Most founders pick a growth partner the same way they'd pick a restaurant: referral, decent website, gut feel, sign. Six months and $90K later, they have a nicely formatted monthly report, a Slack channel that's gone quiet, and no measurable change in pipeline.
The fix isn't a better gut. It's a better interrogation.
The short answer: before you sign with any growth agency, consultant, or embedded pod, get direct answers to 27 questions across seven areas — fit, team, methodology, data, engineering, AI, and commercials. The three that predict outcomes most reliably are: Who exactly does the work day to day?, What did you ship in week one of your last engagement?, and Show me an experiment that failed and what you did next. Vague answers to those three mean walk away, regardless of how good the deck is.
Below is the full checklist, with what a strong answer sounds like and what should make you close the laptop.
How to use this checklist
Don't fire all 27 questions in one call — you'll get canned answers and a tired sales rep.
Run it in three passes:
| Stage | Questions to ask | What you're testing | Time |
|---|---|---|---|
| Discovery call | 1–9 | Fit, seniority, who does the work | 30–45 min |
| Working session | 10–23 | Methodology, data, engineering, AI depth | 60–90 min |
| Contracting | 24–27 | Commercials, IP, exit | Async + legal |
Score each answer 0 (evasive), 1 (adequate), 2 (specific and evidenced). Anything below 38/54 total is a pass. Anything with a zero on Q5, Q11, or Q17 is an automatic no regardless of total score.
We at Growaton get put through versions of this checklist by good founders every month. The ones who ask hardest tend to become the best clients — because they've already thought through what they actually need.
Fit and focus (Questions 1–4)
1. Which specific stage and business model do you do your best work in?
Growth tactics that work for a $50M-ARR e-commerce brand actively destroy value at a seed-stage B2B SaaS company with 40 customers. A partner who says "we work with everyone" works with everyone badly.
Green flag: "Seed to Series C, SaaS, fintech, marketplaces, e-commerce — here's why the marketplace playbook differs from SaaS in our approach to supply-side liquidity." Red flag: A logo wall spanning enterprise banks, local dentists, and crypto.
2. What kinds of companies do you turn down?
The cleanest signal of focus is refusal. A partner with no disqualification criteria has no thesis.
Green flag: "Pre-product-market-fit companies. If you don't have retention signal yet, we'd be optimizing a leaky bucket and taking your money." Red flag: "We can help anyone who wants to grow."
3. What do you think is actually broken in our funnel right now?
Ask this on the first call. Any competent partner has already looked at your site, pricing page, signup flow, and ad library. If they haven't, they're not curious enough to be useful.
Green flag: Three specific hypotheses with reasoning, plus an admission of what they can't know without data access. Red flag: "We'd need to run a full discovery phase before we could say anything."
4. What's your point of view on where growth is going in our category?
You're buying judgment, not hours. A partner without opinions will default to whatever you ask for — which means you've hired an expensive pair of hands, not a thinking partner.
The team: who actually does the work (Questions 5–9)
This is where most engagements quietly fail. The bait-and-switch — senior partners in the pitch, junior account managers in delivery — remains the single most common complaint founders raise about agencies.
5. Who, by name, will be in our Slack channel every day?
Green flag: Named individuals, their LinkedIn profiles, their prior work, and a clear statement of how many hours per week each contributes. Red flag: "You'll be assigned a dedicated account team." Automatic disqualification if they won't name people before contract signature.
6. What's the average years of experience on the delivery team, not the leadership team?
Ask for the delivery team's median. In our experience, the gap between pitch-team seniority and delivery-team seniority is the best single predictor of engagement disappointment.
7. How many other clients does each person on our team work on simultaneously?
Three is manageable. Seven means you get 5.7 hours a week of attention from someone who's context-switching constantly.
8. What disciplines are covered in-house versus subcontracted?
If your growth partner subcontracts engineering, every experiment that requires a code change becomes a two-week procurement exercise. This is the structural reason siloed agencies plateau: the marketer can't ship, and the developer doesn't know what to ship.
Green flag: Product, engineering, data, experimentation, and marketing under one roof, in one pod, on one standup. (This is precisely why we structure around embedded pods rather than departments.) Red flag: "We partner with a great dev shop."
9. What happens if the person we like leaves your company mid-engagement?
Continuity plans, documentation standards, and shared context matter. Ask to see how they document decisions — a real answer includes an artifact, not a promise.
