Rising Stars Muay Thai

A Sacramento fight promotion that, until January, had no website and no data.
We changed that.

January

→

Today

29,376

Events measured

5,779

Sessions

MS Digital Marketing · Cal Poly Pomona
Jarrod Griffin · Juan De La Cruz · Lydia Chen · Bhamini Raghuram Pillai · Christian Gutierrez-Garcia · Karima Ajahiou
Faculty Advisor: Dr. Anthony Kim

SPEAKER: Jarrod · ~90 sec

“Good evening. We’re the MSDM team working with Rising Stars Muay Thai — Jarrod Griffin, Juan De La Cruz, Lydia Chen, Bhamini Raghuram Pillai, Christian Gutierrez-Garcia, and Karima Ajahiou. All six of us speak tonight.

RSMT is a Sacramento fight promotion — a small team led by owner Lino Enriquez. They run live Muay Thai fight nights roughly every quarter, earning on tickets and pay-per-view, both sold through outside vendors rather than their own site.

Until this January they were running blind to all of it: no website, no analytics, no idea who was finding them. Eight months later that same business has 29,376 events measured across 5,779 sessions — from a standing start of zero. And it isn’t just a site that exists, it’s a site that gets found: RSMT ranks in organic search now, and 13 of those sessions came from AI assistants pointing people here when someone asked about Rising Stars. Tiny number, brand-new surface, no regional competitor on it. I led the analytics build and the client relationship.

Our thesis in one sentence. RSMT had no way to measure its digital impact — and the moment we could see it, the data pointed hard at one channel the business wasn’t investing in at all.

Everything tonight builds to one decision we’re asking the client to make — and Karima lands it at the end.”

→ Transition: “Juan — what the client actually asked us to do.”

Project Objective As set with the client

1 · Build the measurement capabilityNo website. No analytics. No customer data.

→

2 · Grow visibility, then ticket & PPV revenueGet found by more people — then convert them.

RSMT sells nothing on its own site. Tickets → Eventbrite · PPV → Akaizo. A conversion is a click, not a sale.

SPEAKER: Juan · ~75 sec

“Thanks, Jarrod. The objective, as we set it with the client, had two halves — and the order matters.

First: build the measurement capability, because none existed — no website, no analytics, no customer data. Second: use it to grow the promotion’s visibility, and turn that into ticket and pay-per-view revenue. Visibility was the client’s own priority — they wanted to be found by more people, and 5,779 tracked sessions since January is the most direct measure of that.

One correction we made early: our first framing had gym memberships in the revenue model. RSMT doesn’t sell memberships — and it doesn’t sell anything on its own site. Tickets go through Eventbrite, pay-per-view through a third-party provider, and the website’s job is to hand intent off to them. We corrected the objective to match the business, not the other way around.

That’s why every conversion number tonight is a click-through, not a sale — it’s purchase intent delivered to a partner, and it’s an upper bound on what actually sold. Ticket clicks, PPV clicks, and lead-form signups.”

→ Transition: “Jarrod — first, the machine that measured everything you’re about to see.”

Turning the Lights On Implementation

On 2026-01-20, RSMT went from zero measurement to a live event-level analytics pipeline.

Site + GA4 / GTMEvery click and view tracked — live 2026-01-20

→

BigQuery exportRaw event stream, snapshotted to versioned files

→

R + DuckDB modelsEvery number in this deck re-computes from data

5,779 Tracked sessions since launch · Jan 20 – Jul 30, 2026
4,505 users · 29,376 events (GA4 Data API, cumulative)

Why a pipeline, not just GA4’s dashboards? GA4’s interface can’t model conversion odds against page depth — and it expires raw event tables at ~60 days. The pipeline snapshots the data so findings stay reproducible after the source is gone.

Source: GA4/GTM property (live 2026-01-20); BigQuery export; R/ga4_data.R

SPEAKER: Jarrod · ~75 sec

“Everything you’re about to hear rests on infrastructure that didn’t exist eight months ago.

On January 20th, 2026, RSMT went from zero measurement to a live event-level analytics pipeline. That’s the date GA4 and Tag Manager went live, and it’s the earliest timestamp in our data — Lydia will come back to why that date is also our biggest limitation.

What’s running: GTM and GA4 on the site, a BigQuery export streaming event-level data, and an R and DuckDB analysis layer on top. 5,779 tracked sessions since launch — 4,505 users and 29,376 events, through July 30.

Why build a pipeline instead of just using GA4’s dashboards? Because GA4’s interface can’t answer the questions we needed — it won’t tell you the odds of conversion given page depth, and it expires its raw tables at 60 days. The pipeline snapshots the data and makes every number in this deck reproducible from a committed file.”

→ Transition: “So what did the data show? Lydia — who is actually buying?”

Your Buyers Decide in One Sitting Audience Diagnosis

Of 235 people who clicked through to buy, almost all did it on first contact.

