PocketScience finds where your site or product loses visitors, prices every gap, delivers the fix, and uncovers what to grow next.
Where your visitors come from
What they do on your site
Watching every visitor
WhoComparison shoppers, mostly from Instagram
IntentClose to buying, checking shipping costs
DoingBrowse 3 products, then add one to the cart
Stuck6 in 10 carts are never paid
The blind spot
The groups they fall into, and what each one wants.
Every click and drop-off, mapped to your funnel.
Every bug, friction point and missed opening, found.
How it works
Four steps. Always running. Nothing to operate.
Your funnel, mapped. Every source, every step, every drop-off.
Bugs, friction and growth openings, found and priced in dollars.
The fix arrives with instructions for Claude Code, Codex or your developer.
Real visitors, before and after. Only the wins that hold up count.
What you get
Every issue. Every opportunity. Delivered like this.
−$18,400 /mo
High confidence · 1,530 sessions
recoverable every month, likely $14k to $22k. We count only the share we expect to win back (about 40%), in an online store doing $200k a month
The problem
The keyboard covers the Pay button on iPhones. 1,530 of 5,350 phone checkouts hit it this month, and 1,450 of those visitors never paid.
Who it affects
Mostly first-time visitors on iPhones, arriving from Instagram and ready to buy.
Why it happens
The checkout page does not scroll when the phone keyboard opens, so the button ends up underneath it.
The fix
A 3-line layout change that keeps the Pay button visible above the keyboard.
Instructions for your agent
No guesswork. Every dollar figure shows its math.
Over the last 30 days, 3,820 of 5,350 phone checkouts were not affected by the bug, and 40% of those visitors paid. We assume the fixed group behaves the same, so we count that share and nothing more.
If we only win back 32%, the figure is $14.7k a month. At 48%, it is $22.1k. Averages vary, so we report it as a range: likely $14k to $22k.
Did it work?
High confidence · 14-day test · 5,350 sessions
+$24,000 /mo
Medium confidence · 60,000 sessions
extra revenue every month, likely $18k to $30k, if about 10% of visitors reach the video. This is a forecast, in an online store doing $200k a month
The opportunity
Visitors who watch the 90-second demo buy at 5.1%. Everyone else buys at 1.0%. Only 2% of the 60,000 monthly sessions ever reach it.
Who it affects
Visitors comparing plans and close to deciding, who leave the pricing page before they reach the video.
Why it is missed
The video sits at the bottom of the pricing page, four scrolls down.
The change
Move the demo to the top of the home page, with a poster frame and one play button.
Instructions for your agent
No guesswork. Every dollar figure shows its math.
Today 2% of 60,000 sessions watch the demo. Moving it to the top should lift that to about 10%, which is 4,800 more viewers. Viewers buy 4.1 points more often than other visitors.
If only 8% reach the video, the figure is $18.0k a month. At 12%, it is $30.0k. It is a forecast, so we report it as a range: likely $18k to $30k.
Did it work?
Medium confidence · 14-day test · 60,000 sessions
−$7,700 /mo
High confidence · 42,000 clicks
of ad spend recovered every month, likely $6k to $9k. We count only the 30% of bounced visitors we expect to keep, on $64,000 of Google Ads spend
The problem
40% of Google Ads clicks leave within 4 seconds. That is 16,800 of 42,000 clicks this month, about $25,500 of spend.
Who it affects
First-time visitors from Google Ads, looking for one specific offer.
Why it happens
The ad promises one specific offer, but the landing page opens with a general headline, so visitors feel they landed in the wrong place.
The fix
Repeat the ad’s offer in the headline and first screen, and pause the 4 keywords that send the most mismatched clicks.
Instructions for your agent
No guesswork. Every dollar figure shows its math.
Pages that repeat the ad’s offer usually keep about 30% of the visitors who would otherwise leave in seconds. We count that share and nothing more.
If we only keep 24%, the figure is $6.1k a month. At 36%, it is $9.2k. Averages vary, so we report it as a range: likely $6k to $9k.
Did it work?
High confidence · 14-day test · 42,000 clicks
Under the hood
Verifying. Evaluating. Customizing. Powered by AI agents that see every visit, with data science behind every number.
Every fix, checked against real visitors.
Every problem and opportunity, valued in dollars.
Tailored to your business, your funnel, your goals.
Funnel agentMapped 4 paths from Instagram, Google Ads, Direct and Email to checkout.
Segment agentNew group found: returning phone visitors from email buy 3x more often.
Issue agentCheckout drop-off on iPhones is up 22% since Tuesday.
Pricing agentPriced the fix at about $18,400 a month recoverable (likely $14k to $22k), with the math shown.
Growth agentOpportunity: demo-video viewers buy 5x more, but only 2% find it. Worth +$24,000 a month.
Verify agentFix shipped. Comparing the next 14 days of real visitors with the 14 before.
Growth
Not just bugs. See where your effort works hardest. Example for an online store, last 30 days: pick a source to see what we’d recommend.
Protect it. Direct is 31% of sessions and your best-converting visitors, at zero cost. Keep the brand pages and the checkout fast.
$0 to act
Works with your coding agent
Paste the instructions into Claude Code or Codex, or hand them to your developer. Or connect one link and your agent picks up new fixes on its own.
Read-only and revocable. Copy yours from Settings → Agent access, and keep it out of your repo.
→ reads each open fix, with the full instructions
→ builds it
→ marks it done, and we check it worked within 14 days
Early access
We’re opening access a few at a time. Tell us about your site or app, and a real person replies within a day.
Prefer email? hey@pocketscience.ai