Be the new customer behind every pop up.
Your audit screenshots the pop up. Obsession’s AI agents follow the brand’s live ad as new customers, sign up through it and take the code to checkout, then show you what happened, with proof. Before a pitch, you get a slide for your teardown. After you win, every test variant is checked past the sign up, every morning it’s live, and the whole funnel every week.
- Ferrow Coffee
- Brindlecup
- Oskerby Roasters
- Tallis & Ash
- From the page behind a live ad
- Its own inbox
- Its own browser, on mobile
- Says it’s AI
- Ferrow CoffeeCode rejected
- BrindlecupNo email
- Oskerby RoastersNo leak
- Tallis & AshOff the ad’s offer
Why nobody sees past the sign up
Your tools see the pop up and the order. Not what happens in between.
The teardown stops at the page
Every pre-sale audit is built from outside: the pages, the heuristics, the speed. None of it shows what came after the sign up.
Your tools wait for a login
Heatmaps, recordings and your testing tool run only after the brand signs. Before the pitch, you see the page from outside and nothing after it.
QA stops at the submit
Your QA checks each variant renders and submits. Waiting for the email and trying the code at checkout, on every variant, every morning a test is live, is a strategist’s whole week. So nobody does it.
Obsession’s AI agents sign up as new customers at every brand, from the page behind its live ads, so you see what happens after the sign up, with proof, before anyone hands you a login.
Before the yes, and after it
It checks the step after the sign up. Your audit still covers the page.
Nothing changes in how you research, test or report. Obsession shows what a new customer gets after the sign up; your team still owns the hypothesis, the test and the call.
- Before the pitchName the brands. Each comes back as a slide for the teardown, with proof.
- After you winEvery variant checked past the sign up every morning it’s live, and the whole funnel every week.
- Where you workSlack, the teardown deck, or a link in the monthly readout.
Step by step
From a prospect’s pop up to a test you can trust.
How the agency in this example finds what happens after Ferrow Coffee’s pop up, wins the account and checks every variant it tests.
Before the pitch
Steps 1 to 4Name the brands
Paste a list, upload a CSV, connect Clay or call the API. Any brand you can name, with no directory to check first.
In this exampleThe agency pastes the 12 brands in November’s pipeline, 4 of them coffee. Ferrow Coffee’s founder has a call booked for Thursday.
- Paste a listNames or websites
- Upload a CSVOne column of websites
- Connect ClayA table and a view
- APISend brands as you go
Pipeline: November
- Ferrow Coffeeferrowcoffee.exampleCoffeeCall Thu
- Brindlecupbrindlecup.exampleCoffeeTo pitch
- Oskerby Roastersoskerbyroasters.exampleCoffeeTo pitch
- Tallis & Ashtallisandash.exampleCoffeeTo pitch
And 8 more from the list you pasted.
Choose where the visit starts
Pick where each visit starts, what the customer does and how long to watch. Each step has its own deadline, so a missing email counts as a leak once its time is up.
In this exampleStart from the page behind each brand’s live ad. Sign up through the pop up on mobile, take the code to checkout and watch for 48 hours. Results to Slack, #new-business.
Each visit starts from
At each brand, the customer
- Signs up through the pop up
- Takes the code to checkout
- Signs up through the footer form
- Opts in to texts
- Leaves a basket
- Places an order
Device
What counts as a leak
- The pop up breaks message match with the ad
- Nothing on screen after the sign up
- No email within 1 hour
- The promised code not in the email
- The code rejected at checkout
Watch
Afterwards
Results
Each brand gets its own new customer
A real inbox, a real mobile number and its own browser, used for that brand only. It opens the page behind the brand’s live ad from the ad’s own link, so no click is paid for, then signs up and shops the way anyone would, saving a screenshot at every step. It says it’s AI when it talks to a person.
In this exampleFerrow Coffee, 10:12 on Monday 2 Nov: opened the page behind its “20% off your first bag” ad, then signed up through the pop up on mobile. The pop up offered 15%.
