How to run the first 90 days, from someone who just did.

Demand gen at an early-stage company moves fast on every front: the market, the product, the team, the buyer. The plan is a useful starting point. The work is figuring out what to keep and what to throw out as you go.

What I want to give you here is the version of my 90 days I'd write knowing what I know now, the one with the trade-offs and the things I'd actually do differently.

A few things to know about the context:

• I joined Vector as the first demand gen hire in seat. If you're walking into something similar, the playbook below maps closely. If you're at a Series C with an existing team, the timelines compress, but the experimentation mindset still applies. Early-stage forces you to ship fast and learn from live campaigns instead of planning in circles, and that's worth borrowing regardless of stage.

• Most of this is anchored in the early-stage startup reality: small team, real pipeline pressure, a stack you'll need to build and not just inherit.

• I'll be honest about what worked and what didn't. That's the only version of this that's useful.

You need enough attribution to know what's working directionally. You don't need a fully built attribution platform with multi-touch and customer journey mapping before your first brand campaign goes live. Those are great things to have, but they're not blockers.

The reframe I'd offer: get the directional reporting infrastructure in place, ship campaigns fast, then build the more sophisticated layers in parallel. If pipeline is growing, you don't need to defend a dashboard.

The pain points matter, but the words buyers use to describe them matter more. At Vector I found a gap pretty quickly between how sales was pitching the product and how customers were actually talking about the problem they'd hired us to solve. That gap is worth looking for at any company you join. It usually tells you something the messaging hasn't caught up to yet.

One useful early test for that gap is a competitor conquest campaign, running ads against the established names in your catI'd tell every new demand gen hire to say this out loud to leadership before they ask: "I'm spending the first 30 days understanding the company, the market, and building the strategy. The second 30, I need to give the platforms time to learn. We'll judge what's working at the 60 to 90 day mark." If you set that expectation up front, you'll get the runway you need. If you don't, you'll spend month two playing defense.egory to see how your product positions next to them. You capture some of their traffic, and you learn how your product positions when it's sitting next to theirs.

What you can build by the end of month two.

If the creative audit went well and you have enough data to tell you that you're directionally producing results, the natural next layer is connecting your campaigns into shared audience structures that feed into each other. This is where compound interest can build. Multiple creative formats and multiple different campaigns and channels. At Vector, that eventually became what we called the B2B surround sound play: one core ICP audience, brand awareness on connected TV, retargeting back into LinkedIn engagement ads and thought leadership, plus a Google search layer to capture the demand the awareness campaigns were creating. We wrote up that exact play in a separate blog if you want the actual mechanics.

You don't have to copy that structure. But the principle to steal is layered campaigns built on a shared audience, not five disconnected campaigns competing for the same budget.

This is also when you start the signal conversation.

If you're going to activate signals or run an ABM motion later in the year, the foundation for that work starts now. You don't have to have a target account list or have a signal architecture finalized yet. You do need to start asking which closed-won deals had which signals present before they closed, because that's the analysis that builds the list. We'll go deeper on this in section 9.

Run an honest retro.

The retro version that helps you is the one where you're willing to kill things that didn't work. The retro version that hurts you is the one designed to make the first 60 days look good.

The questions worth asking:

• Which campaigns produced pipeline. Which produced engagement without pipeline. Which produced nothing.

• Which creative is fatiguing. (Creative is fatiguing faster than it used to, more on this in the AI section.)

• Which channels punched above their weight. Which under-performed and why.

• Which audiences engaged at higher rates. Which ones I'd retire.

Plan the next 90.

By the end of month three, you should have enough signal to say what the next quarter looks like. For most teams at this stage, that means doubling down on the channels that worked, starting the signal or ABM motion, and building out a more mature reporting layer.

The reporting answer at 90 days.

By 90 days you should have enough to show direction, not enough to defend every dollar. The attribution work takes longer than that, and setting that expectation with leadership early is the same play as telling them in month one to give the platforms time to learn.

The framework I use is simple. Three layers: what you need to ship, what you'll need by month two, and what you can wait on entirely. If something doesn't help you ship work in the next 90 days, it's probably not a day-one tool.

The Vector layer.

