Digital PR for SaaS

The SaaS statistics circulating right now are mostly unverifiable. Pick any widely quoted figure about content ROI or churn, follow the citation chain back three or four hops, and you land on a roundup post citing another roundup post, with no sample size, no collection window and no date. Nobody checked. Everyone repeated it.

That is the opening. If your study ships with a real method attached, you are competing against sources that cannot survive a single click of scrutiny, and journalists and answer engines both reward the one that can.

Find a dataset inside the product that nobody else can produce

Start with the question your product is uniquely positioned to answer, then check that your data can answer it honestly. Those are two different tests and most teams only run the first.

Gong built a content function on aggregated sales call analysis because no other company had that corpus. Ramp publishes corporate card spend data showing how fast companies adopt new software categories, and trade press covers it because there is no alternative source. Vanta reports on compliance and trust posture from its own customer base. Each of those is a number a competitor cannot contradict, because the competitor does not have the data.

Run your candidate dataset through four questions before you spend anything:

• Can you aggregate it so no individual customer is identifiable, and does your terms of service permit aggregate reporting?
• Can you state the population in one sentence that a sceptical reader would accept?
• Does the finding matter to someone who will never buy your product?
• Will the number be different next year, so the study can be rerun?

That last question is the one people skip. A study you can rerun annually becomes a reference series, and reference series get cited far more than one-offs because they establish a trend. The original research program playbook covers the operating cadence for running one every quarter without burning out an analyst.

Design the study so the number survives scrutiny

Two kinds of study work for software companies, and they have different failure modes.

Product data studies are cheap to field and expensive to defend. The cost is analyst time, usually two to four weeks, and the risk is selection bias. Handle it by defining the population tightly and publishing the exclusion rules: which accounts you dropped, why, and how many.

Commissioned surveys are expensive to field and easy to defend. A panel vendor will recruit 400 qualified B2B respondents for roughly 6,000 to 18,000 dollars depending on how narrow the screening criteria are. Ask for revenue-band and job-title quotas in writing, and ask what the incidence rate was, because a low incidence rate means the panel struggled to find your audience and the sample may be thin in ways the topline hides.

The segmentation decision is where most SaaS studies lose their value. An undifferentiated figure like ‘content marketing returns 748 percent ROI’ is useless for planning because it does not say at what ACV, what motion or what company stage. Segment your findings by ARR band and by product-led versus sales-led motion, and you produce the only form of the number anyone can act on. That specificity is also what makes the study quotable in a dozen different articles instead of one.

Build the on-page asset so the finding is extractable

The page does more work than the pitch. A journalist lands on it for ninety seconds and decides whether to cite you. An answer engine reads it and decides whether the claim is attributable. Both want the same things in the same order.

Structure it like this:

1. The headline finding as a single sentence, above the fold, with the number, the sample and the period. Not a title. A sentence a person could copy and paste.
2. Three to five supporting findings, each as its own short block with its own number.
3. A chart per finding, as an image with descriptive alt text that restates the number, because alt text is how the chart gets read by anything that is not a human eye.
4. A methodology block: population definition, sample size, collection window, screening criteria, exclusions and known limitations.
5. A downloadable dataset, CSV or spreadsheet, with no form in front of it.
6. A citation block giving the exact wording and the canonical URL to use when quoting the study.

That last item costs ten minutes and changes behaviour. Give people the sentence you want repeated and a lot of them will repeat it verbatim, which keeps your attribution attached to the number as it travels.

Ahrefs is the clearest example of this discipline applied consistently. Their studies publish the dataset scale, the extraction method and the caveats in the body of the post, and the result is that other people’s articles cite the Ahrefs number rather than a competitor’s. We break down how that habit compounds across their whole library in the Ahrefs product-led SEO teardown.

Pitch structure and embargo timing that gets replies

Twenty named journalists beat two hundred addresses from a media database. That is not a moral position, it is a reply-rate observation: targeted pitches to writers who have covered the exact topic in the last six months return replies at several times the rate of untargeted sends, and the coverage that results is in publications your buyers read.

The pitch itself is four short paragraphs.

• Line one states the finding, the sample and the period. ‘Across 12,400 anonymised B2B SaaS accounts between January and December 2025, median trial-to-paid conversion fell from 18 percent to 14 percent.’
• Line two says why it matters to that writer’s specific beat, referencing a piece they wrote.
• Line three offers the embargo date, the dataset and an interview with the named analyst.
• Line four is the method link and your phone number.

