Your First 30/60/90 Days as a New Growth Marketer

The first time I walked into a Segment project nobody owned, I spent a full afternoon trying to figure out why the signup_completed event fired twice for some users and not at all for others. The previous growth lead had left four months earlier. There were three competing definitions of "activated" written down in three different Notion pages, and a Mixpanel project where half the dashboards pointed at deprecated event names. My 1:1 with the CEO was the next morning. He wanted signups doubled by Q4.

If you just started a growth role and the welcome doc looks suspiciously thin, this is the version of the job nobody told you about. The clean dashboard you imagined doesn't exist. The hypothesis backlog isn't there. The friendly data analyst on Slack is shared across four PMs and has 90 hours of queued work. And the CEO already promised the board a number. Here's how to not panic for 90 days.

Why most growth hires stall in the first 90 days

Three failure modes show up almost every time.

The first is shipping experiments before the instrumentation is trustworthy. Week three, you push an onboarding tweak. Week six, the lift looks great. Week eight, someone from data spots that the activation event fires before the user actually completes the activation step. Your win evaporates and so does your credibility.

The second is over-promising in week one. You want to look ambitious, so you tell your manager you'll get activation up 20% by end of Q3. Now every conversation for the next three months runs against that number. You stop investigating problems and start defending a forecast you made before you understood the funnel.

The third, and the one that quietly kills more new growth hires than the other two combined, is treating the CEO's "double signups" as the actual goal. It isn't. It's a request you have to translate into the activation step that's actually broken, the channel that's actually scalable, and the lever you can actually move in 12 weeks. New hires who skip that translation step end up running paid acquisition tests on a leaky bucket and wondering why nothing compounds.

The 30/60/90 below is built around avoiding all three.

Days 1–30: audit, don't ship

Your job in the first month is to earn the right to ship. Nothing more.

Week 1: funnel data audit

Open Segment first. Pull up the source debugger and watch traffic for 30 minutes. Then open Mixpanel or Amplitude and list every event firing on signup, activation, and conversion. Cross-reference with the warehouse if you have one (Snowflake, BigQuery, Postgres replica, whatever). You're looking for three things:

Events that fire when they shouldn't. Duplicate signup_completed events from a client-side and a server-side source. subscription_started firing on the trial start instead of paid conversion. The classics.
Events that don't fire when they should. The "key action" event missing on mobile web. Activation events that depend on a user property that's null for 40% of users.
Properties that drift. plan_type written as "free" in some events and "Free" in others. referrer populated for some signups, missing for others.

You will find at least two instrumentation gaps. I have never seen a growth role where the new hire didn't find at least two. Document them in a single page titled "Funnel data I can't trust yet" and date it. That document is the first artifact your manager will see from you, and it should be specific.

Week 2: identify the activation gap

Now pull the funnel. Visit → signup → key action → repeat use within 7 days. Pick the step with the steepest drop you can actually move.

That qualifier matters. The leakiest step in most B2B SaaS funnels is visit → signup, and most of that leak is solved by paid acquisition spend, brand, and SEO over 12 months. You can't move it meaningfully in your first quarter. Skip it. Look for the activation step instead — the moment a signed-up user does the thing that predicts they'll come back.

If the drop from signup to first key action is 70%, that's your wedge. You don't need 100% of users to hit the key action. You need to find the 20% of users who would have hit it if one obvious obstacle weren't in the way.

Week 3: talk to ten users

Five who activated, five who didn't. Yes, even without a research team. Yes, 30-minute calls. Yes, you have to schedule them yourself.

I know this feels like the part you can skip because you're a "data person." Don't skip it. Five conversations with users who churned at the activation step will tell you more about what to ship in days 31–60 than any dashboard. The first time someone tells you they bounced because the empty state showed three sample projects with names like "Q4 marketing plan" and they assumed the product was for marketers when they were a developer, you'll understand why.

Use Calendly. Offer a $25 Amazon gift card. Most users will say yes if you keep the call to 30 minutes and tell them you're new and trying to understand the product.

Week 4: pick ONE high-leverage fix

By the end of week 4, you have one hypothesis. Just one. Written like this:

Hypothesis: If we replace the empty state on /dashboard for first-time users with a single "Create your first lead source" CTA (instead of three sample projects), activation rate at day 7 will improve from 18% to 25% (+7pp absolute, +39% relative), because 4 of 5 churned users in our research mentioned confusion about where to start.

Surface area: New signups, web app, day 0–7. Instrumentation needed: empty_state_cta_clicked event (currently missing), variant assignment via existing experimentation tool. Expected sample to detect: ~3,200 users per arm at 80% power (current weekly signups: 2,400, so ~3 weeks). Risk: None to existing users; new signups only.

Get sign-off from your manager and the CEO. Tell the CEO you spent month one auditing because you wanted defendable wins, not week-three wins that fall apart at month three. Most reasonable CEOs respect this. The unreasonable ones tell you a lot about what working there will be like.

Days 31–60: ship 2 experiments, build the backlog, set the cadence

Month two is when you start producing visible output. Two experiments, a real backlog, and a weekly readout the rest of the company can actually rely on.

Ship the week-4 hypothesis as experiment #1

Don't wait for instrumentation to be perfect. It won't be. Ship what you wrote in week 4 with the cleanest read you can get. If your experimentation tool has issues, use a feature flag with a 50/50 split and read the result from the warehouse. If you don't have a warehouse, use Mixpanel's experiment reporting and double-check the math by hand.

The point of experiment #1 is not the result. It's proving you can ship a measurable change with a clean read. That's the muscle that compounds.

Run experiment #2 in parallel on a smaller surface

Pick something low-risk: an onboarding email subject line, a pricing page hero, a confirmation modal copy change. Run it in parallel. Two experiments running simultaneously gives you a second shot at learning if experiment #1 produces a null result. It also signals to the org that growth has shipped throughput, not just "this one big thing we're waiting on."

Build a hypothesis backlog of 15–20 ideas

By end of month two, you should have a real backlog. Each entry needs:

Hypothesis (one sentence, "if X then Y because Z")
Surface area (where it lives in the product/funnel)
Expected lift (your honest guess in absolute pp and relative %)
Instrumentation needed (events that exist vs. events to add)
ICE score (impact × confidence × ease, 1–10 each)

Sort by ICE descending. The top 5 become your Q3 roadmap. The bottom 15 become Q4 raw material. When someone asks "what are you working on next?" you point at the backlog instead of inventing an answer.

Set up a weekly experiment readout

Fifteen minutes, every Friday. Three slides.

Slide 1 — Shipped this week
  - Experiment name + surface
  - Hypothesis (one line)
  - Result + p-value + sample
  - Decision: ship to 100% / kill / iterate

Slide 2 — Learned this week
  - One paragraph on what the result means
  - What it implies for the next experiment

Slide 3 — Shipping next week
  - Top 3 from the backlog
  - Why these 3 (link to ICE ranking)

Invite the CEO. Make it boring and predictable. The first three weeks the CEO will probably skip. The fourth week, when they realize the readout always happens at the same time and always says specific things, they'll start showing up. That's when you've earned the credibility to ask for resources later.

Days 61–90: own a metric, present, propose H2

Month three is when you stop being "the new growth hire" and start being the person who owns the funnel.

Pick your North Star

Not signups. Signups is a vanity metric and you already know it. Pick the activation or retention number that actually predicts revenue. For a PLG product, that's usually a weekly retention metric (W1 retention, W4 retention, "users who hit X events in week 1"). For a sales-led B2B product, it's qualified-lead-to-paid conversion or trial-to-paid by cohort.

Defend the choice with data. Pull the cohort analysis: of users who hit metric A in week 1, what % are still around at month 3 vs. users who didn't? If the gap is wide and stable, that's your North Star. Show the cohort curve to your manager. Get them to agree before you put it in front of the CEO.

Present the 90-day report

One deck. Honest. The structure that works:

What I found in the audit. The instrumentation gaps. The funnel as it actually is, not as the previous deck claimed.
What I shipped. Two experiments, results, learnings (including null results, never hide them).
The hypothesis backlog. Top 5 with ICE scores.
Recommended North Star. With the cohort data behind it.
H2 roadmap. Top 5 hypotheses, expected impact in absolute terms, what you need from eng/design/data to ship them.
The ask. One thing. See below.

