Feedback Formulation Cheat Sheet

Last updated: May 11, 2026

The Give Feedback button is the fastest way to fine-tune your messaging, but the AI only acts on what you actually tell it. Vague instructions like "make it better" fall back to defaults; precise instructions stick.

This article is a quick reference of the 30 most common adjustments customers make through Persona Feedback, with the exact wording that works.


Three golden rules

One change per submission. Stack multiple changes in separate feedback rounds — each lands more cleanly.

Pair every "don't" with a "do". If you ban a phrase, give the AI an alternative. Otherwise it invents a near-identical variant.

Quote exact wording when it matters. Templates and CTAs are honoured literally — use quotes.


1. Initial Email Adjustments

These adjustments apply to the first email in your sequence. Select the Initial email step in the Give Feedback modal.

Goal

Formula

Example

Shorten the email

The initial email must be no longer than <N> words.

"We want emails about 80 words. These are too long."

Fix the structure

Start with <X>. Then <Y>. Do not include <Z>.

"I want it to hit pain point immediately."

Format the greeting

Always start with "<greeting>". First sentence on a new line.

"The first sentence needs to be on a new line after the receiver's first name."

Change the CTA — exact wording

CTA in the initial email should be: '<exact CTA>'.

"Change the CTA to: 'Worth a 15-min chat next week?'"

Soften the CTA

CTA should be a <low-effort question / share resource>, not a meeting ask.

"Change the call-to-action to a simple question instead of asking for a meeting."

Add a Calendly / booking link

Include this Calendly link in <step>: <URL>.

"Replace the meeting link placeholder with my link: https://calendly.com/.../30min."


2. Follow-up Email Adjustments

Select the specific follow-up step (Email 2, 3, 4…) in the Give Feedback modal.

Goal

Formula

Example

Stop follow-ups repeating themselves

In follow-ups <list>, vary the opener / drop <reference>.

"Email 2, 3, 4 starts the same way with 'Hi {{First Name}}'. Vary the openers."

Make a specific follow-up shorter

For follow-up <N>, keep it <length / purpose>.

"For the second email, keep it short — essentially just an inbox bump email."

Drop a duplicate reference

Remove <reference> from email <N>, you already mentioned it in email <N-1>.

"Remove Lucid case study reference in email 4, you already said that in email 3."


3. Subject Lines

Select the Subject line step in the Give Feedback modal.

Goal

Formula

Example

Set an exact subject line

Subject line = '<exact subject>'.

"Subject line = Faster software delivery with higher quality."

Set rules for subject lines

Subject lines must be ≤ <N> words. Avoid <buzzwords>.

"Avoid clickbait subject lines, max 6 words."


4. LinkedIn Adjustments

LinkedIn-specific feedback must be tied to a LinkedIn step in the modal — otherwise emails inherit the rule.

Goal

Formula

Example

Change the connection note

LinkedIn connection request must be: "<exact template>"

"LinkedIn connection request must be: 'Hi {{First name}}, thought it'd be great to connect.'"

Adjust LinkedIn DM rules

On LinkedIn DMs, <change>. Do not <pattern>.

"End LinkedIn messages with a clear, low-effort CTA — don't end with open questions."

Adjust LinkedIn follow-up DMs

For LinkedIn follow-up DMs, <do / do not> <X>.

"Do not use the same messaging in first and second LinkedIn DMs."


5. Tone, Style and Language

These rules apply across all messages, regardless of step. They reshape the persona's voice.

Goal

Formula

Example

Tighten tone / writing style

Write with a <tone descriptor> tone. <Specific do>. <Specific don't>.

"Make all the messages slightly more provocative."

Set the output language

All output must be in <language> ONLY. Use <register>.

"The language of all emails and LinkedIn messages must be Italian ONLY."

Ban a phrase or pattern

Never use '<phrase>'. Use '<alternative>' instead.

"Avoid using the phrase 'I came across your profile'."

Remove formatting you dislike

Do not use <em-dashes / bullets / blank lines>.

"Make sure to not use em dashes on the emails."

Avoid past-dated references

Never mention dates or years that are now in the past.

"Never mention years that are now in the past."


