A note on this page. It replaces a URL that summarised a specific third party's annual report. Those findings are not restated here: the report is not available to verify, the figures change annually, and summarising someone else's research from memory is how numbers get miscited. What follows is how to read this genre, which is useful every year rather than for one edition.
Platform data beats survey data on accuracy
A vendor analysing its own platform is measuring behaviour: calls that happened, emails that sent, deals that closed. That is a real advantage over a survey asking people to recall their connect rate, which produces answers counted at least two different ways.
So take the observational findings seriously. When a report says deals with more contacts engaged close at higher rates, that is measured across a large sample and is probably true.
The structural limit is the population
The sample is companies that bought that product. That is not the market, and it skews in predictable directions:
- Toward larger companies, which buy more tooling.
- Toward better-resourced teams, which is exactly the population whose
numbers you should not use as a baseline.
- Toward the vendor's own strong segments.
None of that is hidden or dishonest. It does mean a benchmark from such a report describes well-resourced teams using that product, which is a specific population rather than a norm.
Separate findings from recommendations
The useful split when reading any of these:
Findings are what the data showed. Generally trustworthy, within the population above.
Recommendations are what the vendor suggests you do about it, and they lead toward the product. Not dishonest, and not independent either.
A report finding that multi-threaded deals close better is measured. The recommendation to adopt the vendor's multi-threading feature does not follow from the data; the finding supports multi-threading, not any particular tool.
Three questions per finding
- Is this correlation or causation? Deals with more contacts close better.
Do more contacts cause closing, or do closing deals attract more contacts? Usually both, and the report rarely separates them.
- What is the base rate? A finding that something doubles a rate is
uninterpretable without the starting rate.
- Does the population resemble mine? If the sample is enterprise and you
are self-serve, direction may transfer and magnitude will not.
What to do with it
Treat findings as hypotheses to test against your own data, which is the only population that describes you. Build the baseline described in SDR benchmarks, then check whether the finding holds for you.
The general method for the category is in sales development research.
Where this sits
This is Measurement in the Tenbound Pipeline Architecture Standard. The recurring failure it guards against is importing an external number as a target rather than as a question, which is how teams conclude they are underperforming against a population they do not belong to.