Most agencies sell you a dream wrapped in complexity. They promise viral reach, domain authority, and consistent engagement across every platform. Then they hand you a six-month contract, no guarantees, and a monthly bill that would make your accountant weep. I have watched this pitch land on founders for years, and the part that bothers me is not the price. It is that nobody can check the work.
I run a social desk, so I have a bias. Read the rest with that in mind. But the test I am about to describe works on me too, and I have had clients run it.
Ask what got published last month, and ask to see it.
The outcome promise is the first thing to distrust
Some of the loudest posts lately say organic social is a myth because every success story had money behind it. Half right. A company with a big round buys ads, hires a PR firm, and then credits its LinkedIn account. You cannot see from outside how much of any case came from funding.
The conclusion that organic does not work skips a step. You cannot know whether it works for you until you have shown up consistently for long enough to measure anything. Most owners never get there. They quit at week three, or they sign a six-month contract for work they cannot inspect and find out in month five that half of it never happened.
The useful question is whether you can verify that someone did what you paid them to do. Verification is available on day one. Results are not.
Count the inputs
Ask for a log. One document, one line per action, every line with a URL. Post published, date, platform. Reply sent, to whom, link to the thread. Comment placed on a news article, link to the comment, still live. Short video cut, with the source file it came from.
If an agency cannot hand you that, the pace they described is not the pace they work at. Counts do not prove impact. They prove the work exists, which is the first thing you are paying for and the thing most often missing.
Then read ten entries at random. Open the replies. Does each one answer what the other person actually said, or is it “Great point, thanks for sharing”? Would you sign your name under it? A comment on a news thread should add a fact or a position a reader of that article would care about. If it could be pasted under any article on any site, it is filler, and filler gets ignored by readers and deleted by moderators.
Faster content does not fix this
Plenty of people now say AI made content creation quicker, and they are right. Creation was never the slow part. The slow part is the Wednesday reply to a prospect’s comment that arrives on Wednesday. Take the popular model: one 45 minute call, thirty posts approved and scheduled, a one page report at month end. It is efficient for the agency. It also means nobody is watching your threads between the call and the report.
Scheduled posts do not answer people. A buyer who comments on your post at 9am and gets nothing back by Friday has learned how you treat buyers.
So measure reply latency. Ask for the median time between someone commenting on your content and a human responding. Ask what happens when a post draws a hard question. Ask who answers it, and whether that person knows your product well enough to answer without pinging you. A report that lands once a month is also a thirty day delay on finding out that nothing happened. If the work stalls on day four, you want to know on day six.
Who holds the keys
Ownership is an accountability question because it decides whether you can leave. Your accounts stay in your name, under your logins, with your ability to revoke access in one minute. If an agency wants to run things from its own handle or hold your credentials in a way you cannot audit, that tells you how they plan the exit.
Ask for the name of the person who writes for you. A name, not a team label. Ask whether you approve drafts before they go out or whether you set standing rules and review after. Either can work. Having no answer cannot.
Contract length is the last piece. A six-month minimum moves the risk onto you. The agency gets paid whether or not the work is any good, and the first real review of that work happens after half your commitment is spent. A flat fee with a thirty day money-back window flips that. The desk has to earn the second month.
The founder who goes quiet has the same problem
Launch adrenaline fades. A founder who posted every day for two weeks and then vanished did not have a distribution plan, they had a mood. Hiring a desk is a way to turn it into a system, but only if you can see the system running. Otherwise you go quiet on the desk too, stop reading what it sends, and a month later wonder why nothing came of it.
Put ten minutes on your calendar once a week. Open the log. Read five replies. Check that the news-site comments are live. That is the whole job on your side.
I wrote about the gap between what owners need and what agencies sell in The Real Work Behind Real Results, which I went into here: https://kolbinah.com/en/item/the-real-work-behind-real-results–4b9da14cd049
Measure outcomes on your own turf
An agency dashboard full of impressions is the agency grading itself. Your own data is harder to flatter. Add a “how did you hear about us” field to every intake form. Ask it on every sales call. Write the answers in a spreadsheet and look at it monthly. When someone says a post brought them in, note which post. Six months of that beats any slide deck the agency hands you.
None of this requires a marketing background. It requires asking for the log, opening the links, and reading what went out under your name. A desk that passes that test is worth the fee. A desk that dodges it is charging you for a feeling.
If you want a human desk running your voice, I am happy to get on a call or buy you a coffee. Start at https://guruvox.com or write to apply@guruvox.com.