The short answer
Both sides of this can be automated, so automation is not the choice. The choice is whether you buy attention in bulk or spend it one prospect at a time, and what a customer is worth over a year decides which you can afford. No independent source sets a threshold, but the measured pattern is a trough: acquisition economics are worst in the $25,000 to $50,000 band and better on either side.
The answer is mostly decided by one number: what a single customer is worth to you over a year. The lower that number, the less attention any individual prospect can justify, and the more your growth belongs to ads. The higher it goes, the more the work sits in the conversation.
The measured pattern behind that: Benchmarkit's 2025 benchmarks, from 583 companies reporting full-year financials, found acquisition economics worst in the $25,000 to $50,000 band — about $2.40 spent per dollar of new recurring revenue — and better below it, at roughly $2.20 in the $10,000 to $25,000 band. Efficiency improves above $50,000 and again above $250,000.
So the shape is a trough, not a cliff, and it sits where deals are large enough to need a human selling motion and too small to pay for one. The rest of this explains what to do on each side of it.
| Typical annual contract | Where the money goes | Why |
|---|---|---|
| A few hundred to ~$5,000 | Ads and self-serve | The buyer does not want a call. Per-conversation cost cannot be recovered. |
| ~$5,000 to ~$25,000 | Ads to create the enquiry, then a light sales process to answer it | Deals are won on being answered. Acquisition economics are hardest in this band. |
| ~$25,000 upward | Ads for awareness, individual outreach and research to close | One closed deal justifies weeks of attention. |
Automation applies to any row. It changes what each costs to run, not which row you are in.
Why contract value does most of the deciding
Individual outreach is expensive per prospect. Researching one company properly, writing something that is not obviously templated, following up three times without being irritating — that is real time whether a human does it or an agent does. The cost per conversation is roughly fixed.
Ads are the opposite. The cost per impression is tiny and the work is mostly upfront: build the creative, point it somewhere, let volume do the qualifying.
So the question is not "which is better." It is whether a single customer is worth enough to pay for the attention that individual outreach requires. At the extremes this answers itself. The interesting part is the middle, and the middle is exactly where the measured data says acquisition is hardest.
Small contract values: ads, and stop thinking about outreach
If your typical customer pays you a few hundred dollars a month, your buyer does not want a salesperson. They want to understand the thing, decide quickly, and sign up without a call. Putting a human — or an agent pretending to be one — in front of that decision adds friction to a purchase that was ready to happen.
Start on Meta before Google. Not because Meta is better, but because it is cheaper to be wrong on. Entry budgets are lower, the creative iteration loop is faster, and you can learn what message works for twenty dollars a day. Google search ads capture people already looking for a solution, which is more valuable and correspondingly more expensive per click. Learn on Meta, then buy the higher-intent traffic once you know what converts.
Where agents earn their place at this level: answering inbound fast, keeping the record straight, and telling you weekly what needs a decision. Those cost the same whether you have ten enquiries a month or a thousand, so they scale with you. What does not earn its place is deep research on every individual prospect.
The middle: a real sales process, but a light one
This is the band the Benchmarkit data says is hardest, and the reason is visible in how people buy here. They want to talk to someone before they commit, but not for long and not more than once or twice. Deals are won by responsiveness rather than by depth, so the selling motion has to be human enough to answer and cheap enough not to eat the margin.
The highest-return thing you can do at this level is close the gap between someone raising their hand and someone answering. An enquiry with no reply has nothing holding it, and the buyer keeps looking. If your enquiries sit for a day because you were in back-to-back calls, you are paying full price for leads and collecting a fraction of them.
On how fast is fast enough, the popular numbers do not survive checking. The five-minute rule and its 21x and 100x multipliers trace to two studies from 2007 and 2011, funded by a company selling lead-response software, never independently replicated. What has been tested is whether anything answers at all: a pre-sale chatbot raised sales 16.3% among 44,614 consumers (Fang et al., working paper, October 2025). Build for answering. Treat the specific minute threshold as unmeasured.
Ads still do the finding. What changes is that the follow-up now matters enough to build for.
Large contract values: the conversation is the product
Long, considered sales, several stakeholders, months of thread. Ads can make someone aware of you, but the work that closes the deal is research, memory, and persistence over time — knowing what was promised in July when the call happens in October.
This is where deep individual research and personally written follow-up most clearly pay back, because a single closed deal justifies weeks of attention.
The test, if you only remember one thing
Take what a customer pays you in a year. Divide by the hours you would need to spend to win one. If that number is not comfortably above what an hour of your time is worth, stop selling individually and buy attention instead. Run it on your own numbers, because the answer moves with your close rate and your retention.
Most founders get this wrong in the same direction. They build outreach machinery for a product that should be sold through a funnel, because outreach feels like doing sales and running ads feels like spending money. The machinery is more expensive than the ads, it is just that the cost is hidden in your calendar.