90.6% Converted on the same calendar day as their first touchpoint

78.3% Converted within their first session — no return visit needed

Source: GA4 BigQuery event export, data/raw/ga4_events.parquet · row-level window Feb 27 – Apr 23, 2026 (n = 235 converters) · converters only — directional, not causal

SPEAKER: Lydia · ~80 sec

“RSMT’s buyers decide fast. Of 235 people who clicked through to buy, 90.6% did it on the same calendar day they first landed, and 78.3% converted inside that first session — they never needed to come back.

One caution: this is measured among people who converted. It tells us the buying window is short — it does not tell us how the people who didn’t buy behave. We can’t see a decision that never happened.

But it changes where effort goes. If buyers decide in one sitting, nurture and retargeting are second-order — the first visit has to close. So we recommend the site be built to convert on first contact: a visible purchase path, above the fold, on mobile — measured by the home-to-event-page progression rate.”

→ Transition: “Bhamini, what did the channel data show?”

The Money Went Where the Intent Wasn’t Evidence

Paid Social bought the most volume at the lowest engagement. Organic Search brought the most intent and got no investment.

Paid Social 2,006 sessions at 8.5% engagement · Organic Search 585 sessions at 69.7%. The size of the gap is the finding.

Source: GA4 Data API, data/raw/ga4_api_channel_acq.parquet · cumulative Jan 20 – Jun 23, 2026. Unassigned and (other) excluded.

SPEAKER: Bhamini · ~80 sec

“There’s a contradiction sitting in the acquisition data.

Paid social delivered 2,006 sessions — 52% of all traffic — at 8.5% engagement. Organic search delivered 585 sessions, 15% of traffic, at 69.7%. Roughly an 8-fold difference — and the money was going to the low side.

Some of that gap is structural — ads reach people who weren’t looking, so of course they engage less. But the size of it is the finding: paid was buying top-of-funnel volume while the channel bringing people who already wanted something got no investment at all. And worth knowing — most of that paid volume came from one six-day burst around RSMT 14 in January, with a smaller Instagram push in RSMT 15’s event week. It has never been a sustained investment.

RSMT’s ad budget is finite and owner-funded. So we recommend shifting effort toward organic capture, measured by organic entrances and engagement rate per channel.”

→ Transition: “Christian, the plan that came out of this.”

The Plan: One Flywheel, Not Five Campaigns Strategy · RACE

R · REACH

Get found

Organic search + owned social — where intent already lives.

A · ACT

Explore on-site

Homepage holds ~6s; the event page over a minute — move people between them.

C · CONVERT

Buy

Ticket + PPV click-through — 90.6% same-day; one shot.

E · ENGAGE

Bring them back

Post-fight results page + clips + podcast — feeds Reach.

Engage feeds Reach: results content published after a fight becomes the next event’s organic discovery. A paid impression expires in a day; a results page earns traffic indefinitely.

Source: RACE framework (Dolbec). Dwell times: GA4 Pages & Screens, 30-day window ending 2026-04-30 — Appendix A.6.

SPEAKER: Christian · ~80 sec

“Six analytics objectives map onto one plan, structured on RACE — Reach, Act, Convert, Engage.

Reach: organic search and owned social, where intent lives. Act: the homepage holds about six seconds, the event page over a minute — the job is moving people from one to the other. Convert: the buy path — tickets and PPV, which get one same-day shot. Engage: post-fight results content.

And here’s what makes it a system, not a list. Engage feeds back into Reach — results content published after a fight ranks, and becomes the next event’s organic discovery. A paid impression expires in a day; a results page earns traffic indefinitely. That flywheel is why the plan concentrates on organic.”

→ Transition (Christian continues): “And one piece of that plan has already been through the client and shipped.”

Own the Results AO4 · Client-validated

The one campaign element we took to the client, they reshaped, and we shipped.

We proposed SEO bios for every fighter on the card.

The scale: 40+ fighters per event — hundreds of bios a year to write and keep current, for a very small team.

Client reshaped → we shipped Results-first content from clips they already film, plus the podcast.

Podcast live 2026-06-25, confirmed in GA4 · results page within 48h of Aug 1

Why results content is the target: Search Console shows #1 for RSMT’s own name — but impressions without clicks for “who won” and “how to watch.” /watch/: 793 impressions, 4.0% CTR. Measured, uncaptured demand.

Source: Search Console API pull, data/raw_refreshed_20260701/gsc_pages.parquet; client decision record (AO4)

SPEAKER: Christian · ~85 sec

“This is the one campaign element that went to the client, got changed by the client, and shipped.

Our original proposal was SEO bios for every fighter on the card. Do the math on that: forty-plus fighters per event, multiple events a year — hundreds of bios to write and keep current, for a very small team. The client saw that scale problem before we did, and their feedback was correct and it improved the plan.

What they approved instead: results-first content after every event, built from clips they already film, plus the podcast — bios reserved for high-profile matchups later. Same SEO logic, aimed at content that already exists. Task one — the podcast — is live and confirmed in GA4. Task two, the results page, ships within 48 hours of Aug 1.

