Business Club Varna2 years, 140+ members, zero ad spend
A countdown clock that reset on every visit: 6 to 9 enrolments a month once it was gone
Halcyon Analytics Institute, a private school in San Diego, was sending 1,600 paid visitors a month to its admissions-call request page, and 40 of them filled in the form. 2 agencies had been paid to fix that and neither had. A year after we replaced the site, 60 a month fill it in from fewer visitors, enrolments are up from 6 to 9 a month, and the ad spend is $1,750 instead of $2,100. The ads were fine the whole time; a countdown clock that restarted from 4:59:32 on every page load was not.
About the client
Halcyon Analytics Institute is a BPPE-approved private school in San Diego: 2 founders, 6 instructors and 3 admissions staff selling one $4,500 12-week Data Analytics certificate in 4 cohorts of 30, every seat sold through an admissions-call request form. Enquiries had stopped growing in early 2023, and 2 agencies had been through since, one adding a countdown clock and the other a pop-up and a "Verified School" badge, without moving the monthly count. Helen Okafor, Director of Admissions, had a complaint neither agency asked about: every call began with the caller suspicious of the school.
Private data-analytics school, one 12-week certificate
San Diego, California
11 people
year
Our analysis of the situation
The ads came first: search terms were what you would want, quality scores sat at 7 and 8, and visitors from Google behaved no better on the request page than visitors from Meta, so whatever was wrong happened after the click. Speed and the form were next, a 2.1-second mobile load and a 5-field form that broke nowhere, and 3 weeks of Microsoft Clarity recordings showed people were not abandoning the form; most never reached it.
A fortnight of tallied first questions on admissions calls came back as "is that true", and 5 recorded sessions showed why: a HurryTimer countdown that restarted from 4:59:32 on every load, for a discount that did not exist, a hard-coded "Only 2 seats left" on a cohort with 11 open, and a "Verified School" PNG linking nowhere that appeared only on the $500 deposit page.
For a state-regulated school the clock was an exposure as well as a leak. We chose the request-page completion rate, with the enquiry-to-enrolment rate as the check that new enquiries matched the old in quality, and since sessions were not the constraint we planned to buy fewer of them.
Supporting research
A 2023 study in Behaviour & Information Technology found that limited-time scarcity cues lower a vendor's perceived benevolence [1], a fair description of what the admissions team was hearing on the phone. The FTC's 2022 report "Bringing Dark Patterns to Light" names perpetual countdown timers specifically, and the Commission sent warning letters over them in 2023 [2].
Our strategy to solve the problem
The first decision was whether to split-test the clock's removal or take it out. The founders wanted the test, because the first agency had told them the clock was "carrying the page"; we asked for the numbers behind that phrase and never got them. At 1,600 sessions a month a test would have run most of a year before it could tell us anything, and the FTC point stood either way. The clock and the pop-up came out on launch day, 21 April 2025, and nothing went in their place.
The second was what to put where "Only 2 seats left" had been: nothing, or the real deadline. Nothing was the safe choice and we nearly took it. But Halcyon has a genuine cutoff, because applications close 2 weeks before each start date so the team can run a skills assessment and a set-up call with every enrolee, and a real deadline with its reason attached does the job the fake one was pretending to do. The line at launch read: "The August cohort starts Monday 4 August 2025. Applications close Monday 21 July so we can complete your skills assessment and set-up call before day one." Admissions updates it the day each cohort closes. Of the 6 changes, this is the one that carried the result.
The third was trust marks: keep the "Verified School" PNG and add more around it, as the second agency had proposed, or cap them and move them. We capped them at 2 and put them on the homepage and the request page, where the decision is made rather than paid for. The 2 are Halcyon's BPPE approval, linking to the Bureau's school lookup and to the School Performance Fact Sheet California requires every approved school to publish, and its listing as a Tableau training partner. Both can be checked in under a minute. In July the founders asked for a third, a Trustpilot widget; it ran 3 weeks beside the other 2, completions stayed inside normal weekly variance, and it came down.
