SETHJACOBS

AgileThought · Client: New Horizons · Education · 5-month engagement

The conversion drop was an information architecture bug.

How does a platform “upgrade” erase a third of your sales?

Cart conversions fell 32% after a platform “enhancement,” and telemetry blamed search. I audited the IA against L.A.T.C.H. and found the real defect: price, location, format, and schedule (the four things career-changers buy on) were buried or missing. I rebuilt discovery around them.

Role

Head of User Experience, AgileThought · designed hands-on, led the practice and the vision

Company

AgileThought · Client: New Horizons

Timeline

5 Months

Status

Delivered to client, 2018

Tools

Figma, InVision, UserTesting.com

Impact

32% of conversions recovered

01 · The problem

Telemetry blamed Search. Telemetry was wrong.

Cart conversions dropped 32% after a round of platform enhancements. Telemetry pointed at Search; my brief was to find what broke.

The short version

  • I found two of the five ways to organize information in use. The missing three were what buyers shop by.
  • No research budget, so I built the personas and recruited the testers myself.
  • All 15 participants validated the filter-first redesign. Surfacing price was the biggest single fix.
Catalog and search: price, location, format and schedule as first-class filters

The redesigned catalog. Category, credential, delivery format, brand, level, price, schedule and location surfaced as first-class filters, which is the four missing dimensions from the audit put back where a buyer looks first.

02 · The diagnosis

Two of five

Rather than take the telemetry at its word, I put together a research aggregate: a heuristic audit, interviews, empathy mapping, shadowing, and then checked the site against L.A.T.C.H.: Location, Alphabet, Time, Category, Hierarchy. It had Category and Alphabet. You couldn't browse by where a course ran, when it ran, or how it laddered up to a credential, and those three things are almost the entire basis on which a career-changer decides.

03 · Who I built

Three personas, no budget

THREE PERSONAS I BUILT WITHOUT A RESEARCH BUDGETMVMilitary VeteranLARGEST SEGMENT72%“Will my benefits cover this, and is it near base?”GOALSUse the benefit before it expiresTrain close to homeLand an IT job, not a certificateFRUSTRATIONSPrice is nowhere on the pageNo way to browse by locationPGPostgraduateCREDENTIAL SEEKER“Does this ladder into a certification or not?”GOALSSee how courses stack toward a credentialJudge whether it's worth the timeFRUSTRATIONSNo hierarchy: every course looks standaloneMAMiddle-Aged AdultCAREER CHANGER“I work days. Show me the night classes.”GOALSRetrain around a full-time jobKnow the schedule before committingFRUSTRATIONSSchedule and format buried in a detail pageEach persona's blocker maps to one missing organizing principle: Location, Hierarchy, Time. The audit and the personas told the same story from two directions.
The client wouldn't fund personas, so I built them from the data that already existed and recruited community members to test with. Each of the three turned out to be blocked by a different missing principle, which was the moment the audit stopped being theoretical.

04 · Empathy

Inside the veteran's decision

EMPATHY MAP · MILITARY VETERAN · 72% OF USERSSAYS“Will the GI Bill cover the whole thing?”“Is there a campus near the base, or is this allonline?”“Which of these actually gets hired?”THINKSI have a benefit window and I can't waste it on thewrong cert.Everyone says IT is hiring. Nobody says where tostart.If the price isn't listed, there's a reason.DOESOpens five schools in five tabs and compares byhand.Calls the campus to ask what it costs.Adds a course to the cart, then abandons it.FEELSUrgency about the benefit clock.Skeptical of marketing claims and stock photos.Frustrated enough by hidden pricing to leave.MVVETERANBuilt from interviews, shadowing, and the heuristic aggregate I called The Dissectional, the research the client declined to fund.
Seventy-two percent of users, a benefits clock running down, and a checkout they walk away from. This map is what finally convinced the client that search wasn't the thing that broke.

05 · The audit

I scored the site against all five

FIVE WAYS TO ORGANIZE ANYTHING · WHAT THE SITE ACTUALLY HADLLocationMISSINGBrowse by campusAAlphabetPRESENTA–Z course listTTimeMISSINGWhen does it run?CCategoryPRESENTBrowse by subjectHHierarchyMISSINGDoes it ladder to a cert?Two of five. The three that were missing are the three career-changers shop by.Telemetry said Search was broken. Search was fine. There was nothing to search on.
I scored the site against all five organizing principles and put this in front of the client. It ended the argument about search in about ten minutes, which no amount of describing it had managed.

06 · The audit

Diagnosing a broken IA

The Dissectional

The Dissectional

A heuristic analysis of the current site was performed alongside user interviews, empathy mapping, and shadowing. Findings were aggregated and weighted by prevalence to prioritize ideation decisions.

L.A.T.C.H. Method

L.A.T.C.H. Method

Applied the L.A.T.C.H. framework to audit the existing site's information architecture. Category and Alphabet were the two organizing principles users already engaged with. Location, Time, and Hierarchy were absent.

There was no budget for a formal study, so I ran KLM-GOMS against the recorded sessions we already had. Rigor turned out to be a resourcefulness problem more than a budget one.
The operating principle of a no-budget engagement

07 · Resourcefulness

Research on a $0 budget

The client wouldn't fund personas or testers. So I built three personas out of the data that already existed (Military Veteran at 72% of users, Postgraduate, Middle-Aged Adult), recruited community members to sit with, and ran five participants per persona through iterative prototypes. The budget shaped the method, and I made sure it didn't shape the rigor.

08 · The solution

Surfacing what career-changers buy on

Catalog and search: price, location, format and schedule as first-class filters
Course detail: the outline tab with the live schedule selector
Course detail: the reviews tab, ratings beside the enroll button

The three pages that carry the fix, left to right: catalog and search with price, location, format and schedule as filters; the course detail with its outline and an inline schedule selector showing live class availability; and the reviews tab with the aggregate rating breakdown next to the same selector.

Curriculum and live schedule on one page. Ratings sit next to the enroll button, which is where somebody wants the proof.

09 · Reflections

Results & takeaways

  • One root cause: price, location, format, and schedule were buried or missing. Surfacing them was the fix.
  • The L.A.T.C.H. audit let me explain the gap in one slide: two of five principles present.
  • I built the personas and recruited the testers with no budget at all. The constraint shaped the method. It didn't get to shape the rigor.
  • All 15 participants validated the filter-first direction.