# Marcus Delgado, Agency Principal at Bolt Outbound — read of Campaign Benchmark, June 8, 2026 > 9 years in B2B sales development, ran SDR teams at two SaaS companies, went independent in 2021. Now running LinkedIn automation campaigns for 14 clients, mostly insurance brokers and staffing shops. Hard stop every day at 3pm to pick up my 7-year-old. --- ## How I got here One of my clients asked me last week why his reply rate was "bad." I told him it wasn't bad, but I had nothing to back that up with. No comp data. Just vibes and memory from past campaigns. I Googled "LinkedIn campaign benchmark by industry" and found exactly nothing useful -- some outdated HubSpot blog post from 2022 and a Lemlist article full of cherry-picked numbers. I kept scrolling and eventually landed here after clicking something from a Sales Connector user group post. ## What I clicked first The hero hit the exact itch: "Compare your campaign response rates, reply time, and cost-per-reply against agencies in your vertical and size band." That's the sentence. That's the whole problem I was trying to solve. I actually read it twice because it was so specific. The "vertical and size band" part is what got me -- I've never seen a benchmark product that didn't just throw a global average at you and call it a day. ## Where I paused The honest disclosure section. Specifically this: "we don't have live customers on this idea yet. We shipped the strategy package; you ship the customer conversations." I sat with that for a minute. That is either the most refreshing thing I've read on a product page in years, or it's a very clever way to sell an idea that doesn't exist yet and put all the risk on me. I genuinely could not decide which one it was. The Fermi math showing year-1 take-home as negative seven thousand dollars is the kind of thing that makes you do a double-take. I can't remember the last time I saw a founder voluntarily put a minus sign in front of their own projected returns. ## What I distrusted "Real-Time Cohort Data -- Industry averages update daily as campaigns run across the Sales Connector network." There's no number here. How many campaigns? How many agencies? Fifty people using SC is not a cohort, it's a group chat. If the dataset is thin, the benchmarks are noise. A benchmark that tells me the average insurance agency gets a 4.1% reply rate when that's calculated from six campaigns is worse than useless -- it's actively misleading. I'd want to know the sample size before I put that data in a client deck. Also: the "Anomaly Alerts" feature is listed with zero detail on what an anomaly is, what the alert mechanism is, or what "top performers unlock new patterns" even means. That last phrase reads like it was written by someone who was tired. ## What would convince me Show me the data backing the benchmarks before I pay anything. Not a screenshot of a graph -- an actual table. Something like: "Insurance vertical, 50-500 employee target companies, n=40 campaigns, median reply rate 3.8%, P25 2.1%, P75 6.2%." If you can show me that kind of output exists and is real, I stop being skeptical immediately. Also: one agency owner saying "I put this in a client deck and kept the account" is worth more than everything else on the page combined. ## What I'd ask in an email reply 1. What's the actual size of the Sales Connector campaign network this is pulling data from? Not the total user count -- the number of campaigns with enough data to calculate a meaningful vertical-specific benchmark right now, today. 2. The scoring page shows "financial upside: 2/10" and negative year-1 take-home. Those are your own numbers. Are you saying this is a feature you'd sell as an add-on inside SC, or a standalone product? Because if it's standalone, I genuinely don't understand the business model at that price. 3. When you say "anomaly alerts" -- is that email, in-app, or a thing that notifies the agency when a client campaign drifts? That distinction matters a lot to how I'd position it. ## Verdict: curious-enough-to-reply The brutal self-disclosure almost does too much work here -- I'm replying mostly because I want to understand whether the underlying data is real, not because I'm sold on the product. But "almost too honest" is a far better problem to have than the usual founder theater. If the sample sizes hold up, this solves a problem I couldn't solve with a Google search. --- *Memo by skeptic persona, generated 2026-06-08. Studio breaks own self-grading loop.*