Methodology and experimentation (Questions 10–14)
10. Walk me through your process from kickoff to first shipped change.
You want a named, repeatable methodology — not "we're flexible." Flexibility without structure is improvisation billed hourly.
Our own 4-Phase Growth Framework runs Diagnostics → Measurement → Conversion → Scale, and we're happy to be judged on whether an engagement actually follows it. Whatever framework your candidate uses, make them explain the phases, the exit criteria for each, and where most engagements get stuck.
11. What did you ship in the first two weeks of your last three engagements?
The most revealing question on this list. Real partners have concrete answers: a rebuilt onboarding step, an event-tracking implementation, three landing page variants live in market.
Red flag: "The first month is discovery and strategy." A month of discovery before anything ships is a month of your runway spent on a document.
12. How many experiments do you typically run per month, and what's your win rate?
If they claim a 70% win rate, they're either measuring badly or lying. Microsoft's experimentation team has reported that roughly one-third of tested ideas produce positive results, one-third are neutral, and one-third are negative — a pattern echoed across large-scale experimentation programs (Kohavi, Tang & Xu, Trustworthy Online Controlled Experiments; HBR, "The Surprising Power of Online Experiments").
Green flag: "We aim for 6–10 meaningful tests a month at your traffic level; historically about a third win outright, and we kill the rest fast." Red flag: Any win rate above 50% quoted without qualification.
13. Show me an experiment that failed and what you learned from it.
Ask for the hypothesis, the result, and the next experiment it informed. A partner who can't produce a good failure story either hasn't run enough tests or isn't measuring honestly.
14. How do you decide what to test next — and who has the final call?
You're looking for a prioritization system (ICE, PIE, RICE, or their own) plus a clear decision right. Ambiguity here produces the classic failure mode: two months of arguing about the roadmap.
15. What's your minimum traffic or volume threshold for statistically valid testing?
A partner who promises A/B testing on 400 monthly visitors is selling you a coin flip. Good partners tell you when your volume is too low and pivot to qualitative research, sequential testing, or bigger swings instead.
Data, measurement, and attribution (Questions 16–19)
16. How will you instrument what we can't currently measure?
Most seed and Series A companies have broken or partial event tracking. If your partner can't fix the measurement layer themselves, every result they report is an assertion.
17. What metric will you be accountable for, and how is it calculated?
Get this in writing. "Growth" is not a metric. "Qualified pipeline sourced," "activation rate (defined as X within Y days)," "net revenue retention," or "blended CAC payback in months" are metrics.
Green flag: A single north-star metric plus 2–3 guardrail metrics so they can't win the number by torching another part of the business. Red flag: Reporting centered on impressions, sessions, MQLs, or "brand lift."
18. Who owns the dashboards, and can I see live data without asking you?
You should have read access to the same view they do, updated continuously. Monthly PDF reports are a control mechanism, not a reporting standard.
19. How do you handle attribution when channels overlap?
There is no perfect attribution model. What you want is intellectual honesty: an understanding of the difference between last-touch, multi-touch, and incrementality testing, plus a willingness to run holdouts or geo-tests when the spend justifies it.
20. How do you connect growth activity to unit economics?
Ask how they'd model CAC payback, LTV:CAC, and contribution margin for your business specifically. A partner who optimizes for lead volume without a view of payback period will happily grow you into insolvency. Marketing typically runs around 7–8% of company revenue on average across industries (Gartner CMO Spend Survey) — your partner should know where you sit relative to that and why.
Engineering and technical capability (Questions 21–23)
21. Can you ship production code in our stack, and who reviews it?
If growth requires engineering — and it always does past the low-hanging fruit — you need to know whether they can write, test, and deploy code, or whether they'll be filing tickets with your already-overloaded team.
Green flag: Named stack experience, a code review process, and a statement about how they hand back to your engineers. Red flag: "We'll work closely with your dev team." Translation: your roadmap absorbs their ideas.
22. What's your policy on technical debt and handoff quality?
Growth engineering has a reputation for leaving landmines. Ask what standard they hold themselves to for tests, documentation, and feature flags, and what happens to hacky experiment code once a test wins.
23. What tooling and infrastructure would you introduce, and who pays for it?
Surprise SaaS bills are a recurring source of friction. Get the expected tool stack and monthly cost range in the proposal.
AI and automation (Questions 24–25)
Every agency now claims AI capability. Very few can show payback.
