Conversion Research Sprint
Funnel analysis, session replays, heatmaps, on-site polls and user tests across your top revenue templates. Output is a ranked list of friction points with revenue attached.
San Francisco, CA
Conversion Rate Optimisation in San Francisco, CA
When a click costs what it costs in this market, a friction point on the product page is the most expensive thing you own.
Delivered remotely for brands across San Francisco and California.
Grow · San Francisco
Why San Francisco brands come to us for this
- Tests judged on cohort value at day 90, not first-session conversion rate
- Subscribe-versus-one-time presentation tested against retention, not just add-to-cart
- Price and fee display designed around California's total-price rules from the start
- Cancellation and pause paths built as easy as sign-up, as state law requires
- Qualitative research where traffic is too thin for a valid experiment — stated openly, not fudged
Conversion work here has a different objective function. A brand in a cheaper acquisition market can absorb a mediocre product page and make it up on volume. In San Francisco, where you are bidding against companies that can lose money on a customer for years, the same mediocre page is the difference between a cohort that pays back and one that does not. So we optimise toward cohort value at day 90, not toward the session conversion rate on a first visit — and those two goals genuinely conflict, because the fastest way to raise first-visit conversion is a discount that guarantees the cohort never pays back.
The highest-leverage test surface in this market is the subscribe-versus-one-time decision. Defaulting to subscription raises first-order rate and raises early cancellation; defaulting to one-time protects the cohort and starves it. The answer is not a default at all, it is how the choice is presented: the saving in dollars per delivery, the cadence set to a realistic consumption rate rather than the fastest one, and pause and skip stated as a visible promise. We test those elements against retention, which means results take longer to read and mean considerably more when they land.
California also constrains the test space in ways worth knowing. Price display law here means the total a customer will pay cannot appear as a surprise late in the flow, so the shipping and fee reveal has to be tested as honesty rather than as a delay tactic. Automatic renewal rules require a cancellation route as easy as the sign-up route, which rules out several conversion patterns that still get recommended in generic CRO advice. Working inside those constraints is a design problem, not an obstacle.
Local context
Testing on the traffic a $1M-$5M Bay Area brand actually has
Most San Francisco brands in our range do not have the weekly conversion volume to run a clean A/B test on a secondary page — and a founder who has run experiments at a tech company sometimes assumes they do, because at that company the sample arrived by Tuesday. So we are explicit about where testing works and where it does not. High-traffic templates and the cart-to-checkout path get real experiments with sample sizes stated before launch. Everything else gets sequential improvement backed by qualitative research: session replays on the subscription selector, exit polls on the pre-order page, support-ticket mining, and moderated tests with people who actually buy this category. That research is also where the pre-order finding always shows up — a hardware buyer abandons over an unanswered ship-date question far more often than over price, and the fix is a clearer commitment on the page rather than a bigger button.
Scope
What Conversion Optimisation includes
The same standard of work we run for every client — applied to a San Francisco brand’s realities.
Full service detailPrioritised Test Roadmap
Every hypothesis scored on impact, confidence and effort, with the projected revenue and required sample size stated up front. You always know what runs next and why.
Experiment Build & QA
Tests built and QA'd across devices, browsers and your app stack, with flicker-free rendering and no measurable hit to Core Web Vitals.
Statistical Analysis
Pre-declared sample sizes, sequential-testing guardrails and segment-level readouts by device, traffic source and new versus returning. No peeking, no calling a test at 80%.
Checkout & Cart Optimisation
Cart drawer, shipping thresholds, payment options, express checkout placement and post-purchase upsell, tested against AOV and revenue per session rather than clicks.
Monthly Programme Report
Tests run, results in full, cumulative revenue impact and what the results taught us about your customers. Losers are reported in the same detail as winners.
Scoped and quoted for your San Francisco store
We do not work off a rate card. Every San Francisco engagement starts with a fixed statement of work — named deliverables, named dates, one number — written after we have looked at your store, not before. If a smaller first step would serve you better, we will say so.
Get this scopedHow it runs
From kickoff to results
Measure
We instrument the funnel properly first. Most stores have broken or double-counted events, and you cannot optimise against numbers you cannot trust.
Research
Quant tells us where visitors leave. Qual tells us why. We combine analytics, replays and direct customer input before writing a single hypothesis.
Prioritise
Hypotheses are scored and sequenced so the highest-value, lowest-effort tests run first. The roadmap is shared and you can reorder it.
Test
Two to four concurrent experiments depending on traffic, each run to a pre-declared sample size. No test gets stopped early because it looks good on day three.
