Onboarding
Welcome aboard.
Here's how we win together.
How we work, week by week. Read it once, then point back anytime.
The clearer our decisions, the faster you see results.
Faster results, less homework.
- You don't tell a mechanic how to fix your car. You bring the goal, we bring the tests, your customers' data does the judging.
- About 1 test in 3-4 wins. Long design reviews mostly polish the ones your data will drop anyway, and they are the top reason this work goes slow.
- The fastest way to find winners is more tests, sooner. A live test is the most honest review a design can get.
- So your week stays light on purpose: one clear call on concepts, a quick look at wireframes, everything else on us. Same cost, more tests.
Your part, in full
- Concepts: you say Yes / No / Maybe on each. This is the one that matters.
- Wireframes: we send them in our updates. Tell us only if something breaks a rule, isn't true, or hurts your brand. Otherwise we run.
Then the data talks
- Losers: a change we did not roll out to everyone. If we had shipped it, you would take that loss every day. We drop it, no meeting needed, and use what we learned on the next test.
- Winners: we walk you through the numbers. You decide when it goes live for good.
A partner, not a test vendor.
- We own the program: what to test, in what order, and why. Drawn from 1,000+ tests across 7-9 figure brands.
- Your brand instinct and ideas make every concept sharper. We drive; your brand sharpens it.
- We check each idea against your data and build the version most likely to win.
- Nothing ships for good until the data earns it. Everything below is the rhythm that turns that into steady growth.
Your first four weeks
Concepts locked and wireframed in week one. First test live by week three. Test #2 right behind it.
Kickoff & nail the concepts
- Meet the team, give us access, let us run.
- We pick up where the last call left off and lock the first concepts.
- You call Yes / No / Maybe on each.
Wireframes out, we build
- Approved concepts become wireframes within days, sent in Slack.
- Pixel-perfect Figma, desktop and mobile, interactive where the test is. What you see is what gets built.
- Not a feedback round. Skim them, flag only a dealbreaker, and we go straight into build.
Build, QA & deeper research
- We build and QA the locked wireframes on a dev copy of your live theme.
- Deeper research starts on what to test next.
- Bugs we spotted on day one get fixed along the way.
[ WEEK 3 · CALL ] Your first test goes live.
Same week, a call: we show you the results dashboard and lock next month's concepts.
See an example results dashboard ↗[ WEEK 4 ] We read the early numbers together and launch test #2. From here, we're in the monthly cycle. That's when winners start stacking on your store. Weeks 1-4 get the first tests live, not a finish line.
Month one picks up where the previous call ended.
- No repeating yourself. First slots fill fast with the tests we already discussed, plus bugs we spot on day one.
- Value right away, while we build the deeper research-backed concepts.
- The concept call is the only approval that gates anything. Fast Yes / No / Maybe answers keep you on the week-three track.
- Cadence: two live calls to start (kickoff, then the week-three launch), every two weeks for the first few months, monthly once we're in rhythm.
The monthly rhythm
One loop, every month. This is the engine behind your growth.
Concepts presented
New mockups, built from your data. Mid-month.
We align on slots
We recommend, you say Yes / No / Maybe, we lock what runs.
Wireframes shared
Sent in our updates. Tell us only if something's a real problem, then we build.
Build & QA
We build it and check nothing breaks.
Test goes live
Two-week minimum, even when early numbers look decided.
Analysis & next roadmap
What won, what was flat, what we killed. Every new concept builds on the last test's data. Your logged ideas feed in too.
Weeks 5-8 can feel quiet. That's the work happening.
- Tests are live for their two weeks. Results aren't in yet.
- Next month's concepts are being researched.
- Quiet weeks are when the data gets made.
- Want to see what's running? Your roadmap link always shows it.
How slots work
A slot = one test idea, taken all the way: research, design, build, QA, launch, results explained. We pick what fills each one together. A slot can also be a bigger build.
What a slot gets you
- One idea, start to finish. No extra charge for extra versions.
- Our experience on what to test, and in what order.
