Reported 2026 figures: a fully loaded in-house SDR costs roughly $9,700 to $17,500 a month once benefits, tools, management and attrition are counted. Freelance and outsourced setters run about $1,500 to $4,000. Usage-priced software runs tens of dollars. The comparison people skip is doing it yourself, which is not free: outbound is roughly a day a week, and for a technical founder that day comes out of the product.
You have a product, no pipeline, and three ways to fix it. The internet will tell you the answer is whichever thing the author sells. Here are the numbers instead.
Figures below are third-party reported for 2026 and labelled as such. Treat them as the right order of magnitude rather than a quote.
Option one: hire someone
Reported base salary for an SDR sits around $55,000 to $60,000, with on-target earnings between $83,000 and $85,000. That is the number people compare against, and it is the wrong one.
The fully loaded cost is reported at $116,500 to $210,000 a year, roughly $9,700 to $17,500 a month. The gap is benefits, recruiting, training, tools, software licences, management time and the cost of replacing them when they leave, which in this role they reliably do.
What you get for it is judgement. A good SDR notices that one segment converts at four percent and another at half a percent, and changes what they do about it without being told. Nothing in the software column does that yet.
The threshold question: at $9,700 a month, how many closed customers do you need to break even? At $2,000 annual contract value, roughly five closes a month just to cover the SDR, before anyone has been paid to run the calls. Most early startups cannot get there, which is why most early startups should not be hiring one.
Option two: a freelancer or an agency
Reported 2026 ranges: freelancers around $2,000 to $4,000 a month, offshore platforms under $1,500, nearshore around $2,500 all in, managed US agencies $8,000 to $12,000. Hourly work on marketplaces is reported at $45 to $95.
Considerably cheaper than a hire, and there are two costs that do not appear on the invoice. They usually bring no data licences and no sending infrastructure, so those are your line items on top. And one freelancer is one point of failure: when they are away, the pipeline stops, and you find out about it a month and a half later when the calendar empties.
The other thing worth knowing: someone paid by the hour to book meetings will book meetings. Whether those meetings are with people who could ever buy is a different metric, and it is the one nobody is incentivised to measure.
Option three: software
Usage-priced tools run tens of dollars a month at the volume one founder needs. Per-seat tools run $49 to $150 a month, with the caveats covered in the Apollo comparison. Dedicated AI SDR products at the enterprise end run into tens of thousands a year.
What software is genuinely good at is the clerical majority of the role: finding people who match a description, verifying addresses, writing a specific first message, following up on schedule, logging everything. That is most of the hours, as the breakdown of the job sets out.
What it is not good at is the edges, and it fails differently in a way worth planning for. A person doing the wrong thing does it forty times and usually notices. Software does it four hundred times before anyone looks. The mitigation is not avoiding the category, it is insisting on stopping rules and an activity log you can actually audit.
Option four, the one that is missing from every table: you
Doing it yourself gets priced at zero, and it is the most expensive option on the list for a technical founder.
Running outbound properly is about a day a week. The problem is not the total, it is the shape: most of that day is research, verification and CRM admin rather than writing, so it is a day that feels unproductive and shows nothing for three weeks. Which is exactly why founders start, see silence by week three, and stop one week before the thing would have started working.
And the day comes out of the product. If you are the only person who can build the thing, an hour of your time spent on list research is not worth the freelancer rate you avoided paying, it is worth whatever the shipping delay costs.
The comparison, with the missing column filled in
| Option | Reported monthly cost | Judgement | Fails by |
|---|---|---|---|
| In-house SDR | $9,700 to $17,500 fully loaded | High | Costing more than it returns below ~$2k ACV |
| Freelancer | $1,500 to $4,000, plus data and sending | Medium | Stopping when they do; optimising for meetings booked |
| Agency (US managed) | $8,000 to $12,000 | Medium | Distance from your product |
| Software, usage-priced | Tens of dollars | Low, bounded by its rules | Doing the wrong thing quickly, at scale |
| You | About a day a week | Highest | Stopping in week three, and the product slipping |
How to actually decide
What is your annual contract value? Under about $2,000, a human cannot pay for itself and the question is really software or you. Over about $20,000, a person is affordable and the judgement is worth buying.
Do you have a message that works yet? If nobody has ever replied positively, hiring someone to send more of an untested message buys you a faster no. Find the message first, on small volume, yourself or with software.
Is your time the constraint or your knowledge? If you know how to do this and cannot find the hours, a freelancer or software both work. If you have never done it, a freelancer needs direction you cannot give, and that is the specific gap an expert agent is for.
Operater exists for the last case: you can build anything and were never taught to sell it, so the tool that asks you to describe the campaign is the wrong tool. The honest caveat is the one above. It removes the clerical four-fifths and hands back the decisions, and if people are seeing your product clearly and not coming back, none of this is your problem to solve yet.
Key takeaways
- Base salary is roughly half the real cost of an in-house SDR. The fully loaded figure is the one to compare against.
- A freelancer is cheaper but brings no data licences, no sending infrastructure and no cover when they are away.
- Doing it yourself has a real price: about a day a week, taken from the only thing you are uniquely good at.
- Below roughly $2,000 of annual contract value per customer, a human doing outbound cannot pay for itself. That threshold decides the answer more than any preference.
- The failure mode of software is different from the failure mode of a person: software does the wrong thing quickly and at scale, a person does less of it and notices sooner.