tradeoffs you're accepting.
The demo shows you what the product can do on its best day. These seven questions show you what it'll be like on a normal Tuesday, eighteen months in, when it's load-bearing in your stack. That's the version you're actually buying.
Originally published at getmichaelai.com
Top comments (0)