Implement & Compound
Winners get hard-coded into the theme, losers get documented, and every result feeds the next round of hypotheses. The programme gets smarter each month.
Proof
Conversion Optimisation results
Anonymised under NDA. Figures pulled from the client’s own analytics.
Premium Skincare (DTC)
8-figure DTC brand, ~140 SKUs, US + CA · Shopify Plus (migrated from Magento 2)
Magento 2 cost $9k a month in hosting, extensions and emergency dev before a single feature shipped. Mobile LCP sat at 5.4 seconds and mobile converted at less than half the desktop rate across 68% of sessions. The named constraint: 11,412 indexed URLs and a top-3 organic position on their hero category, so a migration that lost rankings would cost more than the platform ever saved.
+41%mobile conversion rate1.62% to 2.28% over the first 90 days post-launch
5.4s → 1.8smobile LCP75th-percentile CrUX field data, not lab
+27%organic sessionssix months post-migration vs. pre-migration baseline, branded queries excluded
$108k/yrplatform cost removedhosting, extension licences and the standing emergency dev retainer
Performance Apparel (DTC)
~$6M/yr DTC, 900+ SKUs across size and colour variants, US · Shopify Plus
Returns ran at 31% and refund cost consumed the entire paid media margin. One size chart image served 40 different fits, and 62% of add-to-carts started on a collection page that never showed variant availability. The named constraint: no new product photography budget, so every fix had to come out of the existing asset library and the review corpus.
+29%sitewide conversion rate1.71% to 2.21%, five-month average
+14%average order value$84 to $96 once the threshold bar and cross-sell shipped
31% → 22%return ratenine-point drop, roughly $310k/yr in recovered margin
1.9x → 2.4xblended MERwith paid spend held flat throughout
In their words
Clients on this work
CVR 1.9% → 2.7%
“We'd been running our own tests for a year and calling every 3% swing a win. The first thing they did was tell us our sample sizes were garbage, which was not a fun call to sit through. Six months later we're getting wins that actually hold up when you re-measure them. Sitewide CVR went from 1.9% to 2.7%.”
Founder & CEOApparel DTC brand, ~$12M/yr · Los Angeles, CA
Verified client, 2025FAQ
Conversion Optimisation in San Francisco — your questions
Enough for the homepage, your main collection and the cart-to-checkout path, and not enough for most product-page variants read in a sensible timeframe. We will tell you which of your templates clear the bar and give you the projected runtime for each before anything launches. For the rest, sequential change guided by research is more honest than declaring a winner from an underpowered test, which is how teams end up shipping noise.
Only if your retention data says the resulting cohort survives, and most brands have not checked. Defaulting to subscribe reliably lifts first-order conversion and reliably lifts month-one cancellations, so the net can be negative once you count support load and refunds. We usually test presentation rather than default: the saving expressed per delivery, a realistic cadence, and an explicit pause promise on the page.
It removes a category of test rather than blocking the work. Patterns that hide mandatory fees until late in checkout are not available to you, which is fine because they were also destroying trust and generating chargebacks. What is still very much testable is where and how the honest total appears — early on the product page, in the cart, as a shipping threshold — and that placement moves conversion more than the concealment ever did.
The ship-date commitment and what happens if it slips. Buyers of an unreleased product abandon over uncertainty far more often than over price, so the tests worth running are about specificity: a stated window rather than coming soon, a clear description of when the card is charged, a plain refund policy before the ship date, and a named channel for updates. We test those as content, then test whether a deposit structure outperforms full payment upfront.
Around 40,000 sessions and 800 orders a month is where a testing programme becomes statistically viable. Below that, tests take months to reach significance and you are better served by research-led redesign work and analytics fixes. We will tell you honestly which bucket you are in.
The first test goes live in week three, after research and instrumentation. Meaningful cumulative impact typically shows around month four, once six to ten tests have run. CRO is a compounding programme, not a one-month fix, and anyone promising otherwise is selling best practices.
Some do, and that is normal. A loser is still information: it rules out a hypothesis and sharpens the next one. We report losses in the same detail as wins because a programme that only ever produces winners is one that is not being measured properly.
Next step
Conversion Optimisation for your San Francisco brand.
Thirty minutes with the strategist who would actually run your account. We screen-share your store, read your data live, and tell you the three highest-value things we can see from the outside.
Shopify or Shopify Plus stores doing $150k/mo or moreFounder, CEO or eCommerce lead on the callNo deck and no pitch — we open your store instead
Prefer to write it out? [email protected] gets a real reply the same business day, Mon-Fri, 9am-6pm MT.