Section D · Execution

Day-Of Tactics & Your 30-60-90

The plan to walk in with, the questions to ask, the traps to avoid, and the closing statement. Reread the morning of.

The 30-60-90 to walk in with

The single strongest thing you can bring to a first-data-hire interview is a concrete, confident plan for your first 90 days. It signals that you've already inhabited the role. Deliver it verbally, unprompted, when they ask "how would you approach this job" — and don't over-promise. The honest truth, which you should say out loud, is that a first data hire produces no meaningful insight for roughly the first three months while the foundations get built. Saying so is a credibility signal, not a weakness. This mirrors the build plan in 03 · Building the Data Stack — here it's compressed into something you can say.

Window	Theme	What you actually ship
Days 0–30	Listen & audit	Learn the business, KPIs, and systems; interview every stakeholder (product, eng, finance, marketing, leadership); audit every data source and what's actually logged; map the safe-vs-unsafe data surface (02); align a plan with the VP of Business Ops. Publish a short "what we will never collect, and why" guardrail doc — it earns trust with a privacy-first founder and sets the cultural tone.
Days 31–60	First light	Stand up the warehouse (start on Postgres or BigQuery) + ingestion (Airbyte/Fivetran + custom loaders for gateway logs and on-chain) + dbt v1; ship the first dashboards on the metrics that already matter — MRR and its movements, free→paid conversion, engagement on content-free metadata, gross margin after inference COGS. Get one real decision made off your numbers.
Days 61–90	Foundation set	Ship a v1 conformed data model with standardized metric definitions (one source of truth, robust to rapid product iteration); stand up experiment-tracking foundations in PostHog (05); and deliver a 6–12 month roadmap with costs, hires, and risks named. Now you can start answering strategic questions proactively.

The framing that lands

"My first 90 days are about earning the right to have opinions. I don't ship a churn model on day 30 — I ship trustworthy MRR, conversion, and margin numbers, a metric layer everyone agrees on, and a written privacy guardrail. The models come after the foundation, because at a company that deliberately doesn't log user content, the foundation is the hard part."

Verify these the morning of

The company's public facts move. Stale numbers are unforced errors, and being precise about live figures is literally the job. Spend ten minutes refreshing:

Pricing — the company's live pricing page. Confirm the tiers (Free / Pro ~$20 / higher paid tiers) and the credits ratio haven't shifted.
Model list — the company's live model catalog. It rotates; don't name a retired model.
Token price & market cap — CoinGecko or CoinMarketCap. Know roughly where the staking token sits today, and remember the treasury is partly denominated in it.
The posting — re-read the exact job listing in case responsibilities changed.
Any fresh news — a quick search for the company catches a funding update, a token event, or a product launch worth referencing.

Use the product

If you haven't already, run a few prompts through the product and generate an image the night before. First-hand product experience — "when I used it, I noticed X" — beats any amount of secondhand research in the room.

Questions to ask them

Your questions are an interview round you control. Ask ones that only a senior operator who understands this company would ask — they double as evidence you've done the work.

On the core tension: "How does leadership actually feel about the privacy-versus-measurability trade-off? Where has not-logging genuinely cost you a decision, and where has it saved you?"
On what's blocked: "What's a decision the team is currently making on gut feel because the data isn't there yet? That's probably where I'd start."
On the token: "How do you think about on-chain token metrics versus fiat revenue? Do you treat the staked token as committed demand, or watch them separately?" (Shows you understand 06 and 04.)
On success: "Six months in, what does it look like for this role to have clearly worked — and what would tell you it hadn't?"
On build vs buy: "For a team of roughly 50 people, how much appetite is there to self-host the analytics stack for privacy reasons versus adopt managed tools? Where's the line?"
On the culture clause: "The posting asks this person to advocate against unnecessary data collection. How often does that tension actually come up, and how do you want it navigated when growth wants more tracking?"

Traps to avoid

These lose the loop

Treating privacy as an obstacle. The instant you frame no-logging as "a problem we need to work around," you've failed the alignment screen. It's the product. (Reframe: 02.)
Over-indexing on fancy ML. This is a build-the-foundation, own-the-truth role. Leading with deep-learning churn models reads as someone who'll build window dressing instead of plumbing.
Stating unaudited numbers as fact. Say "reported ARR in the tens of millions, though that's self-reported." Precision about certainty is the job.
Ignoring the token economy. Most candidates forget the entire on-chain data source. Mentioning it — with the consented-linkage ethics line intact — is a differentiator.
Being cynical about the mission. A crypto-libertarian founder screens hard for genuine belief in privacy and free speech. Don't fake it — but if you're not aligned, this isn't your company.
Proposing a heavyweight enterprise stack. Snowflake + a five-tool modern data platform for a team of roughly 50 people signals poor judgment. Lean, self-hosted, sequenced.