Make one ask

A dedicated analyst. An experimentation tool budget. Eight hours a week from a backend engineer. A research panel subscription. Whatever unblocks the next quarter. New hires who don't ask in the 90-day report rarely get resources later. The 90-day mark is the moment you have the most leverage you'll have all year, because you've shipped, you have data, and the CEO hasn't yet absorbed you into the existing org budget.

Ask for the smallest specific thing that unblocks the biggest specific outcome. "10 hours/week of analyst time will let me run 4 experiments per quarter instead of 2, which on the current backlog ICE rankings should produce ~$140K in incremental ARR over H2." That ask gets approved. "I need more resources" doesn't.

The "experiment but no learning" trap

You can ship 12 experiments in a quarter and learn nothing. It happens constantly. The pattern looks like this:

Underpowered samples. You read a result at 40% of required sample, see a 5% lift, ship it. Six weeks later you can't tell if the lift was real.
No clean control. You changed the empty state and the welcome email in the same week. Activation went up. You don't know which one moved it.
Instrumentation drift mid-test. Halfway through the test, a release changed how the activation event fires. Your pre/post comparison is now meaningless.
Vanity metric reads. You measured clicks instead of conversions. Clicks went up. Conversions didn't.

Design experiments that produce a learning even when the result is null. That means: pre-register the hypothesis, the metric, and the sample size. Don't change other variables on the same surface during the test window. If the result is null, write down what that null result tells you about the hypothesis (sometimes a null result kills an entire branch of the backlog, which is genuinely useful). A null experiment with a clean read is more valuable than a "winning" experiment with a dirty one.

Working without a data analyst

You probably won't have one. Or you'll share one across four PMs. Three things to set up so you stop being blocked:

SQL basics you actually need. Joins (inner and left), window functions (ROW_NUMBER, LAG, cumulative sums), and cohort retention queries. That's it. You don't need to be a Hex power user. You need to be able to write the cohort retention query and check it against Mixpanel.
Three dashboards you build yourself. A funnel dashboard (visit → signup → activation → retention by week), an experiment results dashboard (per active experiment, with sample, lift, p-value), and a North Star trend dashboard (your chosen metric, weekly, with the previous quarter as context). Build these in Mixpanel or Amplitude, not in the warehouse. Speed of iteration matters more than query elegance.
How to negotiate analyst time when there is one shared. Don't ask for "help with growth analysis." Ask for one specific query, with the SQL you've already attempted, and a deadline that's three days out. Analysts say yes to this. They say no to "can you help me think through activation."

The CEO conversation

The "double signups by Q4" framing is almost never the actual goal. The actual goal is usually revenue, which means activation and retention matter at least as much as raw signups. Reframing without sounding like you're pushing back is a skill. Here's a script that's worked for me:

"I want to make sure I deliver on the signups goal. To get there responsibly, I spent the last month auditing the funnel. Here's what I found: we're losing 70% of signed-up users at the activation step, which means even if I double signups, the revenue impact is roughly half of what you'd expect. I've identified the activation fix that I think gets us 60–70% of the way to your revenue target with about 40% of the work of doubling top-of-funnel. Can I show you the plan? If you still want me to prioritize raw signups after seeing this, I'll switch."

That's not pushing back. That's showing your work. Most CEOs respond well to it because you've translated their request into something measurable, you've done the homework, and you've left the final call with them.

90-day scorecard

Grade your own first 90 days. Honest answers only.

Eight or more yeses and you're set up to compound for the rest of the year. Five to seven and you've got specific things to fix before month four. Below five, find a peer growth marketer at another company and trade notes — something structural is in the way, and it's almost never a personal performance issue.

The growth hires who survive year one are almost always the ones who spent month one auditing, not shipping. You're not behind. You're paying the tax that lets you compound for the next eight quarters instead of burning credibility in the next eight weeks.

Growth Marketer Playbooks