6. Brand, Content and Social Proof

These rules govern what the AI says about your product, customers, and proof points.

Goal

Formula

Example

Restrict topics (whitelist)

Only speak about <topic A>, <topic B>. Do not mention <off-topic>.

"ONLY speak about signal intelligence."

Define / refine pain points

Lead with this pain: <pain>. Do not mention <off-pain>.

"Tell a better story, give a pain and say how we can solve it."

Narrow personalization signals

Personalize based on <allowed signal>. Do not personalize on <banned signal>.

"Do not focus on any recent events. Personalize based on LinkedIn bios and current title."

Replace or drop a case study

Use <new social proof>. Do not reference <old>.

"Please no longer use the social proof GymShark."

Correct numbers / claims

The correct figure is <X>, not <Y>. Never invent numbers.

"200,000 buyers, not 500,000."

Fix product / brand name

Always refer to <product> as '<exact name>'. Never use '<wrong name>'.

"Never reference RiviaConnect, only RiviaHealth as the product."

Reframe the value narrative

Frame the conversation as <value-first / outcome>, not <X>.

"Always frame Green Impact as adding value, not as avoiding risk."


7. Campaign-Wide Rules

Rules that must hold across every message in the campaign — initial, follow-ups, and LinkedIn together. This is the most-used category in customer feedback.

Custom instruction caveat: <custom_instruction> overrides every other persona section. Use it sparingly — too many or conflicting rules confuse the AI and produce inconsistent output. If you have many related rules, edit the matching persona tab directly instead.

Goal

Formula

Example

Add a campaign-wide rule

Always / Never <do X> when <condition>.

"Always start communication with 'Hi {{First name}},' — first sentence on a new line."

Frame as existing-customer outreach

These are existing customers. The goal is <expansion / value-add>, not net-new.

"These are existing customers — we're reaching out to add value, not prospect."

Anchor to an event

This is outreach for <event, date, venue>. Reference it in <step>. Link: <URL>.

"This is outreach for an upcoming conference: Asembia 2026."

Pin a useful link

When relevant, reference this link: <URL>.

"Include this URL: https://www.questionpro.com/xday/2026-uk/"

Add an FAQ for AI Smart Replies

If a lead asks <question>, answer with: '<exact response>'.

"If they ask about pricing, answer: 'Plans start at...'"


Common Pitfalls

The most frequent reasons feedback doesn't stick:

Too vague. "Make it better" or "rewrite the whole email" — the AI falls back to defaults. Be specific about what and where.

Forgot to select the step. Feedback applied to "all steps" globalises a rule that should have been per-step. Always pick the exact step in the modal.

Banned a phrase without offering a replacement. The AI invents a near-identical variant. Always pair "don't" with "do".

Multiple rules in one feedback. Submit them separately so each lands cleanly in its own tag.

Wrong channel tab. LinkedIn-specific feedback written generically often leaks into emails. Say "On LinkedIn DMs..." or select the LinkedIn step explicitly.

Conflicting rules. Adding a rule that contradicts an existing one produces inconsistent output. If unsure, open Full Prompt Mode and check what's already there.


What Give Feedback Cannot Change

Some elements are configured outside Persona Feedback:


Once you've submitted feedback, click Regenerate sequence to see changes. If the results aren't right, use Undo and try a more specific instruction.

Quick Triage — "Where Will My Feedback Land?"

If you're not sure which tag your rule will end up in, scan this checklist:

  • Tone, voice, register, language, or a phrase/formatting to ban → Context > Mandatory language rule

  • A rule for a specific channel or step → the matching section under Email rules, LinkedIn rules, Phone call rules, or Video rules (e.g. follow-up wording → Email rules > Mandatory follow-up email rules)

  • ICP, seniority, company size, or tech stack → Context > Target audience

  • A buyer problem, pressure, or consequence → Context > Pain points

  • Product or brand name → Context > Product name

  • Positioning, angles to lead with, or capability-to-pain mapping → Context > Values

  • A company, case study, or proof number → Context > Case studies

  • A buyer question and its answer → Context > Frequently asked questions

  • Applies to every message regardless of channel → Context > Custom instruction (use sparingly — overrides everything else)

Related Articles