24. Where specifically do you use AI in delivery, and what does it change for us?
Green flag: Concrete workflows — programmatic SEO page generation with human editorial review, LLM-assisted qualitative analysis of support tickets, automated lifecycle messaging, code generation that a senior engineer reviews. Plus a clear statement of what they don't automate. Red flag: "We're AI-native" with no examples, or worse, unedited AI content shipped to your blog.
25. How do you measure the ROI of the AI tooling you deploy on our account?
The honest answer involves cost-per-output, cycle time reduction, and quality holds — not vibes. This is the question that separates teams using AI to increase throughput from teams using AI as a pricing narrative. (We've written a full teardown on proving whether LLM spend actually pays back because the gap between claimed and realized AI ROI is where a lot of budget currently disappears.)
Commercials, contracts, and exit (Questions 26–27)
26. What's the pricing model, what's the minimum term, and what's the exit clause?
Retainers, pod-based pricing, performance fees, and equity swaps all have different incentive geometries:
| Model | Aligns on | Risk to you | Best when |
|---|---|---|---|
| Hourly / time-and-materials | Hours billed | Scope creep, slow pace | Narrow, well-defined projects |
| Fixed retainer | Deliverable volume | Coasting after month 3 | Ongoing, predictable workstreams |
| Embedded pod | Shipped outcomes on a cadence | Higher monthly cost | You need multiple disciplines moving weekly |
| Performance / CPA fee | Volume of the paid metric | Quality erosion, gaming | Mature, well-instrumented funnels |
| Equity-inclusive | Long-term value | Cap table dilution, misaligned horizon | Early stage, capital constrained |
Green flag: A 30-day rolling notice after an initial 90-day period, and a written statement of what triggers a mutual exit. Red flag: A 12-month lock-in with no performance break clause. If they need a year of contractual protection, they don't trust their own first 90 days.
27. What do we own when this ends — and what does offboarding look like?
Code, creative assets, ad accounts, analytics properties, documentation, and experiment archives should all be yours, in your accounts, from day one. Ask for a written offboarding plan before you sign, not after you're unhappy.
Red flag: Ad accounts or tracking properties held under the agency's business manager. This is the most common form of hostage-taking in the industry.
The five answers that should end the conversation
If you only remember one section, remember this. Walk away when you hear:
- "We can't name the team until after signature." Delivery quality is people, and they're hiding the people.
- "The first 30 days are discovery." Discovery and shipping run in parallel. Anything else is billable stalling.
- "Our win rate is around 70%." Either the measurement is wrong or the honesty is.
- "You'll get a monthly report." Weekly cadence with live dashboards is now the baseline, not a premium feature.
- "We'd need a 12-month commitment." Fine to want one; a red flag to require one with no break clause.
What a good first 90 days actually looks like
Use this as your benchmark when a partner describes their plan. Anything materially slower deserves an explanation.
| Timeframe | What should exist |
|---|---|
| Week 1 | Data and tool access, first instrumentation gaps identified, at least one shipped change (copy, page, tracking) |
| Week 2–4 | Baseline metrics agreed and documented, prioritized experiment backlog, 2–4 experiments live |
| Week 5–8 | First statistically meaningful results, one channel or funnel stage measurably improved, weekly shipping rhythm established |
| Week 9–12 | A repeatable operating cadence, a working measurement layer you own, and a clear scale-or-kill call on each initial bet |
The pattern we've seen across seed-to-Series C engagements is that the trajectory of the first 90 days almost perfectly predicts the trajectory of month 12. If it's slow at the start, it doesn't accelerate later — it just gets more expensive.
Before you run the checklist on anyone else
Two things worth doing first.
Decide what model you actually need. A specialist agency, an embedded pod, a fractional operator, and an in-house hire solve different problems at different price points. If you have one clearly defined channel bottleneck and internal engineering capacity, a specialist is cheaper and better. If your bottleneck spans product, data, and marketing at once — which it usually does between Series A and B — a single senior pod that can ship across all three avoids the coordination tax entirely.
Be honest about whether you're ready. Growth partners multiply what exists. If retention is broken, a partner will just help you acquire churning customers faster. Have revenue, some product-market-fit signal, and a willingness to let a partner change the product — not just the ads.
If you'd like a second opinion on your funnel before you shortlist anyone, book a free growth diagnostic. You'll get a specific read on where your growth is leaking and what we'd test first. No obligation, and you're welcome to take the findings to whichever partner you choose — including none.