And Search Console tells us why this is the target. RSMT ranks number one for its own name, but only gets impressions — not clicks — for the searches that matter after a fight: ‘who won,’ ‘what’s his record,’ ‘how to watch.’ The watch page alone: 793 impressions at 4% click-through, average position 5.6. Measured, uncaptured demand — and a results page is what captures it.”

→ Transition: “Jarrod — did any of it work?”

Did It Work? Three Observable Outcomes Effectiveness · KPI

DAILY SESSIONS · EVENT SPIKES ACROSS THE SAME THREE CYCLES

Source: GA4 event timeline; matched 12-day pre-event windows and completed event−14 to event+7 cycles. Purchase intent = tracked ticket/PPV handoff, not a confirmed sale. Observational, not causal.

SPEAKER: Jarrod · ~60 sec

“Here’s the short answer: yes, we see progress — across three different outcomes, not just one.

From left to right, the audience is getting larger, more of that traffic is arriving organically, and the people who reach the site are showing stronger purchase intent. I don’t need you to memorize every label; the point is that all three indicators move in the same direction.

The WFC intent result is still marked pending because that event cycle isn’t complete. We’d rather leave the box open than pretend we have an answer.

The timeline underneath is the reality check. Fight traffic is naturally spiky, so we’re comparing the same pre-event window each time instead of treating one big event day as sustained growth.

This is directional evidence, not proof that our work caused every change, and purchase intent means a ticket or PPV handoff — not a confirmed sale. But compared with where this project started, we can now see a consistent, reviewable pattern.”

The Pattern Held Across Three Cycles Effectiveness · Replication

What is ours
The measurement layer and organic strategy — none of this was visible before.

How to read it
Directional: small windows (n = 231–305) · WFC 194 is a WBC co-promotion.

Source: Matched pre-event windows across three cycles — ibm6950/week9/window_compare.R; complete days only

SPEAKER: Jarrod · ~80 sec

“The clearest version of that pattern is in this chart.

The grey section is every other channel; the red section is organic search. Across the three cycles, the grey barely moves while the red gets larger. In other words, almost all of the additional pre-event visibility came from organic.

We made the comparison as fair as we could: the same twelve days before each Saturday event, using complete days. It still isn’t an experiment. WFC 194 is a larger WBC co-promotion, and that alone may create more search interest. So I would not say, ‘our strategy caused this result.’

What I am comfortable saying is that the pattern repeated. It also isn’t a paid-media artifact; the paid activity does not line up with the lift shown here.

That distinction matters. One event can be noise; three events give the client something worth watching and acting on. The measurement layer now makes the next step straightforward: keep publishing ownable organic content, test the next event cycle, and use the dashboard to see whether the red section keeps growing.”

→ Transition: “With that, Bhamini can show what is live, what is staged, and where the team chose to focus.”

Built Where the Leverage Was What shipped · what’s staged

One measurement layer live · one element shipped ◐ · five plans staged.

✅ BUILT & RUNNING

Measurement & analytics stack · Jarrod GA4 + GTM · BigQuery · Looker dashboard · conversion tracking — live since 2026-01-20.

AO4 — Own-the-results organic · Jarrod ◐ partial Podcast live 2026-06-25 · results page within 48h of Aug 1.

⏸ ANALYZED & STAGED — plan + KPI complete, awaiting one client walkthrough

Plan	Owner
AO1 Website funnel + CTA	Bhamini
AO2 Los Angeles market pilot	Lydia
AO3 Owned-social calendar	Christian
AO5 Immediate-intent capture (QR → email)	Juan
AO6 Engagement-based fighter booking	Karima

One finding steered where we didn’t build: Instagram is already RSMT’s #1 front door — so we instrumented what works and built into the gaps.

Source: Team implementation record; channel mix from ga4_api_channel_acq.parquet

SPEAKER: Bhamini · ~75 sec

“Here’s the honest accounting. We led with measurement, then built marketing where the data pointed — and we deliberately left the social content side alone, because it was already working.

On the left, what’s built and running. The measurement stack — GA4 and Tag Manager, the BigQuery pipeline, the dashboard, conversion tracking. And AO4, the results strategy — partially shipped: the podcast is live, the results page ships this weekend.

On the right, five plans analyzed and staged — each with a named owner and a KPI ready to run, all waiting on one client walkthrough.

And where not to spend: social is already RSMT’s strongest front door — Instagram is the number-one traffic source. So we didn’t rebuild what works; we instrumented it and put the build effort into the gaps: the website funnel and organic search.

[If asked paid vs organic: most of that Instagram volume came from the January paid push around RSMT 14. Organic Instagram is smaller but engages far better — 61.5% engagement rate against paid social’s 8.5%. That contrast is the same argument we’re making about organic search.]

So: one measurement layer live, one element shipped, five plans staged — for an owner running a small team.”

→ Transition: “Juan — what does the client actually keep when we leave?”

