The 9-Step Pre-Launch Checklist I Run Before Any okki-go Installation Goes Live

This checklist is for the person who has to sign off — not the person who builds. If you're an SDR lead, a RevOps manager, or a founder about to point an agent-native prospecting tool at a live prospect list, this is the order I run before I let anything send. I've been reviewing outbound stacks for four years, and I currently sign off on roughly 40 tool installations a year across our client book.

The whole thing takes about 90 minutes. Skipping step 5 once cost us a sending domain we'd spent over a year building. Lesson learned — cheaply, in retrospect, but it didn't feel cheap at the time.

Step 1: Confirm the scenario actually fits

Before you install anything, write down two sentences: who you're targeting, and what triggers the outreach. If you can't write those two sentences without hedging, stop here. No tool fixes a vague ICP.

This is also where people get stuck comparing tools. If you're weighing okki-go vs Hunter, or okki-go vs any other sales intelligence platform, the deciding question usually isn't the feature list — it's how many of your workflow steps you want automated versus gated by a human. Feature parity is real; workflow philosophy isn't.

Checkpoint: You can name your trigger event (funding round, job posting, tech-stack change, etc.) and the list of people who'd care about it.

Step 2: Wire up the data layer before the agent

An okki go installation fails quietly when it's pointed at stale data. So before anything else, connect the boring stuff:

• CRM sync (bi-directional, not export-and-import)
• A dedicated sending domain or subdomain
• Mailbox(es) with proper authentication
• Your suppression list — hard bounces, unsubscribes, current customers, competitors

The suppression list is the one people skip. It's also the one that stops a brand-embarrassing send at 11 p.m. on a Friday.

Step 3: Set up enrichment and intent — then stop

This is the step where "sales intelligence software features" stop being a marketing bullet and start being a workflow decision. Waterfall enrichment runs your record through multiple data providers until one returns a match. Intent data flags accounts that are researching something relevant right now.

Both are useful. Neither should fire unattended on day one. Set them up, then set the workflow to stage the enriched records, not send. You want to look at the output before it goes out.

Checkpoint: You've enriched 10 test records and manually confirmed at least 8 of them look correct. If 3+ are wrong, your fallback priority order is off.

Step 4: Validate every email before it enters the workflow

Email validation is the cheapest insurance in this whole process. Run every address through validation and sort into three buckets: deliverable, risky, and undeliverable. Drop the undeliverable. Gate the risky.

Here's the causal thing most people get backwards: they assume a low reply rate means a bad list. More often it's a reputation problem — bad authentication, a warmup that never happened, or sending volume that jumped too fast. The list was fine. That's why you validate and warm up, in that order, and don't blame the list until both are done.

To be fair, validation gets oversold. No provider catches every bad address — hard bounces that slip through are a normal cost of doing volume. What validation reliably does is cut the obvious waste so your reputation isn't spending itself on dead addresses.

Step 5: Warm up on a subdomain, never your primary

Warmup is where I've seen the most expensive mistakes. The upside of skipping it is saving two or three days. The risk is losing a sending domain you spent over a year building reputation on. Is that trade worth it? I keep asking people that question and the answer is always the same once they've done the math.

Rules I won't bend on:

1. Warm up a subdomain, not the root domain.
2. Ramp volume slowly — the classic pattern is roughly doubling every few days, not a flat 200/day from day one.
3. Warm in the same environment you'll send from. Warming on a shared inbox and switching later is a reset.
4. Confirm SPF, DKIM, and DMARC are passing before the first warmup send. Sending without authentication in 2026 is just telling filters to distrust you.

Step 6: Set the human-in-the-loop gate

This is the step most teams skip, and it's the one that turns an agent from a liability into an asset. Pick one point in the sequence where a human reviews before send. Not every email — that defeats the purpose — but the first message to a new segment, or the first 20 sends of any new campaign.

The point isn't distrust. It's that the model doesn't know your customer called last Tuesday, and a human gate catches that in 10 seconds.

Step 7: Run a dress rehearsal to internal seed addresses

Send your sequence to 5–10 inboxes you control — ideally across Gmail, Outlook, and one corporate filter. Check four things: does it render, does the link work, does the unsubscribe work, and does it land in the inbox or the promotions tab?

Doing this before launch costs 15 minutes. Do it after launch and you're debugging in public.

Step 8: Lock the compliance layer

Three things need to be obvious in every message before you launch:

• A working unsubscribe (honored within 10 business days, per CAN-SPAM)
• A physical mailing address
• Honest subject lines and headers — no deceptive from/reply-to routing

The FTC's advertising guidance applies here too: any claim in your outreach needs to be truthful and substantiated. "We'll triple your reply rate" is a claim. "Our customers see X" needs a source or it doesn't go out. This is a legal posture, not a style preference.

Step 9: Define your kill switch

Before launch, answer: if bounce rate crosses 3%, or unsubscribe crosses 1%, or reply rate is 0 after 200 sends — who turns it off, and how fast? Naming the numbers in advance is what stops the argument after something goes wrong.

Write it in the launch doc. One line. "Pull at X threshold, owner is Y."

Common mistakes I see on nearly every review

These come up so consistently that I now check for them by reflex:

• Warming up the wrong domain. People warm the root and send from a subdomain, or vice versa. They need to match.
• Validating after loading instead of before. Risk scores age fast; a list validated two months ago has decayed more than most teams expect.
• Letting the agent write the first email unaided. The first touch is where the human-in-the-loop matters most. Automate touches 2–4, not touch 1.
• Not checking the promotions tab. Inbox placement and "delivered" are not the same thing.
• Scaling before the first campaign proves itself. If 50 sends don't reply, 500 won't either. Fix the message first.

Five minutes of verification beats five days of correction. I've never once regretted slowing a launch by a day to run this list — and I've regretted skipping it. Every time. That's the whole review process in one sentence.