Writing Outbound from Prospect Evidence, Not Templates

Most "personalized" agency outbound is still a template with better merge fields.

First name, company, title, maybe a city. The body could have been sent to anyone in the same industry. Prospects can smell that in one glance. So can your client's brand reputation after a few weeks of it.

Evidence-based writing starts from something observed about this prospect or account, then connects that observation to a relevant ask. Tokens are not evidence. A job post, a product launch, a hiring spike, a public complaint about a workflow, a tech change, or a reply from last quarter can be evidence.

What the research pattern keeps repeating

Public operator writeups in 2025-2026 keep landing on the same split:

• Tight ICP and account-level signals move reply rates more than prettier copy alone
• One real research point beats five cosmetic personalization tokens
• Mail-merge "personalization" (name + company + role) barely moves the needle
• Teams that automate research and keep human judgment on the final ask outperform teams that automate bland paragraphs

Ektie's outbound sequence notes put the median cold reply band for a well-defined ICP around the mid single digits, with loose ICP work falling lower and tight ICP plus account research climbing higher. Their sharper claim is directional: account-level research improves replies by a multiple over pure template sends, while first-name-plus-role tokens do not. Treat third-party ranges as directional, not as a promise for your book of business.

SyncGTM's personalization guidance makes the mechanism plain: personalization works when it references a real signal or trigger. The data layer is what makes that scalable. Without signals, you are decorating a generic message.

Agency case writeups that claim jumps from low-single-digit template replies into higher bands almost always describe the same system: research hooks extracted per prospect, drafts grounded in those hooks, humans still owning judgment. When the research is fake-generic ("saw you are growing fast"), the gains disappear.

Evidence vs flattery

Evidence:

• "You opened two AE roles in Austin last month, which usually means ramp pressure on outbound quality."
• "Your latest release notes shipped SSO, and your pricing page still treats mid-market procurement like self-serve."
• "You replied in March that Q3 was better timing. Q3 is here."

Flattery dressed as research:

• "Loved your recent LinkedIn post."
• "Congrats on all the growth."
• "As a fellow innovator in the space..."

If a stranger could paste the line onto 500 accounts without changing a fact, it is not evidence.

How agencies should staff the write step

Separate the layers.

1. ICP and entry rules decide who deserves a draft at all.
2. Evidence collection pulls a small set of verified signals per prospect.
3. Drafting turns one signal into a hypothesis and an ask in the client's voice.
4. Human review catches wrong identity, tone mismatch, and offers that do not match the workspace.
5. Send windows and compliance decide whether the message may leave.

GTM's motion names this on purpose: write from prospect profile and signals in the client voice, after find and qualify. The write step is not "generate 2,000 emails." It is "produce sendable messages that can be defended if a client asks why this person got this angle."

A practical evidence budget

You do not need a biography. You need one strong hook and one clean ask.

For most agency outbound seats:

• Spend research depth on accounts that cleared qualification
• Cap signals at two or three candidates, then pick one
• Prefer recent, public, or first-party signals over vague enrichment adjectives
• Drop the prospect if you cannot find a non-embarrassing hook and the ICP is only a maybe

Volume without evidence creates work for the shared inbox: confused replies, angry replies, and unsubscribes that should have been preventable at write time.

Templates still have a place

Templates are scaffolds, not finished messages. Keep a structure for opener, bridge, proof, ask. Fill the opener and bridge from evidence. Keep legal and unsubscribe blocks consistent. Do not confuse a scaffold with a campaign that only swaps names.

What to measure instead of vanity personalization scores

• Percent of sends with a logged evidence reference
• Positive reply rate by evidence type (hiring, product, timing callback, trigger event)
• Client escalations about "weird personalization"
• Time from evidence capture to send inside the approved window

If evidence references are empty but "personalization" dashboards look green, your metrics are lying.

CTA

GTM helps agencies find, qualify, and write from prospect evidence inside isolated client workspaces, then send, reply, and book with compliance controls on. Join the waitlist at https://trygtm.com/waitlist.

Sources

• Ektie: What outbound sequences taught us about B2B reply rates
• SyncGTM: Personalize outbound sales emails
• adai.news: Cold outreach personalisation agent case notes
• GTM product notes on write-from-evidence