What RSMT Keeps When We Leave Client Capability

Two assets the client operates without us — and they outlast every tactic in this deck.

📊 Looker Studio dashboard — live Day-to-day KPIs, self-serve. Before this, “how did the last event do?” meant asking someone. Now it’s a link.

📝 Conversion tracking on both lead forms Fighter Application (/for-fighters/) and Inner Circle newsletter (/contact/). The only two lead paths RSMT has — both now fire tracked events.

Source: RSMT Looker Studio dashboard (client account); GTM conversion tags on /for-fighters/ and /contact/

SPEAKER: Juan · ~75 sec

“Thanks, Bhamini. Two of the things we built are things the client uses without us — and they’re the ones that outlast this project.

First: a Looker Studio dashboard, live, that the client uses for day-to-day KPIs. Before this, answering ‘how did the last event do’ meant asking someone. Now it’s a link.

Second: conversion tracking on both lead-capture forms — the fighter application and the Inner Circle newsletter. Those are the only two lead paths RSMT has, and both now fire tracked events.

This matters because RSMT is owner-operated. A capability the client can run alone is worth more than a campaign we run for them for eight weeks. When this engagement ends, the dashboard keeps answering questions. Measured by: whether the client can self-serve KPIs without contacting us — and they can, today.”

→ Transition: “And Jarrod built one more thing that goes further than a dashboard.”

The Bottleneck Was Never the Analytics Capability · Built & tested

The blocker was publishing to the website — so we removed that step.

✉️ SITE EDITS, BY EMAIL · ~$3.22 · ~10 min per edit

Client emails a sentence“Add the October show” — poster attached

→

Identity check → AI editsReads the date and venue off the poster

→

Live preview emailed backThe change, already on the page

→

Reply YES → publishedNo filenames, no git

🎙️ Podcast → page Episodes the owner appears on become pages optimized for search and AI answer engines — same tracked redirects, same measurement. Nothing new to film or write.

Status: site edits live-tested in production — client walkthrough held until after Aug 1 so the event stays the focus · podcast pages newly live, no performance data yet

Source: mail-poller (private) + rsmt/_docs/AI-PIPELINE.md; test records: HANDOFF.md 2026-07-29 · STAGE-3-HANDOFF.md 2026-07-30

SPEAKER: Jarrod · ~75 sec

“Two more things — the pieces I’d argue matter most for a client like this.

Every recommendation tonight assumes RSMT can act on it. It’s a very small team with no marketing staff. The bottleneck isn’t knowing what to publish — it’s the twenty minutes and the technical confidence a website change takes. So the last thing we built removes that step.

First: the client emails a sentence. ‘Add the October show,’ poster attached. The system checks the sender is really them, an AI agent edits the site — reading the date and venue off the poster — and emails back a live preview of the change already on the page. They reply YES; it publishes. About three dollars and ten minutes an edit, no filenames, no git.

Second: the podcast lane — shipped today. A /podcast/ hub and the WFC 194 episode, drafted automatically from the transcript, optimized for search, with the same tracked redirects feeding the same conversion measurement.

Now the honest part. The site-edit pipeline works — I’ve used it many times for real changes that are live on the site right now. But every request through it has been mine, not the client’s: it works, and it isn’t in the client’s hands yet. That’s deliberate — we’re holding the walkthrough until after Saturday so the event stays the focus. And the podcast pages are new, so there’s no traffic or ranking data on them yet. Capability, not result, labelled that way on purpose.

**These are the first things we’ve built that shorten how long it takes RSMT to act.”

→ Transition: “Lydia — the constraints on all of this.”

What Bounds These Findings Limitations

Weigh these before you act on the recommendation — they change how much confidence each number deserves.

1 · Measurement floor — no website before 2026-01-20; no pre-project baseline, every “before” is within-project.

2 · Attribution — a conversion is a click, not a confirmed sale; rates are lower bounds on intent.

3 · Small samples — 231–305 sessions per window, 235 converters; intervals stay wide.

4 · Observational — nothing randomised; association and timing, not campaign effect.

5 · Expiring history — BigQuery drops raw tables at ~60 days; row-level modelling covers 2026-02-27 onward.

6 · Zero baselines — four of five elements score attainment, not improvement.

7 · Platform limits — cross-device collapse, incognito undercount, social→Direct mis-assignment.

None of this invalidates the direction — treat precise percentages as directional.

Source: Project data-collection record; GA4/BigQuery retention limits — Appendix A.12

SPEAKER: Lydia · ~72 sec

“Weigh these constraints before you act, not after.

The big one: our measurement floor is January 20th, 2026 — the day GA4 went live. RSMT had no website before it, so there’s no pre-project baseline and no way to reconstruct one. Every ‘before and after’ here is inside the project window.

Second, attribution: RSMT hands off to Eventbrite and Akaizo, so a conversion is a click, not a confirmed sale — every rate is a lower bound on intent. Third, samples are small. Fourth, nothing was randomised — association and timing, not cause. Fifth, BigQuery expires its daily tables at ~60 days, so event-level history only reaches back to February 27th; RSMT 14 survives only as channel totals.