Recovery patterns

You will hit a question you can't fully answer. Senior candidates recover; they don't freeze.

When you don't know: say so, then frame the approach out loud. "I don't have a memorized answer, but here's how I'd reason about it, and here are the two assumptions I'd need to check." Structured thinking beats a lucky guess.
On a fuzzy prompt: don't dive in. Decompose it aloud first — restate the question, name the decision it informs, state your assumptions, then answer. This is the exact skill the role tests.
After a rough round: reset. Each interviewer usually scores independently; a weak SQL round doesn't doom the loop. Take a breath, close the tab, start the next one fresh.
When challenged: if an interviewer pushes back, don't crumble or dig in blindly. "That's a fair point — here's what would change my mind, and here's what I'd still hold to." Defending a view and updating on evidence is what they're testing (08).

Your closing statement

When they ask "anything else" or "why you," have a short, genuine close ready that ties your background to the company's specific challenge. A template to adapt:

Adapt, don't recite

"What drew me here is that this company has the one analytics problem I find genuinely interesting: building a real data function without the crutch of user surveillance. Most companies would call that a handicap. I think it's a discipline — you fly on instruments, you're honest about what you can't see, and you earn trust by never collecting what you don't need. I've spent [X years] doing [your relevant thing], and I'd want my first 90 days here to be about proving we don't need to watch users to understand the business. That, and the privacy mission is one I actually believe in."

Keep it under 45 seconds. Specificity and sincerity beat polish — a crypto-libertarian founder can smell a canned answer.

Final checklist

☐ Can I explain what the company is, how it makes money, and why privacy is a strategic choice — in 60 seconds?
☐ Can I explain the token's staking-for-compute (the pro-rata rule) and the compute credit without notes?
☐ Can I deliver the six-beat answer to "build analytics without logging users" (02)?
☐ Can I draw the safe-vs-unsafe data map from memory?
☐ Can I name the reference stack (warehouse → Airbyte → dbt → Metabase → PostHog) and defend it for a 45-person team?
☐ Can I write MRR-movement and retention-cohort SQL fast, and handle the window-in-WHERE gotcha (07)?
☐ Do I have four behavioral stories ready — first-hire ambiguity, advocating against data collection, explaining uncertainty, and killing something?
☐ Have I refreshed pricing, model list, and token price this morning?
☐ Have I actually used the product?
☐ Do I have six sharp questions to ask them?

If most of those are checked, you're ready. Go back to 00 · Start Here for the reframe one last time, or the guide hub to jump to any chapter. Good luck — and remember the line that carries the whole loop: you're not flying blind, you're flying on instruments, deliberately privacy-safe ones.

✓ Check yourself

Answer from memory before opening each one.

Why does the plan ship a dashboard in days 31 to 60, before the conformed model exists, and what does the guardrail document in the first 30 days do?

Show answer
A visible early win buys the trust that the foundation work needs, so it is dashboard first and model second. The document on what you will never collect earns trust on the privacy side before any code exists. Together they say you are earning the right to have opinions. See the 30-60-90.
The morning of the interview you refresh a short list of facts. What is on it, and why is a stale number more costly here than at most companies?

Show answer
Live pricing and tiers, the model list, token price and market cap, the exact posting, and any fresh news. The role is about precision over what is certain, so an outdated figure undercuts the very skill being hired. See verify these the morning of.
A candidate calls no-logging "a problem we need to work around" and proposes Snowflake, a streaming layer and a catalog for a team of about 50. Which traps do they hit, and why do they cost the loop?

Show answer
They treat privacy as an obstacle when it is the product, and they propose a heavyweight stack that signals poor judgment. Both fail the alignment and restraint tests. The smaller answer is lean, self-hosted and sequenced. See traps to avoid.
Name the reference stack in order and defend why it fits a company of about 50.

Show answer
Postgres as the warehouse, Airbyte for ingestion, dbt for models, Metabase for BI, and PostHog for experiments. It is the smallest set that reconciles the numbers and ships a dashboard, it keeps data in the trust boundary, and each addition must earn its place. See the final checklist.

Takeaway

Walk in with the 30-60-90: listen and audit, ship first light, then set the foundation, and do not promise a model early.
Refresh live facts on the morning of, because precision about certainty is the job.
Ask questions that probe the tension between privacy and measurability, since only someone who has done the work would ask them.
The traps are privacy as an obstacle, leading with fancy machine learning, quoting unaudited numbers as fact, ignoring the token economy, and proposing a heavyweight stack.
When stuck, decompose the question aloud, state your assumptions, and update on evidence without crumbling or digging in.

Return to the hub to revisit any chapter before you go in. → Guide hub