Next Run Test
When To Use
You keep changing five things at once and learn nothing. The review gave you one candidate change. This answers: how do I try it so that, a few runs from now, I can say whether it was worth keeping?
When Not To Use
If you have not reviewed your own counts yet, run Outreach Review first; a test with no starting point proves nothing. This is not a statistical test, and with small runs it never will be: it is a disciplined way to change one thing.
Inputs
- The one change from your Outreach Review.
- Your counts so far for the thing you will compare against.
- How many runs you are willing to wait before looking.
If you have none of this, I start from the change you describe in one line and mark the plan as a first draft with no starting count.
Approach
One change at a time: plan it, run it, look at what happened, decide. This improvement cycle is widely taught and is described here generically. The discipline is in writing the expectation and the keep-or-drop rule before the runs, because afterwards any result can be explained. With about ten people a run, one run tells you almost nothing, so the plan says how many runs you wait. The failure it prevents: a new subject line, a new reason and a new channel in the same week, and no idea which one mattered.
Workflow
- Ask at most three questions: the one change, what you expect it to do and why, and how many runs you will wait.
- Write the change as one sentence about your own action: "I will [do this] where I used to [do that]". If it takes two sentences, it is two changes; pick one.
- Write the expectation and the reasoning before any run, in your words. This is the line you will read back later.
- Say what you will count, from columns already in your Outreach Log, and what the count was before. Nothing new is tracked about the people.
- List what stays the same during the test: who you choose, the channel, the run size, the warm-first order.
- Set the look date as a number of runs, not a result. You do not peek and stop on a good week.
- Write the rule in advance: what would make you keep the change, what would make you drop it, and what counts as "cannot tell", which with small numbers is the likely answer and still a result.
Output Format
# Next Run Test
The change: I will [do this] where I used to [do that] | Starts with run: [number] | Look after: [number] runs
## What I expect, written before the runs
[Your expectation, and why.]
## What I count
| Count (from the log) | Before the change | After [number] runs |
|---|---|---|
| [replies / first calls, of people written to] | [count of count] | [to fill in] |
## What stays the same
- [who you choose] · [channel] · [run size] · [warm before cold]
## Keep, drop or cannot tell
| Outcome | What would show it (set now) |
|---|---|
| Keep | [your rule] |
| Drop | [your rule] |
| Cannot tell | [your rule] |
## Decision
[Your name] runs the change for [number] runs and decides on [date] to keep it, drop it or call it "cannot tell".
Done When
- The change is one sentence about your own action.
- The expectation and the keep-or-drop rule were written before the first run.
- What is counted already exists as a column in your log.
- The look date is set as a number of runs.
Quality Bar
- One change only; a second idea waits for the next test.
- The test changes what you do, never how much you send: the run stays at about ten.
- No claim of proof: small counts support "seems worth keeping", not more.
- Nothing about the people is tracked, scored or compared.
- You choose the change and judge the result from your own counts; nothing is automated to run the test.
Next
Run reach-run-of-ten (Run of Ten Plan) to plan the next run with the one change in place.
About the makers
This pack is made by 90n (90n.ai): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).
1---2name: reach-next-run-test3description: Sets up a test of one change over your next runs, with the change you will make, what you expect and why, what you will count, how many runs before you look and what would make you keep it or drop it. Use for "run reach-next-run-test", "next run test", "test one change", "improve my outreach", "A/B test my cold email", "what should I change in my outreach", "try a different approach", "experiment with my outreach", part of the Claude Guide for Outreach Pack by 90n.4metadata:5 source: https://skillmd.com/skills/polar-bear-org/reach-next-run-test6---78# Next Run Test910## When To Use11You keep changing five things at once and learn nothing. The review gave you one candidate change. This answers: how do I try it so that, a few runs from now, I can say whether it was worth keeping?1213## When Not To Use14If you have not reviewed your own counts yet, run Outreach Review first; a test with no starting point proves nothing. This is not a statistical test, and with small runs it never will be: it is a disciplined way to change one thing.1516## Inputs17- The one change from your Outreach Review.18- Your counts so far for the thing you will compare against.19- How many runs you are willing to wait before looking.20If you have none of this, I start from the change you describe in one line and mark the plan as a first draft with no starting count.2122## Approach23One change at a time: plan it, run it, look at what happened, decide. This improvement cycle is widely taught and is described here generically. The discipline is in writing the expectation and the keep-or-drop rule before the runs, because afterwards any result can be explained. With about ten people a run, one run tells you almost nothing, so the plan says how many runs you wait. The failure it prevents: a new subject line, a new reason and a new channel in the same week, and no idea which one mattered.2425## Workflow261. Ask at most three questions: the one change, what you expect it to do and why, and how many runs you will wait.272. Write the change as one sentence about your own action: "I will [do this] where I used to [do that]". If it takes two sentences, it is two changes; pick one.283. Write the expectation and the reasoning before any run, in your words. This is the line you will read back later.294. Say what you will count, from columns already in your Outreach Log, and what the count was before. Nothing new is tracked about the people.305. List what stays the same during the test: who you choose, the channel, the run size, the warm-first order.316. Set the look date as a number of runs, not a result. You do not peek and stop on a good week.327. Write the rule in advance: what would make you keep the change, what would make you drop it, and what counts as "cannot tell", which with small numbers is the likely answer and still a result.3334## Output Format35```markdown36# Next Run Test37The change: I will [do this] where I used to [do that] | Starts with run: [number] | Look after: [number] runs38## What I expect, written before the runs39[Your expectation, and why.]40## What I count41| Count (from the log) | Before the change | After [number] runs |42|---|---|---|43| [replies / first calls, of people written to] | [count of count] | [to fill in] |44## What stays the same45- [who you choose] · [channel] · [run size] · [warm before cold]46## Keep, drop or cannot tell47| Outcome | What would show it (set now) |48|---|---|49| Keep | [your rule] |50| Drop | [your rule] |51| Cannot tell | [your rule] |52## Decision53[Your name] runs the change for [number] runs and decides on [date] to keep it, drop it or call it "cannot tell".54```5556## Done When57- The change is one sentence about your own action.58- The expectation and the keep-or-drop rule were written before the first run.59- What is counted already exists as a column in your log.60- The look date is set as a number of runs.6162## Quality Bar63- One change only; a second idea waits for the next test.64- The test changes what you do, never how much you send: the run stays at about ten.65- No claim of proof: small counts support "seems worth keeping", not more.66- Nothing about the people is tracked, scored or compared.67- You choose the change and judge the result from your own counts; nothing is automated to run the test.6869## Next70Run reach-run-of-ten (Run of Ten Plan) to plan the next run with the one change in place.7172## About the makers7374This pack is made by 90n ([90n.ai](https://90n.ai)): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).