None of that invalidates the direction — it means treat the precise percentages as directional.”

→ Transition: “Karima, what we learned.”

Reflection What changed our thinking

EXPECTATION

Value would come from executing six marketing objectives.

EVIDENCE

Bios for 40+ fighters per event couldn’t scale; the highest-intent channel had no investment.

ADJUSTMENT

One measurement layer + one element the client could sustain — reshaped by them.

LEARNING

An implementable plan beats a comprehensive one.

FUTURE ACTION

Instrument before you campaign — sequence by what the client can sustain.

That dashboard will still be answering questions after every tactic in this deck is stale.

SPEAKER: Karima · ~75 sec [Karima’s own revision, 2026-07-30 — kept in her words.]

“What changed our thinking?

When we started this project, we thought the biggest value would come from executing all six marketing objectives. But as the project progressed, two things changed our perspective.

First, the client pointed out that our original idea — creating SEO bios for more than 40 fighters every event — simply wasn’t sustainable for a small team. Second, our own data showed that the highest-intent channel was also the one receiving the least investment.

Because of that, we adjusted our approach. Instead of trying to execute everything, we focused on building one strong measurement system and one campaign the client could realistically maintain long after we’re gone.

The biggest lesson we’ll take away is that, for an owner-operated business, an implementable plan is more valuable than a comprehensive one. Build the measurement first, then invest in the marketing the client can actually sustain.”

→ Transition: “Which brings us to the decision.”

The Decision

DO — Lean into organic website content: results pages, podcast write-ups, fight recaps

Starting with the WFC 194 results page within 48h of Aug 1 · $0 — built from clips and audio you already produce · KPI: organic entrances vs the 138-session pre-event baseline

HOLD — Media spend, until the organic plays have a cycle behind them

If you look to expand to a new market or venue, Los Angeles is the option the data supports — 12.96M reachable within an hour’s drive vs Sacramento’s 3.1M.

DON’T — Rebuild paid social at its old weight

It delivered 52% of sessions at 8.5% engagement, while organic delivered 15% of sessions at 69.7%.

SPEAKER: Karima · ~85 sec · SLOW DOWN — pause between the three [Karima’s own revision, 2026-07-30 — kept in her words. Every figure she wrote reproduces; each is inlined from the same scalar the slide reads, per §1, so her spoken number can never drift from the one behind her.]

“So, after everything we’ve shown tonight, here’s the decision our data supports.

First, lean into organic website content — results pages, podcast write-ups, and fight recaps.

Begin with the WFC 194 results page, publishing it within 48 hours after the August 1 event. It costs nothing because it’s built from content the team is already creating, and we’ll measure success by comparing organic entrances against the current 138-session pre-event baseline. This recommendation has the strongest evidence and the clearest path to purchase intent.

Second, give these organic efforts one full event cycle before making any major marketing changes. Once you have another cycle of data, you’ll be in a much stronger position to decide where future investment should go. And if RSMT does choose to expand into a new market, our analysis supports Los Angeles, with nearly 13 million people within an hour’s drive compared to Sacramento’s 3 million.

And finally, don’t return paid social to its previous level. It generated 52% of sessions but only an 8.5% engagement rate, while organic generated just 15% of sessions with a 69.7% engagement rate. Based on the evidence, organic should remain the priority.”

— STOP HERE. No thank-you slide. Take questions if offered. —

— Q&A PREP (own the answer for your area) —

Q: What should we do first after your project ends? A (Karima): The results page, within 48h of Aug 1. Zero cost, uses clips already filmed, most evidence behind it.

Q: Who maintains the website once you’re gone? A (Jarrod): “That’s the question we built the last piece for. RSMT is a very small team — there is no web person to hand it to. So instead of handing them a process, we removed the step: the client emails a sentence — ‘add the October show,’ poster attached — the sender is identity-checked, an AI agent makes the edit and reads the date and venue off the poster, and emails back a live preview of the change already on the page. They reply YES and it publishes. They never see a filename or git. About three dollars and ten minutes an edit. Honest status, and I’d rather say it than be asked: nine of eleven stages are proven in live end-to-end testing, but every request through it so far has been mine, not the client’s. It’s tested; it isn’t in his hands yet — deliberately, because we’re holding the walkthrough until after Saturday so the event stays the focus. What I’m claiming is the direction, not a result.”

Q: How do you keep an AI-drafted page from publishing something false? A (Jarrod): “Names never come from a transcript — two independent speech-recognition systems rendered our own client’s name four different ways, so names come from the fight card and the client profile already in the repo. An earlier draft carried a bout result, a fighter’s record, and an injury that existed only in the transcript — all three came out, and there are no direct quotes, because auto-captions are paraphrase-grade. The Muay TV episode is attributed and linked, not passed off as our content.”