- Clear results, and a lesson for the next test.
How to think about it
- Slots are limited. The biggest opportunities go first.
- Not every idea needs a slot. Easy fixes we just do.
- The goal: a long list of ideas, so we're never short on what's next.
Finding a winner can take more than one try.
- About 1 test in 3-4 beats the original. Our job is to beat that average.
- Some parts of your site need two or three tries before one wins.
- A flat or losing test isn't wasted. You paid for the answer, and it shapes the next version.
- Winners stay up for good and add up. The score that counts is the total, not the last test.
From a safe build to a clean finish.
Every test follows the same path. Your live site stays safe the whole way.
From build to live
- We never swap your theme. We add hidden code to it, or add changes through Intelligems. The original stays exactly as it is today.
- We build and test on a copy of your live theme, never on live.
- Before each launch we update that copy from your live site. The test is your current site plus our changes, so your latest fixes are always in.
- You get a preview link first. Switch between the original and each variation yourself.
- We only push the files our test touched. Nothing else moves.
- We launch Mon-Thu, so someone is watching. No Friday launches.
While it runs, and after
- Each variation runs on its own. If one does badly, we can switch it off without touching the rest.
- Winner: we walk you through the numbers. You raise anything that bothers you, then we put it live for good. Your call.
- Loser: we remove the test code and clean up. Your theme goes back to normal. That variation would have cost you money at full traffic. We use what we learned on the next test.
Asked on every kickoff
Three questions every new client asks. Here are the answers up front.
Already running your own tests?
- Keep them running. On day one we look at what's live and what's next, so nothing clashes.
- Your own tests stay yours. We advise, check, and help fix them on a call. No slot used.
- Want us to go deeper on one? We take it into a slot and add variations.
How do we measure success?
- Revenue per visitor first: how many people buy and how much they spend, in one number.
- Conversion rate on its own moves with ads, the season, and where your traffic comes from. No site change controls those.
- We push for an early win. It really adds up around month three, when winners stack up and each test builds on the last one.
Is this a 30-day project?
- Testing takes time, and not all tests win. That's how you find what actually lifts revenue, instead of shipping guesses and calling the month done.
- Insights, winners, and avoided losers compound. Each test makes the next one sharper, and winners stay on your store. Those are the things that move the business.
- A loser is not wasted spend. It is a change we caught before it hit every visitor. We call it in the same update we end it, with the gap vs the original.
- A 1-month sprint that promises every test is a big winner sets the wrong score. The program is built so the total over time is what grows the business.
Your ideas don't get lost. They get scheduled.
Send ideas anytime. Here's how we keep them safe without breaking a running test.
The Idea Vault
- Every idea you send, we write down in the Idea Vault.
- We check each one against your data, fast.
- The strongest ones go first on next month's roadmap.
- Urgent and safe? We usually just do it, no slot used.
The lock
- Once concepts are approved, the roadmap is locked.
- The design locks when wireframes go out. A change after that sends the test back to build and testing, so it slips about a week.
- The lock gives each test real time for research, design, and checks.
- Live tests run their full two weeks. No swapping things around.
- New ideas wait for the next roadmap. That's on purpose, not a delay.
A test changed halfway is a test that can't win.
Send an idea mid-month and the answer is always: "Great, we'll write it down, check it against your data, and slot it next."
The habits that make this work
Small things that keep everything clean, fast, and easy to follow.
1. Keep conversations in their thread
- Reply in the same thread as the topic.
- One topic per thread.
- New channel messages for a thread that exists.
- Concept A feedback inside concept B's thread.
2. Feedback happens once, on the concept
- Feedback on the concept itself, in the doc or Figma.
- Yes / No / Maybe, plus one line on why.
- Feedback spread across DMs, docs, and calls.
- A separate doc of your own.
- Design notes on wireframes. That step is a heads-up, not a review.
3. Call concepts fast, then let it run
- Call concepts Yes / No / Maybe, on the call or same day.
- Let live tests run their full two weeks.
