Recommendation: go hybrid, with PLG as the top of the funnel and sales-assist on top of it. Don't choose one motion over the other.
Your data already points this way. Signups are growing while revenue is flat, so self-serve is producing volume but not dollars. Your biggest accounts come from sales calls, so the buyers who matter aren't converting through the product alone. This matches the skill's PLG-vs-sales finding: self-serve brings volume at low ACV (~$5K), and sales-led brings high ACV (~$50K). In the skill's parallel test, sales won about 10x on revenue despite about 10x less volume. That test isn't your data, so treat it as a hypothesis for you to check.
Run the decision tree on your product
- Can a user get value in under 10 minutes without docs? For an API this is often yes, so measure time to first API call.
- Can they self-serve implementation? Small projects usually can. Production integrations at larger companies usually can't.
- Is the buyer the user? Your developers are the users. The budget owner at a large account usually isn't a developer.
If the answers are yes, yes, no, the tree gives hybrid. Check each answer against your own activation data.
What to do
- Keep the free tier and self-serve checkout for accounts under about $10K.
- Add sales-assist for $10K–$50K accounts. Trigger it on product-qualified signals:
- several users from one company
- usage near the free-tier limits
- questions about SSO, SLAs, security or team pricing
- Make outreach specific. For example: "Your team is calling endpoint X from 12 services. Want 15 minutes on rate limits?" Don't send "I saw you signed up."
- Leave sub-$5K accounts alone. Early sales contact on small deals hurts the self-serve motion.
- Look at your closed-won accounts first. Did they start as free signups and then get a call, or did they come in through sales without ever using the product? If most started in the free tier, the PQL triggers will work. If not, the free tier is mostly serving hobbyists.
- Run a 4–6 week test. Compare PQL-triggered outreach against a no-contact control group on conversion and ACV.
- Fix activation. If time to first value is over 10 minutes, shorten it with sample data, a working quickstart and less setup before the first call.
Reviewing channel economics (SEO blog, paid search, community events)
Report every metric by channel and by acquisition source, not as a blended number. The weakness in most reviews is that they stop at signups.
| Metric | Why it matters |
|---|---|
| CAC (spend ÷ new users) | Include content production and staff time for SEO and events, not just ad spend. |
| Signup → activated (first successful API call) | Shows whether a channel brings real developers or just clicks. |
| Signup → paid, and → $10K+ | Your revenue comes from large accounts, so measure this too. |
| 30- and 90-day retention | Cheap users who churn cost more than expensive users who stay. |
| LTV and payback period | Use revenue, and revenue by account size, not signup counts. |
Then apply the rule:
- CAC below LTV × margin: scale. Check 90-day retention first. If it's above 60%, put about 3x the budget there. If it's below, fix activation first.
- CAC about equal to LTV × margin: optimize, but don't scale.
- CAC above LTV × margin: give it 4 weeks to improve, then kill it.
Build a growth equation for each channel before you scale it. Examples are 1 post → N visits → X% signup, $1K → clicks → signups, and 1 event → attendees → users. Validate with 4 weeks of data, then scale on that basis.
Channel-specific notes (my expectations, to be checked against your numbers)
- SEO blog: often the cheapest per user and the best retained. Segment by post type, because tutorials and integration guides probably bring developers who build, while top-of-funnel posts bring browsers.
- Paid search: probably the highest CAC and the lowest retention of the three, but test that. Look at which keywords bring activated accounts, and cut the ones that only bring signups.
- Community events: small volume, but attendees may be the most likely to become sales-assisted accounts. Track attendee → account → pipeline through CRM, not signups alone. Events are the channel most likely to look bad on a self-serve-only view and good on a pipeline view.
- Sales-sourced accounts: put your sales calls in the same table as a fourth channel, with fully loaded CAC. That's how you compare them fairly with the other three.
Cadence
- Review monthly. Reallocate quarterly. Test two new channels a month.
- Forecast baseline, upside and downside, and update the forecast monthly.
- Write a one-page playbook for each channel that works.
What I need from you to turn this into numbers
- spend and new users by channel
- activation, paid conversion and retention by source
- ACV and LTV by source
- the origin story of your top 20 accounts
I haven't seen any of your data, so everything above about your channels is a framework or a hypothesis, not a finding. If you share even a rough export, I can build the channel table.
Real run in an isolated folder with only this skill installed. Only the Skill and Read tools were enabled, so nothing was fetched from the web and no file was written. The example is fictional. The model answered from SKILL.md alone and said it had no real channel data.
What it does
Helps the model reason about self-serve growth instead of assuming it always works. It covers a PLG reality check (the author's six-month parallel test where sales-led won on revenue), a growth equation that maps inputs to outputs per channel, channel economics (CAC, conversion, retention, LTV, payback), time to first value, the point where self-serve breaks and sales assistance should step in, growth forecasting in three scenarios, and a one-page playbook habit.
How it works
- The model checks whether value is clear in minutes, setup is trivial and buyer equals user.
- It builds a growth equation and a keep / scale / kill rule for each channel.
- It suggests sales-assist triggers based on usage, team expansion and buying signals.
Good for
Developer tools and B2B SaaS deciding between PLG and sales, fixing freemium conversion and prioritising channels.
Pure instructions: no scripts, no network access, no file writes, no accounts. Figures such as the $5K to $50K inflection and the 4-week kill rule are the author's experience, not benchmarks. The skill does not see your analytics, so any numbers come from what you provide. Treat the output as a plan to test, not a forecast.