Q: Who at RSMT will be trained to maintain the tracking and interpret the dashboard? A (Jarrod): “The owner, or nobody — and I want to be straight that that’s the business constraint, not an oversight. RSMT is a small, owner-operated promotion with no marketing staff. There is no analyst seat to train into. That’s why the deliverable is a Looker link and not a training programme: a recommendation this client cannot execute is worth nothing. The dashboard is built so the question ‘how did the last event do’ is a link, not a skill.”

Q: Which recommendation has the strongest evidence? A (Jarrod): The organic pre-event lift — 87→138, Poisson RR 1.59 [1.2–2.1], p<0.001, now tracking consistently in a third cycle. Only claim with a like-for-like reference window.

Q: What did you change after feedback? A (Christian): Per-fighter bios couldn’t scale — 40+ fighters per event, hundreds of bios a year for a very small team. The client caught that and we rebuilt AO4 around results-first content and the podcast. Podcast shipped 6/25.

Q: What is the biggest implementation risk? A (Christian/Jarrod): Post-event turnaround — the results page must ship inside 48h with a small team doing clip capture, and 90.6% of converters act same-day. Slip a week and the organic window has closed.

Q: Which KPI matters most and why? A (Jarrod): Organic entrances to results pages. Leading, free to move, and it compounds — a paid impression expires in 24h, a results page earns traffic indefinitely.

Q: What should we NOT do, despite being tempting? A (Karima): Rebuild paid social at old weight; commit to LA before a second cycle of evidence.

Q: How do you define a conversion — is that an actual ticket sale? A (Jarrod/Bhamini): Click intent to Eventbrite or Akaizo — strongest signal measurable without a post-purchase pixel. All rates are lower bounds. Appendix A.3.

Q: Isn’t 90.6% same-day just proving people who bought, bought fast? A (Lydia/Jarrod): “That’s exactly right, and it’s why we call it directional. What we can conclude: among people who converted, 90.6% did so the same day — the buying window is short. What we cannot prove: that shortening the path causes more people to convert, because we only observe the ones who already did. That’s a design question and it needs an A/B test on the CTA — which is what we’re recommending you measure next cycle.”

Q: Your report says 92.1% same-day but the slide says 90.6% — which is right? A (Jarrod): “Both, for different windows — and we report the window with every figure rather than pooling them. The deck reads the frozen February-to-April event archive, 235 converters, because that window has row-level timestamps that survived BigQuery’s 60-day table expiry. The report reads the full live pull through yesterday, 517 converters. Same conversion definition, different observation windows. Picking one primary measure across both artifacts is the first thing on our list after tonight.”

Q: Why LA over Anaheim? A (Lydia): Higher median income ($91,853 vs $90,893), more census tracts (3,085 vs 3,044), stronger existing social presence. Appendix A.9.

Q: Why logistic regression? A (Jarrod): Binary outcome, interpretable odds ratios, standard for click-behaviour data. McFadden R² = 0.274. Appendix A.4.

Appendix

Supporting detail — available for Q&A
A.1 Data Sources · A.2 Technology Stack · A.3 Conversion Definition
A.4 Regression Full Output · A.5 Channel Performance Table · A.6 Page Funnel
A.7 Last-Touch Before Conversion · A.8 Journey Distribution
A.9 All Six Markets · A.10 All Fighter Rankings · A.11 Traffic Sources · A.12 Limitations

A.1 — Data Sources & Collection

Source	Dataset	Coverage	Key Variables
GA4 (event stream)	`ga4_events.parquet`	RSMT 13 launch → Apr 30, 2026	event_name · page_location · link_url · user_pseudo_id · ga_session_id · device_category · traffic_medium
GA4 Data API	`ga4_api_channel_acq.parquet`	All-time channel-level aggregates	sessionDefaultChannelGroup · sessions · engagementRate
Google BigQuery	Raw streaming export	Full event stream, lossless	Complete GA4 schema, exported via `bigrquery` R package
ACS 2023 Census	60-min isochrone (6 venues)	Census tract 5-yr estimates	population · median_income · pct_age_18–34 · pct_age_35–54 · census_tracts
Instagram (manual)	RSMT 15 promo posts	12 matchup posts	likes · comments · engagement_score = likes + 2×comments

Collection context: Google Tag Manager and GA4 went live 2026-01-20 — the earliest event date in the GA4 export, and the project’s hard measurement floor. RSMT had zero analytics infrastructure and no website before this project, so no pre-instrumentation baseline exists. BigQuery streaming export began at site launch and is continuous; row-level event history is retained ~60 days, so behavioural modelling covers 2026-02-27 onward.