- Concepts that sit for a week. Every day late pushes your results back.
- Asking the whole team for opinions, especially as a big group chat on a call.
- Changes while a test is running. Tests need time.
- Deciding from week-one numbers.
Don't overthink the concept call. The test decides, not the approval.
- A Yes isn't a bet that it wins. Nothing stays on your site until the data earns it, so a Yes is low-risk by design.
- Two weeks of real customer data beats weeks of internal debate.
- Not sure? Launch it and let your shoppers decide.
4. Save new ideas for the next roadmap
- Send new ideas anytime. We write them down.
- Tell us the brand things only you know.
- Adding new tests to a month that's already locked.
- Changing plans every week. It ruins tests that are running.
Where everything lives
Three places, always current. Pinned in your channel on day one.
Your live roadmap
One link. Every concept, status, wireframe, and result. Always current.
Concepts & feedback
Mockups you comment on directly. All feedback goes here.
Your Slack channel
Day-to-day. One topic per thread, replies in-thread.
What you can expect - and what we ask
Clear commitments, both ways.
From us
- We push for an early win from day one.
- Concepts built on your data, never guesses.
- Every idea checked against your data and answered, fast.
- Bugs we spot, we tell you about and often fix.
- Wireframes in days, results explained every month, Slack answered fast.
- A roadmap you can check anytime.
- Exited founders, not account reps.
From you
- Feedback in one place, on the concept.
- Reply in-thread, one topic per thread.
- Call concepts Yes / No / Maybe, fast.
- Let tests run their full two weeks.
- Send new ideas anytime, just not into the month that's already running.
- Tell us what you know about your brand.
- Not every test wins. No honest agency tells you otherwise.
- The ones that win stay up for good and add up. We push hard for an early one.
- No agency controls the market or how much people want your product.
- But nobody will give your growth a more careful, steady shot than we will.
Questions? That's what your channel is for.
Drop them in your Slack channel, in a thread of course. Now let's go find your first win.
[ Review your first four weeks → ]Early numbers lie.
Here's how to read them.
A test is a measurement, not a verdict. The first few days are the messiest.
Once you know what early numbers mean, a slow start won't worry you.
Why the first days swing
When the numbers are small, one order changes everything. That's the math, not the result.
The trap
- 1 buyer out of the first 10 visitors = 10%.
- The other variation gets 0 out of 10 = 0%.
- Neither number is real yet. That's far too few people.
What's actually happening
- A zero in one variation early on is almost always noise.
- Variations jump up and down, then settle as more people come through.
- That's small numbers settling, not a losing variation.
- Lots of orders isn't the same as enough time. 300 orders in five days is still only five days.
- Weekday and weekend shoppers act differently. We've seen tests flip after week one.
An early number isn't a result. It's just noise.
Big early swings are normal, even when the order counts look big. We watch the trend over time, not the first few days.
When we call it
Three things we look at together, not a strict rule. Orders and confidence can show up early. Time can't.
Orders per variation
What we like to see before we trust the numbers. Smaller stores run longer, or we read a steady result on less.
Minimum time live
So weekday and weekend shoppers both count. Hitting the order count early doesn't shorten it. Some tests run longer on purpose.
Confidence to beat
How sure we are the winner really beats the original. On smaller stores, a clear steady lead can win below this.
- Why we don't decide early: that's how you ship a "winner" that was really just a lucky weekend.
- We've seen strong week-one leads flip in week two.
- We weigh all three, plus your store's size and how steady the trend is. It's a judgment call, not a formula.
Revenue per visitor. How many people buy and how much they spend, in one number. Conversion rate on its own moves with ads, the season, and where your traffic comes from, so we treat it as backup, not the answer.
Testing tools update from day one: win chances, summaries, forecasts, all built on whatever data is in so far. A confident answer on a few days of data is a snapshot, not the final call. If something really needs action sooner, we'll tell you.
The exception: a test that's tanking.
- The two-week minimum stops us judging a test on noise. It is not a reason to keep losing money.
