Scaling Link Outreach Without Losing Quality

What does scaling outreach actually mean?

Scaling means getting more placements from the same standard, not lowering the standard to get more sends. A scalable system has four parts: repeatable qualification checks, segmented lists, pitches written for a specific page, and a review cycle that tells you when quality is slipping.

What breaks when you scale on volume?

Three things, in order:

1. Relevance falls. Lists get padded with sites that do not cover your subject, and editors reject those pitches first.
2. Deliverability falls. Mass sending with weak personalisation raises complaints. Google’s guidance for bulk senders (close to 5,000 messages or more a day to personal Gmail accounts) is to keep the user-reported spam rate below 0.1% and prevent it from ever reaching 0.3% or higher (Google (n.d.), email sender guidelines FAQ). Google also says rates above 0.1% already hurt inbox delivery for bulk senders, and rates of 0.3% or higher make a sender ineligible for delivery mitigation.
3. Link quality falls. Placements on weak sites are the ones most likely to disappear. Ahrefs (2022) found that 66.5% of links to a sample of 2,062,173 sites had rotted since January 2013, and 47.7% of the links it counted as lost sat on pages dropped from its index.

We removed a set of figures from the earlier version that claimed specific reply rates for different send sizes and a 70% drop in placements. They had no source. What remains is the mechanism: weak targeting lowers replies, and complaints cap how far you can push volume.

How do you simulate the trade-off?

The simulator compares two sending plans with your own rates, and checks the volume against the Gmail spam-rate limits. Fill it with data from your last campaigns. The defaults are placeholders, not benchmarks.

Outreach funnel simulator

Compare two sending plans using your own rates. The defaults are placeholders, not benchmarks.

Plan A: broad

Emails sent Reply rate (%) Placements per reply (%)

Plan B: targeted

Emails sent Reply rate (%) Placements per reply (%)

Plan A: replies / placements

-

Plan A: placements per 100 emails

-

Plan B: replies / placements

-

Plan B: placements per 100 emails

-

Gmail spam-rate check

Google asks bulk senders to keep the user-reported spam rate below 0.1% and never reach 0.3% or higher. It applies to mail to personal Gmail accounts.

Gmail recipients in the period Spam complaints

Complaint rate

-

Complaints at the 0.10% target / 0.30% ceiling

-

Text version of the simulator.

• Replies = emails x reply rate.
• Placements = replies x placement-per-reply rate.
• Placements per 100 emails = placements / emails x 100.
• For the Gmail check: complaint rate = complaints / Gmail recipients.
• At 2,000 Gmail recipients, 0.10% is 2 complaints and 0.30% is 6.
• With the placeholder rates (a broad plan of 2,000 emails at 1% replies and a targeted plan of 100 emails at 10% replies, both at 20% placement per reply) the broad plan returns 4 placements from 2,000 emails and the targeted plan returns 2 from 100.
• The numbers illustrate the arithmetic only.
• Claim strength: derived, placeholder inputs.

What are our working rules for a scalable programme?

These are operating rules from our own practice, not measured benchmarks. Claim strength: opinion.

Set the goal per campaign. Guest posts, link insertions and resource-page updates need different pitches. A guest post pitch needs a topic outline and samples, and an insertion pitch points to a specific gap on an existing page.

Segment before you write. Group prospects by niche, business model and content theme, so each message speaks to one editorial focus.

Qualify by page, not by domain. Check topical match, real organic visibility, editorial standards, and whether the link would sit naturally in context. Treat DR as a filter, and use the site qualification checklist. For newer domains, sites that are open to contributors can beat large publications that rarely accept outside work.

Write short pitches. Reference a specific page, say why it helps their readers, and ask one easy question. We keep subjects short and bodies to a few sentences. We do not claim a magic word count.

Follow up sparingly. Two or three follow-ups at intervals, each with new context, then close the thread.

Split manual and structured work. Hand-write to top publications and key relationships. Use structured sequences for broader qualified lists.

Start relationships before you need links. Cision’s 2025 data says 85% of journalists name an introductory email, even without a story, as the best way to build a relationship (Cision, 2025). Journalists are not the same audience as blog editors, so treat it as a proxy.

Where does AI help, and where does it not?

Use AI for research: pulling author names, summarising recent articles and flagging relevant topics. Do not let it invent facts or write unreviewed pitches. In Cision’s 2025 survey, 72% of journalists worried about factual errors in AI-generated PR content, which is a reason to have a named person check every claim before sending (Cision, 2025).

How do you protect deliverability?

• Set up SPF, DKIM and DMARC on every sending domain. Gmail’s bulk-sender guidance requires authentication.
• Keep the user-reported spam rate in Google Postmaster Tools below 0.1% and never let it reach 0.3%. Spam rate is calculated daily.
• Send only to verified addresses, limit links in the body and avoid promotional language.
• If complaints rise, cut volume and tighten the list first. Do not rewrite the subject line first.

What should you track beyond link counts?

• Placements per qualified prospect.
• Share of placements live, indexed and correctly linking at 90, 180 and 365 days (see our cohort method).
• Cost per durable link.
• Referral quality and on-page engagement.
• Spam rate and bounce rate for each sending domain.

Warning signs of declining quality: placements that lose organic visibility, links removed or switched to nofollow within months, and links placed on pages that nothing else links to. Audit live placements quarterly, and run one test per cycle so you know what changed the result.