A.2 — Technology Stack

Tracking & Ingestion

Tool	Purpose
Google Tag Manager	Event triggers, custom tag deployment
Google Analytics 4	Behavioral event collection
Google BigQuery	Cloud warehouse + raw export
`bigrquery` (R)	BigQuery API auth & pull

Storage & Query

Tool	Purpose
Apache Parquet + `arrow`	Columnar in-memory format
DuckDB + `DBI`	SQL over Parquet, no server needed

Analysis & Modeling

Package	Use
`dplyr`	Data wrangling (lazy eval over Arrow)
`ggplot2`	Publication-quality charts
`forcats`, `scales`	Factor ordering, number formatting
`stats::glm()`	Logistic regression (base R)

Reporting & Output

Tool	Purpose
Quarto	Reproducible documents & slides
RevealJS	HTML presentation engine
Posit Connect Cloud	Shiny app (AO2 census map)
Excel + Tableau	AO6 Instagram engagement scoring

A.3 — Conversion: How We Measured It

A session is counted as a conversion when it contains at least one of:

Page view → /go/tickets or /go/ppvRSMT redirect slugs. GA4 fires the page_view event before the external redirect fires, capturing the click-through moment. Both ticket and PPV paths are captured.

Outbound click → eventbrite.comLive ticket purchase platform. Tracked via GA4 outbound link event + GTM click trigger on the purchase button element.

Outbound click → akaizosports.livePay-Per-View streaming platform. Same GTM click-tracking mechanism as Eventbrite.

What this captures:

High intentA visitor who clicked to the purchase platform — the strongest measurable signal short of a post-purchase pixel

What this does NOT capture:

Confirmed transactions (no Eventbrite pixel)
Users who typed the URL directly after leaving
In-person or phone ticket sales

Key implication: All conversion rates here are lower bounds. True conversion rate is higher. A post-purchase pixel for RSMT 16 would close this gap and enable confirmed-purchase measurement.

A.4 — Logistic Regression: Full Model Output

Model: glm(conversion_flag ~ session_event_count + page_view_count + device_category + traffic_medium, family = binomial) · McFadden R²: 0.274 · N: 1,433 sessions

Predictor	Odds Ratio	95% CI	p-value
(Intercept)	0.044	0.026 – 0.076	< 0.001	***
session_event_count	0.488	0.426 – 0.56	< 0.001	***
page_view_count	10.306	7.078 – 15.008	< 0.001	***
device_categorymobile	0.778	0.48 – 1.26	0.3076
device_categorytablet	0.000	0 – Inf	0.977
traffic_mediumorganic	1.634	1.154 – 2.314	< 0.01	**
traffic_mediumpaid	0.270	0.166 – 0.439	< 0.001	***
traffic_mediumreferral	0.365	0.116 – 1.144	0.0838
traffic_mediumsocial	1.199	0.326 – 4.406	0.7845

Reference category: desktop device · cpc (paid social) medium. Intercept OR reflects baseline log-odds and is not standalone-interpretable. *** p<0.001 · ** p<0.01 · * p<0.05

A.5 — Channel Performance: Complete Table

All acquisition channels — sessions, % of total traffic, and engagement rate

Channel	Sessions	% of Total	Engagement Rate
Paid Social	2,006	52.2%	8.5%
Direct	1,027	26.7%	54.3%
Organic Search	585	15.2%	69.7%
Organic Social	208	5.4%	61.5%
Referral	17	0.4%	58.8%

Source: GA4 Data API (ga4_api_channel_acq.parquet). Unassigned and (other) excluded. Engagement rate = sessions containing ≥1 engaged event ÷ total sessions.

A.6 — Page Funnel: Every Number

Homepage → Event Page → Ticket Page — full metrics from GA4 Pages & Screens

Page	Views	Users	Avg. Engagement	Observation
Homepage	1,212	841	~6 seconds	Highest traffic; most visitors exit without a clear next action
Event Page	175	—	~1 min 03 sec	~10× longer dwell than homepage — users who arrive here read the card
Ticket Page	123	—	—	Final step before Eventbrite redirect
PPV Page	122	—	—	Parallel purchase path (Akaizo)

Homepage first-step navigation (Path Exploration · 1,028 homepage sessions):

Immediate next page from homepage	Users
PPV page	122
Ticket page	101
Event page	78
Exit / bounce	remaining majority

Source: GA4 Pages & Screens report + Path Exploration tool. 30-day window ending Apr 30, 2026.

A.7 — Last Page Before Conversion (Full)

Page visited immediately before the buy-link click — all categories

Page Category (last before conversion click)	Conversions	% of Total
Home Page	187	82.7
Events Page	20	8.8
Dedicated Tickets/PPV Page	18	8.0
Fighter Profile	1	0.4

The homepage triggers the most final conversions despite being the page with the shortest dwell time. It functions as a closing page: content pages (fighters, fight card) build intent; the homepage must execute the purchase click. A sticky “Buy Tickets” CTA above the fold converts this already-warm traffic with zero additional friction.

A.8 — User Journey: Days & Sessions to Conversion

Converters beyond 14 days or 5 sessions represent a small tail and are excluded here for readability.