- If a variation is clearly and badly down, and stays down, we can stop it early.
- First we check the setup, the number of days, and whether the gap holds through a full week, weekend included.
- A confident week-one scoreboard usually isn't enough. The math is sure about a week that isn't finished.
- We watch for this and tell you fast. You can flag it too. We pull the plug together.
How orders get counted
The most common mix-up: it looks like every order went to the original. Here's what's really happening.
How the split works
- Visitors are split about 50/50, and each person stays in their group.
- One group sees the original page. The other is sent to the test page.
- Every order a visitor places counts to their group.
People stay in their group
- Once someone is in the test group, all their orders count to it.
- That's true whether they buy on the test page or come back days later and shop normally.
- So the test group measures everyone who started on the new page, not just that one page.
An early result that looks one-sided usually isn't a tracking bug. It's too few people, plus the grouping above. We check the split is working, then let the numbers build.
Normal vs worth a flag
Most early weirdness is normal. A few things are worth a message.
- A 0 in one variation on day one or two.
- Variations swapping the lead back and forth early on.
- A variation behind before a full week is in, even if the tool sounds sure.
- The dashboard calling a winner or loser in week one.
- Big percentage swings on small daily numbers.
- One variation getting little or no traffic. That means a broken split.
- The wrong page showing, or the test page not loading.
- A clear gap that keeps holding after the first full week.
- Anything that just looks off. We'll check the setup.
We run tests on Intelligems. These are their own rules for reading results. Want the longer version? Intelligems: why early results look different, then stabilize ↗
Numbers look strange? Send it over.
Drop the test in your channel and we'll review the setup and the live data with you. No guessing.
[ Back to the top → ]What disciplined CRO does
A concentrated dose of what is possible. Operators and founders, in their own words.
Why is CRO so powerful? When you 2X your conversion rate, your costs stay nearly the same, so profit soars.
Rand Fishkin@randfish↗Revenue up 500% in two years. Consistent testing works. We deployed around a dozen winners. Compounding growth.
Clayton Ferguson@ecomclay↗Most "CRO wins" online are flashy homepage tests. The stuff that actually moves the needle is boring, repeatable, and compounds.
Julian@JuliansDigital↗The ugliest CRO wins are the highest ROI. +35% CVR killing a discount field. +85% CVR adding trust copy.
Daniel Budai@thedanielbudai↗We added $50,000 this BFCM by showing the price with the coupon already applied. +5% revenue per visitor.
Clayton Ferguson@ecomclay↗A database of over 150 CRO ideas. Generating over $150 million in extra revenue.
Samuel Hess@LessEgoMoreData↗Messaging changes can unlock 10-30% conversion lifts without touching your traffic strategy.
Noah@Noah_rkf↗PSA Shopify founders: we saw an immediate 10% boost to eCVR fixing this. 100% ROI there. Do it.
Clayton Ferguson@ecomclay↗You don't want to guess your best pricing. You need a painted-door test. Most brands leave 10-20% margin on the table.
Julian@JuliansDigital↗We weren't tweaking colors or buttons. We were fixing a fundamental flaw. Your revenue is tied to your customer's emotional trust.
Adebiyi Samuel@UxBeta↗One of the highest-ROI optimizations for any DTC brand: optimizing the navigation menu. Prioritize it.
Osh@oshalchemy↗We're seeing personalizations add 20-30% revenue-per-visitor lifts on certain audiences.
Clayton Ferguson@ecomclay↗Most CRO focuses on product pages. The real lever? The offer. We tested checkout: +6.35% CVR at 99.99% confidence.
Jacob Elbaum@ElbaumJacob↗Your biggest advantage is running with test ideas and launching quickly. You have the authority and you believe in CRO.
Julian@JuliansDigital↗A tiny urgency cue, big results. Added "Only 2 left": +6.4% ARPU, $9,280 extra revenue.
CustomFit@CustomFitAi↗Using Claude to clone Replo pages: +10% lift to performance, -10% CAC.
Clayton Ferguson@ecomclay↗