The fourth was the testimonials: rewrite the stock-photo carousel or replace it. We replaced it with 3 named graduates, one for each kind of applicant admissions sees. Marcus Bell changed career and is now a Junior Data Analyst at Harborline Logistics. Dana Ruiz stayed in her role as Reporting Analyst at Coastal Pacific Credit Union and took the course to stop handing her SQL to someone else. Tom Whitaker, Operations Manager at Torrey Outdoor Co., needed to read analysis rather than produce it. Each says what they did before and what changed in their work, nothing about salary or placement rates. The graduates gave written consent and their employers confirmed the roles.
Below them we described what happens on the admissions call: 20 minutes, a skills baseline, a written recommendation within 48 hours, and a plain statement that the recommendation is sometimes "not this course, or not yet."
The last was the ad budget, the money paid to Google and Meta rather than to us. We could have kept $2,100 a month running into a page that now converted, and the founders half expected us to. Instead we cut the Meta prospecting campaign to a third of its budget, moved the rest of Meta to retargeting people who had already reached the request page, and trimmed Google to the course-intent terms that were doing the work. Spend settled at $1,750 a month from June 2025.
What we needed from Halcyon: the real cohort dates and cutoff, refreshed every cohort without fail; sign-off to remove devices they had been told were working; 3 graduates willing to be named; and the instructors' credentials, listed with the year each was earned.
Before and after
Before is March 2025, the last full month on the old site. After is March 2026. Both fall inside the application window for the May cohort. Enrolments are attributed to the month of the enquiry that produced them.
| Metric | Before (Mar 2025) | After (Mar 2026) | Change |
|---|---|---|---|
| Monthly ad spend (Google + Meta) | $2,100 | $1,750 | −$350 (−17%) |
| Sessions to the admissions-call request page | 1,600 | 1,400 | −200 (−12.5%) |
| Completed admissions-call requests | 40 (2.5%) | 60 (4.3%) | +20 (+50%) |
| Enrolments from web enquiries | 6 | 9 | +3 (+50%) |
| Enquiry-to-enrolment rate | 15.0% | 15.0% | flat |
| Cost per completed request | $52.50 | $29.17 | −44% |
| Cost per enrolment | $350.00 | $194.44 | −44% |
The enquiry-to-enrolment rate stayed at 15 percent, which is the point. The extra 20 requests a month close at the same rate as the original 40, which is what we would expect if the site had stopped turning away people who were already interested. 6 changes went live together on one school's traffic, and we cannot say which did how much.
Before, the admissions team took 40 calls a month and spent the first 10 minutes of most of them defending the school against its own website, while the founders paid $2,100 a month to fill cohorts that averaged 22 of 30 seats. Now the team takes 60 calls that start with the course, the May 2026 cohort closed full at 30, and the spend buys a smaller audience that has already shown some interest.
Summary
Halcyon's request page was losing people between the ad and the form: a timer that reset on refresh, a seat count that was wrong, and a badge nobody saw until the deposit page had each given visitors a reason to check up on a school with nothing to hide.
The ads, the page speed and the form had all checked out; a fortnight of tallied first questions and 5 recorded sessions pointed at the page itself.
We removed the devices, published the real cutoff with its reason, moved 2 checkable marks to the pages where people decide, named 3 graduates, and bought less cold traffic.
Enquiries rose from 40 to 60 a month and enrolments from 6 to 9, with the monthly ad budget $350 lighter and the enrolment rate unchanged.
References
- "Running out of time(rs): effects of scarcity cues on perceived vendor benevolence", Behaviour & Information Technology (2023). tandfonline.com
- Federal Trade Commission, "Bringing Dark Patterns to Light" (September 2022). ftc.gov
Final word from the client
The first week after the new site went up I kept waiting for someone to open with 'is that true'. Nobody did. 2 years of that, and it was a clock. I am still a bit annoyed about it.