Compare cold email sequence steps and rewrite the losers

Build me a sequence copy lab for my Instantly cold email campaigns. I want to know which individual emails inside my sequences are dead weight, and Instantly's own reporting makes that impossible to compare across campaigns, so today I paste results into spreadsheets to work out which follow-up is worth keeping.

The main screen has a control bar at the top: a multi-select campaign picker populated from List Campaigns, a date range picker, and a minimum sends field that defaults to 50. When I choose campaigns and a range, load step level numbers for each selected campaign from Get Campaign Steps Analytics, and load the actual sequence copy for each campaign from Get Campaign. Join the two by step position so every row carries both the numbers and the words.

Render one row per step in a single table: campaign name, step number, the subject line, the first couple of lines of the body with a click to expand the full copy, then sent, opened, replied, open rate and reply rate. Default the grouping to step position, so step three of one campaign sits directly next to step three of another, and give me a toggle to group by campaign instead. Every column is sortable and the default sort is reply rate descending.

For each selected campaign also pull its overall performance for the same date range from Get Campaign Analytics, and use that as the campaign's own baseline reply rate. Flag a step as a rewrite candidate when its reply rate is below its campaign average and it has sent at least the minimum sends threshold. Show the flag as a badge on the row and a count of candidates at the top of the table. Steps under the threshold render muted and are never flagged, so tiny samples do not get called out.

Each row has a Rewrite this step button that kicks off a background agent. The agent reads three things: the underperforming step's current subject and body from Get Campaign, the top performing steps across my other campaigns as reference material, which it finds from the step analytics I already have plus Get Campaign for their copy, and the campaign's own settings from Get Campaign such as sending schedule, daily limit, tracking options and audience, so the replacement matches how that campaign actually runs. It then drafts exactly two replacement variants, each with a subject line, a body, and a short written reasoning explaining what it changed and which high performing step it borrowed the pattern from.

The agent writes its output back into the app as a pending proposal, and the row shows a proposal ready state. Opening the proposal gives me a side by side view: the original step on the left and the two variants beside it, each with its reasoning underneath. I can pick one variant and approve it, or discard the whole proposal. Only on approval does the app call Update Campaign to write that one variant into the campaign sequence at that step position, leaving every other step and every campaign setting untouched.

Two rules matter more than anything else. Never auto apply a copy change: no call to Update Campaign may ever happen without me explicitly choosing a variant and approving it, and show a confirmation naming the campaign and step number before the write goes through, because campaign edits are live and affect sends already in flight. And always respect the minimum sends exclusion so small steps are never flagged.

Persist pending proposals so they survive a reload, and keep a history of applied rewrites recording the campaign, step number, date, the old copy and the new copy, shown on a second tab. From that history I want to be able to push the previous wording back through Update Campaign if a rewrite underperforms. Also remember my last selected campaigns, date range and minimum sends threshold per user.