No attachments. No PDF. No ‘hope this finds you well’. The subject line carries the number, not your company name.

Outlet type	Embargo lead time	Honours embargoes	What they want
Trade publications	5 to 7 working days	Usually yes	Exclusive angle plus an interview
Tier one business press	7 to 10 working days	Yes, if formally agreed	National relevance and a named expert
Independent newsletters	Same day, no embargo	Often not	The raw data and a chart they can reuse
Analysts and consultants	Same day	Not applicable	Segment-level data to cite in their own work
Podcasts	2 to 4 weeks	Not applicable	The analyst as a guest, not the report

Timing detail worth respecting: do not launch a study in the week a big funding round or acquisition is announced in your category, and do not launch in the last two weeks of December. Beyond that, the day matters less than the runway you gave the writer to ask a follow-up question.

Send me the sample size in the first line or I will not open the attachment.

That line is a composite, assembled from the same complaint stated three different ways by editors covering software. It is the most consistent feedback in the category and the easiest thing to fix.

Measure citations, not placements

Placement count is the metric agencies report because it goes up when they pitch aggregator sites. Replace it with four measures that track whether the number entered circulation.

The third row is new and it is where the category has no shared practice. Build a fixed list of thirty to fifty questions a buyer would ask an assistant, run them monthly across ChatGPT, Perplexity, Gemini and Google AI Mode, and record which domains get cited. Keep the prompt set frozen so the comparison means something, and note the model version each time because an update can move results without anything changing on your site. Our SaaS AI citation benchmark publishes the method and the quarterly results if you would rather compare against a running dataset than build one cold.

Two honest failure modes. First, unlinked mentions usually outnumber linked ones by two to four times, and chasing every one with a link request wastes more time than it recovers. Convert the top ten by domain and let the rest sit, because an unlinked mention still feeds entity recognition and still gets picked up by answer engines. Second, roughly half of the studies you run will underperform. Two in four is the realistic base rate, and the ones that fail usually fail because the finding only mattered to people already using the product.

Connecting a citation to pipeline is the part nobody has solved cleanly, and you should say so internally rather than inventing a model. What works in practice is a self-reported attribution field on the demo form asking how the buyer first heard of you, checked quarterly against the studies you published. When ‘saw your data on churn’ starts appearing in that field, the program is working, and that single free-text answer is more reliable than any last-touch report on a channel where the click often never happens. Pair it with the organic traffic forecast calculator to keep the ranking half of the return visible too.

What this costs and who has to do it

A single campaign runs 12,000 to 30,000 dollars all in, and the split depends on the source: a fielded survey or your own product data.

• Panel survey fielding: 6,000 to 18,000 dollars, skipped entirely for product-data studies.
• Analyst time: two to four weeks, internal, and this is the line most teams forget to cost.
• Charts and design: 2,000 to 5,000 dollars for eight to twelve chart images with proper alt text.
• Pitching: 3,000 to 8,000 dollars for a freelance ex-journalist per campaign, or an agency sprint.
• Page build and internal linking: roughly a week of a marketer and a developer.

Four campaigns a year at that cost sits at roughly 60,000 to 100,000 dollars, which is comparable to a mid-range agency retainer and leaves you owning four datasets instead of a folder of placements. We put that comparison side by side in the SaaS link building guide, and if you need to defend the spend against a paid alternative, the SaaS SEO ROI calculator will model payback against the same budget in paid search.

Staffing reality: this needs one person who can interrogate a dataset and one person who can write a pitch, and they are rarely the same person. Companies that try to run digital PR out of a generalist content role produce studies nobody covers, because the pitching half never gets the attention it needs.

What to do next

Block two hours this week and list every table in your production database that could be aggregated into a market-level number. Most SaaS companies find two or three candidates and have never looked. Pick the one that answers a question someone outside your customer base would care about, and scope it for next quarter.

If you have no product data worth reporting yet, commission a 400-respondent survey on the single question your category argues about most, and publish the segmented result. That is the cheapest way to enter the citation graph. The lesson on earning links and AI citations covers the pitch templates and the tracking sheet, the SaaS SEO pillar shows how research investment sits against technical and content work, and if your growth plan leans on page volume rather than research, programmatic SEO for SaaS is the other half of the authority equation.