A.9 — All Six Markets: Full Comparison (AO2)

ACS 2023 · 60-minute drive-time isochrone · candidate venue addresses

Market	Census Tracts	Reachable Pop.	Median HH Income	Age 18–54 (tract mean)	Pop Δ vs Sacramento
Sacramento (baseline)	711	3,166,025	$89,844	48.3%	—
San Jose	1,124	5,014,076	$147,459	50.7%	+58.4%
San Francisco	1,232	5,249,936	$143,095	50.1%	+65.8%
Riverside	1,658	8,207,709	$94,760	49.5%	+159.2%
Los Angeles	3,085	12,960,180	$91,853	51.1%	+309.4%
Anaheim	3,044	13,203,228	$90,893	51.3%	+316.9%

Why Los Angeles over Anaheim (Anaheim is marginally larger by population):

Higher median income ($91,853 vs $90,893) → stronger premium ticket potential
More census tracts (3,085 vs 3,044) → broader geographic audience spread
Higher existing social media presence (20 tracked website users vs 13 in Anaheim)
Venue: 1111 S Figueroa St (Crypto.com Arena area) vs 2695 E Katella Ave (Honda Center)

Source: ACS 2023 5-year estimates, Census Bureau. Drive-time isochrones generated at the census-tract level. Age 18–54 = pct_age_18–34 + pct_age_35–54.
Method note: the age column is an unweighted mean across tracts, as submitted by the section owner (AO2, Lydia Chen). Weighting Los Angeles by tract population gives 52.8% — the figure the written report uses. The market ranking is unchanged either way.

A.10 — All 12 Fighter Matchups: Complete Rankings (AO6)

RSMT 15 Instagram promotional posts · Engagement Score = Likes + 2 × Comments

Rank	Matchup / Post	Likes	Comments	Engagement Score
1	Taylor vs Tiffer	428	41	510
2	Houston vs Hall	420	41	502
3	Phetamphone vs Handy	288	37	362
4	Sanchez vs Villanueva	268	10	288
5	Harris vs Sokolova	246	7	260
6	Jack (solo)	197	30	257
7	Logan (solo)	206	23	252
8	Rodriguez vs Sardi	206	14	234
9	Mayorga vs Licea	201	10	221
10	Bullie vs Hargrove	163	13	189
11	Carlos (solo)	135	24	183
12	Underwood vs Roberts	113	14	141

Follower context (top 3 by follower count): Carlos — 2,457 followers · Vera Sokolova — 2,450 followers · Darian Houston — 1,559 followers.
Koby Taylor (Taylor vs Tiffer, rank #1, score 510) has far fewer followers than Sokolova (rank #5, score 260) or Carlos (rank #11, score 183) — the two most-followed fighters on the card. Follower count is not a reliable booking predictor — matchup excitement drives engagement.

A.11 — AO3: Traffic Sources + AO5: Conversion Timing

AO3 — Traffic Source Breakdown (total events)

Source	Events
google	1,946
(direct)	1,894
ig (Instagram in-app)	634
bing	71
yahoo	40
facebook.com	30
lm.facebook.com	30
duckduckgo	26
l.instagram.com	19
m.facebook.com	11
l.facebook.com	5
ecosia.org	4
instagram.com	4
chatgpt.com	2
linktr.ee	2

Total page views: 2,668 · Unique users: 1,095 · Avg: 2.44 pages/user

AO5 — High-Intent Conversion Timing

90.6% Same-day: converted on the same calendar day as first visit

78.3% Single-session: converted within their first session — no return visit needed

AO5 Method: Converter = session with a ticket/PPV click-through (Eventbrite/Akaizo). First-touch date vs. conversion date, measured among converters (n = 235). Same finding as the Audience Diagnosis slide — the buying window is short.

Implication: Capture high intent at the moment it peaks — direct social and email links to the event page, not the homepage — because buyers rarely return for a second session.

A.12 — Study Limitations & RSMT 16 Outlook

Known limitations:

Click intent ≠ confirmed purchaseWe measure redirect clicks to Eventbrite/Akaizo. The gap between click and confirmed transaction is unknown without a post-purchase pixel. All conversion rates are lower bounds.

Two event cycles = directional, not causalRSMT 14 (Jan 31) and RSMT 15 (Apr 18) establish behavioral patterns. Causal claims need more cycles and controlled variation — these findings should seed experiments, not replace them.

No demographic data on site visitorsGA4 does not surface age/gender in this export. Demographic findings (Los Angeles 52.8% age 18–54, population-weighted across tracts) come from ACS Census at the market level — not from actual visitor profiles.

What RSMT 16 data adds:

Third cycle → trend validationThree data points confirm whether channel and funnel patterns are stable across events. RSMT 16 is the first real test of whether these findings are systematic — not event-specific.

Pre/post optimization comparisonIf homepage CTA and direct social links are implemented before RSMT 16, the data will directly measure impact — converting these recommendations from theoretical to evidenced.

Longitudinal fighter engagement baselineThree events of Instagram scoring data builds a fighter performance ranking over time — not a single-event snapshot. More reliable as a booking and promotion criterion.