- Identity
- AI, says so to a person
- Inbox
- ferrowcoffee.5b2c@test.useobsession.com
- Mobile
- +1 303 555 0143
- Browser
- Its own session, on mobile, Denver
- Found 3 live ads in public
- Opened the 20% ad’s page from its link
- No discount in the cart: the link carried no code
- Pop up: 15% off your first bag
- Signed up; “Check your inbox” shown
What happened after the sign up, brand by brand
After 48 hours, every brand has its slide: the offer against the ad, what showed on screen, the minutes to the first email and the code at checkout. A form that takes the email and shows nothing is a leak too.
In this example12 brands done. 4 pop ups broke message match with the ad, 6 sent no email within an hour and 3 codes were rejected at checkout. 6 have 2 or more leaks, Ferrow Coffee among them.
- Pop up breaks message match
4of 12
- No email within 1 hour
6of 12
- Code rejected at checkout
3of 12
- 2 or more leaks
6of 12
| Brand | Offer | First email | Code | Leaks |
|---|---|---|---|---|
| Ferrow Coffee | 15%, ad 20% | 7 h | Rejected | 3 leaks |
| Brindlecup | Matches | none | none | 2 leaks |
| Tallis & Ash | 10%, ad 15% | 4 min | Works | 1 leak |
| Oskerby Roasters | Matches | 2 min | Works | No leak |
12 of 12 done. Brindlecup’s pop up took the email and showed nothing, which counts as a leak too.
What real checks found
93% of stores we shopped on every journey had a leak. 77% of the leaks were silence.
128 stores shopped so far, watching what came by email after a sign up, a browse, a basket and a checkout. Silence is the leak a testing tool can’t see: the email that never came.
- Broken links · Drinks store · New subscriberThe welcome email came, but every product link and button opened a closed development store.
- Wrong email · Fashion store · Basket leftA buyer left a basket, and 2 hours later got a browse email instead.
- Broken email · Cosmetics store · Basket left2 cart emails came, both with no product name and a blank total.
- Cart silent · Natural beauty store · Basket leftA subscriber got the welcome, then left a basket. No cart email in 48 hours.
The pitch
Steps 5 and 6A slide for the pre-call teardown
Each brand’s slide, ready for the teardown and the paid audit after it: the ad, the pop up, the email and the checkout, each screenshot dated, on 1 link anyone can open, including the brand.
In this exampleFerrow Coffee: the ad said 20% off, the pop up 15%. The welcome email came 7 hours later, and FERROW15 was rejected at checkout on a subscription.
- Opened the page behind the 20% ad
- The pop up offered 15%; signed up on mobile, “Check your inbox” shown
- No email after 1 hour
- Welcome email with FERROW15, sent at 17:23 by Ferrow’s email platform, by its own headers
- FERROW15 rejected on a subscription
The prospect beside its rivals
Name its rivals with the list. Each gets its own new customer, so the pitch shows the brand’s sign up next to theirs.
In this exampleWednesday’s pitch prep: Ferrow’s first email came after 7 hours, and the 5 rivals named on Monday averaged 4 minutes. Of the 6, only Ferrow’s code fails on a subscription, and none shows its code on screen.
- First email
7 hrivals 4 min
- Message match with the ad
4of 6 brands
- Code works on a subscription
5of 6 brands
- Code shown on screen after the sign up
0of 6 brands
| Brand | Offer | First email | Code |
|---|---|---|---|
| Ferrow Coffee · Prospect | 15%, ad 20% | 7 h | Rejected |
| Wendlow Coffee · Rival | Matches | 2 min | Works |
| Gorsedale Roasters · Rival | Matches | 6 min | Works |
| Linnet Roast · Rival | Matches | 1 min | Works |
| Mallin & Crane · Rival | 10%, ad 15% | 9 min | Works |
| Abberly Coffee · Rival | Matches | 2 min | Works |
Obsession #new-business
Ferrow Coffee is ready for Thursday: 3 leaks after the sign up, each with proof, beside 5 rivals. The slide is in the teardown.
After you win
Steps 7 to 10Every variant, checked before launch
Give it each variant’s preview link. Before a test goes live, a new customer signs up through every variant on mobile and desktop, waits for the email and takes the code to checkout.
In this exampleFerrow signs, and 3 quick wins ship that week: the pop up matches the ad, the welcome email sends at once and the code works on a subscription. Test 1 shows the code on screen after the sign up. The night before launch, both variants sign up, both emails come within a minute and both codes work.
| Variant | On screen | First email | Code |
|---|---|---|---|
| A · Code by email | “Check your inbox” | 40 s | Works |
| B · Code on screen | “Your code: FERROW20” | 38 s | Works |
Each variant reached from its preview link, on mobile and desktop, the night before launch.
A variant that stops working mid test
While a test is live, a new customer signs up through every variant each morning, so a silent app update is caught that morning, with the proof, before anyone calls a winner.
In this exampleFriday 20 Nov, day 5 of test 1: B is ahead on opt-in rate, not yet significant. After an overnight update to the pop up app, B takes the email and gives no code. Fixed by 09:00, and the 26 hours since the last good check are left out of both arms.
- The morning check: a new customer signs up through A and B on mobile
- A: the code by email in 40 seconds
- B: a blank where the code should be; alert sent with the proof
- B: no email after 1 hour either
- The agency finds the app updated overnight and reconnects B
- Checked again: B shows FERROW20 and emails it in 38 seconds
Obsession #ferrow-coffee
Test 1, variant B: since an overnight app update, the pop up takes the email and gives no code, on screen or by email. Last good check Thursday 07:00. Proof attached.
The whole funnel, every week
Every Monday, a new customer runs the client’s funnel from the page behind its live ads: the pop up, the email and the code at checkout. The check from the pitch stays on record, so the before is there from day 1.
In this exampleFerrow, from the pitch to Cyber Monday: the first email from 7 hours to 38 seconds, and a code that works on a subscription.
| Check | Offer | First email | Code |
|---|---|---|---|
| 2 Nov, the pitch check | 15%, ad 20% | 7 h | Rejected |
| 16 Nov, test 1 live | Matches | 40 s | Works |
| 30 Nov, Cyber Monday | Matches | 38 s | Works |
The first row is the check from the pitch, so the before is on record from day 1.
The month, in 1 readout
Every variant as a customer got it, the breaks caught and the before and after, every value linked to its proof. Ready for the monthly readout.
In this exampleNovember: 3 quick wins shipped, both variants of test 1 checked every morning, 1 break caught and fixed the same morning, none open.
- Quick wins shipped
3
- Variants checked every morning
2
- Caught and fixed the same morning
1
- Leaks open
0
- Fixed
- The pop up matches the ad; the welcome email sends at once; the code works on a subscription
- Caught
- Test 1: B gave no code after an overnight app update, fixed the same morning
- Open
- None
- Against 5 rivals
- The fastest first email of the 6
- Before and after
- 2 Nov: first email after 7 hours, code rejected. 30 Nov: 38 seconds, code works
Once you have the leaks
What you do with the leaks.
The opener
Their own funnel and your first test, in 1 email.
The proof shows what a new customer gets today. Your plan shows what you’d change first.
From Jonah, your agency to Sam Ortega, Ferrow Coffee · Subject Your 20% ad, and a 15% pop up
Hi Sam,
Before Thursday’s call, we followed your “20% off your first bag” ad as a new customer on mobile. The pop up offered 15%, the code came 7 hours later, and it was rejected on a subscription.
I’ve attached the slide and the first test we’d run. Worth walking through on Thursday?
Jonah
Proof: Ferrow Coffee, 2 to 4 NovPlan: after the sign up
Written by your audit lead, from the brand’s proof.
Your plan
Ferrow CoffeePlan
Quick win
Match the pop up to the ad: 20% off your first bag.
Quick win
Send the code at once, and let it work on a subscription.
Test 1
Show the code on screen after the sign up.
Winning clients
Lead the teardown with what happens after the sign up
Name the brands you’re about to pitch; each comes back as a slide, with proof.
Ferrow Coffee: 20% in the ad, 15% in the pop up, a code that fails on a subscription.
Testing
Check every variant past the submit
Each variant’s sign up, email and code, before launch and every morning it’s live.
Test 1, B: ahead on opt-in rate and giving no code, caught that morning.
Paid traffic
Check the page behind every ad
Connect a client’s ad accounts, read only, and every live ad’s page is opened from its link each morning.
A client’s ad sending clicks to a sold out bag, found at 07:04.
Reporting
Put the customer in the readout
Every variant as a customer got it, every value linked to its proof.
Ferrow, November: 3 quick wins, 1 break caught the same morning.
Questions
What CRO agencies ask.
Obsession’s AI agents follow any brand’s live ad as new customers, sign up through the pop up and take the code to checkout, then show you what happened, with proof. Before a pitch, you get a slide for your teardown; after you win, they check every test variant past the sign up, every morning it’s live, and the whole funnel every week.
Because the conversion isn’t finished at the sign up. A code that never comes, or one rejected at checkout, can lose the first order from your highest intent traffic: paid visitors who already gave their email. That’s CVR and revenue per visitor, in the step neither your audit nor your testing tool can see.
Your audit reviews the pop up as it looks. Obsession signs up through it as a new customer from the brand’s own ad, then records what arrives, when, and whether the code works at checkout, with proof the brand can open.
A VA can sign up through 12 pop ups in an afternoon. Obsession also keeps every inbox open for 48 hours, tries each code on a subscription the minute it lands, keeps each email as it arrived with its headers, and does it again every morning a test is live.
3 brands you’re about to pitch and 2 of your clients, with their OK. At each, an AI agent follows a live ad as a new customer, signs up through the pop up, takes the code to checkout and watches the inbox for 48 hours. You get each brand’s slide within 4 days, with proof, in any format.
Yes. A client’s store is checked with the client’s OK, and they install nothing. A brand you’re about to pitch you can check before the call.
No. It reads the brand’s live ads in public and opens each one’s page from the ad’s own link, so it sees what an ad visitor sees and no click is paid for. A brand with no live ads starts from its home page.
Through your testing tool’s preview link for each one. A new customer signs up through every variant before launch and every morning while the test is live, so a silent app update is caught that morning.
Barely. It reaches each variant through its preview link, a few visits a day at most, and every visit carries a parameter you can add to the test’s and the client’s exclusions.
It won’t change a bid. It stops paid clicks that became sign ups from dying at checkout, and with a client’s ad accounts connected, read only, it opens every live ad’s page each morning and flags any that’s sold out, broken or off the ad’s offer.
Anything a new customer should get and doesn’t, by a deadline you set: a pop up that breaks message match with the ad, nothing on screen after the sign up, no email within the hour, the promised code missing, a code rejected at checkout, a link to a closed store.
No. It arrives the way any shopper would, from the brand’s own ad or page. Your testing tool and heatmaps still run inside, after the yes.
If you choose. A code check stops at checkout. A real order goes on a card you fund, never on fake payment details.
Yes. The slide carries your name and logo. The proof link opens on a page anyone can check, under Obsession’s name or, with white label, yours.
Not from us. Obsession shares what a brand did, never who asked, unless you choose to be named. Your list, your settings and your slides stay yours.
Because Black Friday pop ups and codes change every week this month, and nobody signs up through them again. 5 brands, 4 days, nothing to install, and the slides are ready for January, when brands review their agencies.
Up to 1,000 stores in 48 hours, so a whole pipeline or every client store in one go.
Every agent does only what any customer can do, and says it’s AI when it talks to a person. It never uses fake payment details or gets round a CAPTCHA. Any company can opt out at useobsession.com/agents. This is not legal advice.
The agency, the brands in the 10 steps and their .example addresses, the people, the ads, the test, the phone number and every figure on their screens are made up. The 93%, the 128 stores and the 77% are real, the 4 findings come from real checks with every store, brand and product changed, and the sample output is a real run. Clay and Slack are trademarks of their owners, named to show where a list can come from and where results go.
Try it on 5 brands.
Name 3 brands you’re about to pitch and 2 of your clients, with their OK. We’ll follow each one’s live ad as a new customer and send what happened after the sign up within 4 days, with proof.
We keep what you tell us to set up your spot and email you about Obsession. Privacy notice