Published benchmark reports average across companies that share nothing but a software category, which is why every team reading one assumes it sits on the good side of the median. We collect the same numbers from a screened set of operators at genuinely comparable companies, and we publish the sample size next to every range, including when the sample is small.
Timeline
3 to 6 weeks
Pricing
Quoted per scope
Commitment
Per project, no retainer
01/
Who We Ask
Sourcing and screening
+Growth and marketing leaders at companies matched on buyer, price point and motion
+Finance and revenue-operations leads who owned the numbers rather than presented them
+Former operators at direct competitors, on the economics they inherited and left behind
+Agency and fractional leads who can give ranges across several comparable accounts
Every expert is at least six months removed from any company they are asked to discuss. No material non-public information may be requested or shared, calls are monitored, and the audit trail is kept for seven years under the FieldSignal compliance framework.
02/
When To Run It
6 trigger moments
+The board is asking whether your acquisition cost is normal and nobody can answer
+Planning season, and next year's targets are being set off internal history alone
+A channel looks expensive and you have nothing outside the company to compare it against
+Sales cycle length is being blamed on the market with no external reference point
+You are sizing a growth team and guessing at the shape of it
+You are preparing to raise and expect to be asked about efficiency
03/
Questions We Put To Them
6 examples
Q.01
What did you pay to acquire a customer in this channel, fully loaded?
Q.02
What share of leads from this channel reached a real opportunity?
Q.03
How long from first touch to signature, and where did the time actually go?
Q.04
How many people ran this, and which parts did you outsource?
Q.05
What changed in the last year, and was it you or the market?
Q.06
Which of your own numbers would you not trust if someone else quoted them to you?
04/
What You Get
Deliverables
+Ranges for acquisition cost, conversion and cycle length, split by channel and segment
+The sample described honestly: how many operators, how comparable, how recent
+Team shape at your stage: headcount, seniority mix, and what comparable companies outsourced
+Where you sit against each range, and which gaps are worth acting on rather than merely noting
+The anonymised raw responses, so you can check the reasoning instead of taking the median on trust
05/
What We Need From You
Before the kick-off
+Your own numbers, calculated the way you calculate them. We will normalise, but we need the raw definition first
+A clear statement of what counts as comparable to you: buyer, price point, motion, geography
+Willingness to contribute your figures to the anonymised aggregate. Reciprocity is a large part of why operators answer
+Realism about sample size. A tightly defined comparable set is more useful and smaller than a broad one
06/
Benchmarks To Distrust
Why most published numbers mislead
06.1
Vendor-published category averages
Collected from the vendor's own customers and framed to make the vendor's category look efficient. Directionally interesting and structurally biased, and the methodology section usually tells you so if you read it.
06.2
Blended acquisition cost
A single blended figure hides the one channel doing the work and the three being subsidised by it. Ask for the split before acting on it, and be suspicious when the split is not available.
06.3
Any range without a sample size
A benchmark that does not say how many companies it rests on is not a benchmark. We publish ours, including the times it is twelve rather than fifty.
06.4
Numbers from a different price point
Acquisition economics scale with contract value. A benchmark drawn from companies charging ten times your price is describing a different business with the same job title.
06.5
Medians presented without spread
The distance between the top and bottom quartile is usually more informative than the middle, and it is the first thing dropped when a benchmark is turned into a slide.
06.6
Anything more than two years old
Paid channel costs, platform rules and buyer behaviour have all moved sharply since 2023. An old benchmark is a historical record rather than a target.
07/
A Worked Example
Across our published work
07.0
What nine channel tests showed in aggregate
The situation
Across the nine channel-discovery engagements published on this site, the same channels came up repeatedly against different buyers, price points and categories.
What we did
We looked at how each channel performed across all nine, and at what separated the wins from the losses rather than at the headline result.
What came back
Paid search won outright twice and lost outright four times, and the split was always whether category search demand already existed. Paid social carried the consumer engagements and lost every considered B2B one. Outbound won or shared the win in seven. You should not apply any of that to your own plan blind. It is the kind of pattern a benchmark engagement exists to establish properly, with a defined sample and a stated method.
08/
What This Will Not Tell You
Limits, stated upfront
Every research method has a boundary. Here is where this one sits, so you can decide before you commission it rather than find out at the read-out.
08.1
Self-reported numbers carry self-reporting bias
Operators round in flattering directions, and finance and marketing rarely calculate acquisition cost the same way. We ask how the number was built, publish the definition alongside it, and accept ranges rather than push for precision that does not exist.
08.2
A tight comparable set means a small sample
The more precisely you define comparability, the fewer companies qualify. You can have twelve genuinely comparable respondents or fifty loosely comparable ones. We will tell you which you are getting before we start.
08.3
We cannot benchmark you against a named company
Against the segment, yes. Against a specific named competitor's current figures, no. That runs into the cooling-off rule and, for public companies, into material non-public information.
08.4
A benchmark is not a target
Being above the median is not automatically a problem and being below it is not automatically fine. What the number is for is starting a better internal argument than the one you are currently having.
09/
Where We Have Run This
2 anonymised studies
GTM-02
AI FASHION IMAGING · CHANNEL TEST
Gen-AI SaaS · fashion e-commerce imaging
An AI fashion-imaging SaaS tested five channels to find which one reached brand and e-commerce buyers.
GTM-03
SELF-STORAGE SAAS · CHANNEL TEST
Vertical SaaS · self-storage operators
A self-storage management SaaS tested five channels to find where its operator buyers actually convert.
10/
Questions
5 answers
01How do you get people to share real numbers?
Anonymity, reciprocity and payment. Responses are anonymised before they reach you, participants receive the aggregate, and operators are compensated for their time. We also accept ranges instead of pushing for false precision, which raises both the response rate and the honesty.
02How large is the sample?
It depends on how narrowly you define comparable. A broad category question can reach fifty or more responses. A tightly defined set of direct comparables might be twelve. We agree the target before starting and report the achieved number either way, including when it lands short of what we hoped.
03Can you benchmark us against named competitors?
Against the segment, yes. Against a named company's current figures, no. That runs into the cooling-off rule and, for public companies, into material non-public information. What we can source is what former operators there experienced at least six months ago.
04How do you handle everyone calculating CAC differently?
We ask how the number was built before we ask what it is: whether it includes salaries, agency fees and tooling, and whether it is blended or paid only. Then we normalise where we can and publish the definition where we cannot. A benchmark without a stated definition is close to useless.
05Do we have to share our own numbers?
It is not mandatory, but it is a large part of why operators agree to answer. Contributing to the anonymised aggregate is the norm and it improves what you get back.
11/
Related Engagements
3 suggestions
03
CHANNEL TEARDOWNS
Timeline · 2 to 4 weeks
One channel, taken apart by the operators who ran it. Economics, failure modes, preconditions.
02
CHANNEL DISCOVERY
Timeline · 5 to 8 weeks
Five channels tested against one qualified bar. You get a ranking, the reasoning, and a budget plan.
01
GROWTH OPERATOR INTERVIEWS
Timeline · 3 to 5 days to first call
Calls with the VPs of Marketing and Heads of Growth who already ran your motion somewhere else.
Brief us on GTM benchmarks.
Response under 4 hours, Mon–Fri. First call in 3 to 5 business days.