Two pieces of the foundation above are where Vector fits naturally: an audience builder that doesn't require you to have a target account list before you can run anything, and engagement visibility that goes beyond "someone clicked, someone didn't." Our ICP Builder lets you define your audience by who they actually are. Our Reveal product tells you, by name, who clicked your ads and visited your site. Both are things I needed on day one.

The reason I'm putting this in a kit for new demand gen hires (and not in a list of "later" tools) is that the longer you wait to build the audience layer right, the more budget you spend reaching people you don't actually want. That's why this category is foundational for me.

What you need to launch.

Genuinely minimum. Three things:

• Conversion tracking from website to ad platforms. Tag manager set up, conversion actions defined, UTM parameters defined and taxonomy defined, the basics. If this isn't in place, fix it before you launch anything.

• Offline conversion tracking from CRM to ad platforms. So the ad platforms know when a demo or trial signup became a qualified lead, and when that lead closed. This is what teaches the platforms to find more of your real buyers, so they stop optimizing toward bots or non-ICP form-fillers.

• Last-touch and first-touch reporting in your CRM. Even if it's basic. Even if it's just utm parameters tied to the lead source field. You need to be able to tell a directional story by month two.

What you can wait on.

Everything else, including the things that probably feel essential right now. Multi-touch attribution. Customer journey mapping. Impression-level account influence. A purpose-built attribution platform. All of these are valuable. None of them block your first campaign.

What this looked like for me.

I used in-platform reporting and CRM last-touch for the first 60 days, paired with Google Looker Studio and Supermetrics for the manual reporting work. Around month two I bought a reporting tool to handle offline conversion tracking and customer journey mapping, because the manual work had grown beyond what I should be doing by hand.

The framework I use thinks about signals in three dimensions. Fit and relevance happen at the company level. Engagement happens at the person level. The most valuable accounts and contacts sit at the intersection of all three.

The narrowing exercise.

I worked closely with our product marketing manager and our RevOps Leader Sara to do a full ICP analysis, including signal detection in recent closed won deals. We also looked at closed-lost, sales qualified and sales disqualified. Here's an exercise that actually produces something usable:

1. Pull your last 10 to 20 closed-won deals.

2. For each one, work backward through the signals that were present before they closed. Look across all three dimensions.

3. Build a hypothesis list of signals you'd target. I started with around 35.

4. Narrow to the five or six signals that showed up most consistently. Those become your foundation. Your hypothesis.

Going from 40 to five or six feels brutal. It is. But the discipline of narrowing is what makes the signal-based campaigns work later, because you're targeting the patterns that actually predict your wins instead of whatever's easiest to query. Complex signal architecture is earned over time. Get your wins in with the most important signals.

Why this work usually starts sooner than you think.

I'd planned to do this kind of work toward the end of the year, but it ended up being one of the bigger pieces of my first 90 days. Once you start asking which closed-won deals had which signals present, that question pulls on a lot of threads: the audience definition for your campaigns, the target account list, how the team talks about who's worth pursuing. You don't have to activate a full signal-based motion in month two, but you do need to start the conversation.

The other reason is that signals shape who your broader campaigns actually reach. Even if you're running brand awareness, knowing your signal layer informs how you narrow the audience down from broad reach to the people most likely to be in market. Starting on the architecture early pays for itself even if you don't activate the full motion for another quarter.

Start in chat. With work you're already an expert at.

The most useful version of AI for a demand gen leader is small and specific: build a workflow for the thing I do every week or every day, where I know what good looks like, and get the output to 90% of how I'd do it myself. Then you use it daily or weekly, and from there you can start moving into more advanced workflows. Automating the entire stack can wait.

If you start in chat, in your domain expertise, you have taste. You can tell when the output is wrong. AI is fast and confident, and if you don't have the expertise to push back, the confident output starts to look like the right output. That's how you end up with polished, structured, very confident AI slop in your campaigns.

Starting here gives you a chance to refine your frameworks, strategy, and context about how you think and your company. The context layer is extremely important to get right because it's the foundation for everything that comes after.

Where Vector MCP fits, honestly.

Vector MCP is one example of an AI workflow worth setting up early in your tenure, because most of what it answers are the questions you're going to be asking yourself every single morning anyway. The questions I use it for most: