A/B Testing
in
Aberdeen.
A/B Testing in Aberdeen the right way looks nothing like what most agencies sell. No call-it-early. No "the variant's trending up". One test at a time, properly powered, called at 95%.
I run a structured experiment programme: 1–2 properly-sized tests every month, each rooted in real conversion research, each documented win, loss or null result. After six months you have a private library of what actually works for your customers — and a measurably better-converting site.
1–2 tests / month · 95% significance · Rolling monthly
Bad A/B testing is worse than no A/B testing.
A test called three days early on 200 sessions is a coin flip with extra steps. It tells you nothing — but it feels like it told you something, so you ship the variant. That's not evidence; it's noise dressed up as data.
"Looks like a winner."
- Test launched on a "what if?" idea from a Slack message.
- Sample size never calculated. "We'll see what happens."
- Called on day 4 — "variant is up 14%, let's ship it".
- Three weeks later, conversion is flat. The "winner" was noise.
- Test forgotten. No record. Six months later you run the same one.
"Test held. Result conclusive."
- Every test starts with a data-backed hypothesis, scored, prioritised.
- Sample size calculated up-front. We tell you the test length on day one.
- 95% statistical significance threshold. Held until reached.
- Mobile + desktop segmented. Blended winners that lose on mobile? Caught.
- Every test logged — searchable knowledge base you keep.
Hypotheses scored. Best one first.
Every test idea goes into a backlog and gets an ICE score — Impact, Confidence, Ease. The highest-scoring hypothesis runs next. Each one written in plain English: "If we change X because of Y, we expect Z to improve." You sign off before we build.
- Hypotheses sourced from analytics, recordings, interviews
- ICE scoring — Impact × Confidence × Ease (out of 10 each)
- You see the full backlog — and pick what runs first if you'd rather
-
#1Single-step checkout (mobile)I:9 C:8 E:7ICE 8.0Next up
-
#2Sticky CTA on long pagesI:7 C:8 E:9ICE 8.0
-
#3Reviews above hero, not belowI:8 C:6 E:9ICE 7.7
-
#4Headline · concrete promise rewriteI:8 C:7 E:7ICE 7.3
-
#5Pricing FAQ above the foldI:7 C:6 E:8ICE 7.0
-
#6+ 14 more in backlogUpdated weekly from research
We don't call tests early. Ever.
Before any test goes live we calculate the sample size needed to detect a meaningful effect at 95% confidence — based on your actual traffic, your actual conversion rate, and the minimum effect we'd consider worth shipping. Then we hold until we hit it. No early calls when "the variant's trending up". No mid-test peeks that fool you.
- Sample size calculated up-front, test length quoted on day one
- Held to 95% statistical confidence — no early shipping
- Mobile + desktop segmented — blended winners don't fool us
Every test, documented. Forever.
Win, loss or null — every test we run gets written up: the hypothesis, the variant, the result, what we learned, what it means for the next one. Six months in, you have a private knowledge base of what actually works for your customers. That's the part most agencies skip — and it's the part with the longest payoff.
- Searchable archive of every test — yours forever
- Losses written up properly — they teach you the most
- Each test informs the next — hypotheses get smarter monthly
-
TEST-041WinSticky CTA on mobile+8.4% rel. · 97% conf. · Shipped 12 Mar
-
TEST-040NullFree-shipping banner+1.2% · 38% conf. · No effect
-
TEST-039LossHero video instead of image-6.1% rel. · 96% conf. · Reverted
-
TEST-038WinForm: phone before email+14.7% rel. · 98% conf. · Shipped 28 Feb
"Daniel produced a brilliant website for us which has had a lot of positive feedback for its clarity and ease of use. He listened attentively to all our suggestions and guided us towards a structure which would best reflect our content."
Us, vs the other ways to run tests.
Bad A/B testing is worse than none at all — you end up shipping noise and trusting the wrong things. Here's the difference between our programme and the alternatives.
One loop. Repeated every month.
Same process every month. Each cycle ships one tested change and adds a learning to the knowledge base.
Backlog review
Top-scoring hypothesis from the backlog gets picked. Sample size calculated, test length forecasted, you sign off the brief.
Build variant
Variant built and QA'd. Anti-flicker handled. Tracking double-checked. Variant ships behind a 50/50 split.
Hold to significance
Test runs to the calculated sample size. No early calls, no peeking that affects the call. Status communicated weekly — without peeking at results.
Call + document
Winners shipped. Losers reverted. Nulls noted. Every test written up in your knowledge base — what we learned and what's next.
Aberdeen businesses I've helped get found.
Eyeconic Beauty
Now ranking #1 across 24+ local search terms, with a 174% lift in Google impressions. Oldmeldrum.
The Garioch Fish Bar
An appetite-led website for one of Inverurie's best-loved chip shops — built to show off the awards and pull in online orders.
Inverurie West Parish Church
A warm, accessible website that makes service times, events and online giving easy to find. Inverurie.
Got a question? Likely answered.
What Aberdeen business owners ask before signing up for a structured testing programme. If yours isn't here, just ring.
How much traffic do I need to run A/B tests?
Which testing tool do you use?
How long does a typical test run?
Why won't you call tests early?
Do you run multivariate tests?
What happens to the knowledge base if I leave?
Conversion Rate Optimisation across the North-East.
Based in the Granite City, optimising websites right across Aberdeenshire — wherever your customers are based.
Let's run experiments that ship real wins.
Drop me a line and we'll have a 20-minute chat about your traffic volume, your tracking, and whether A/B testing is the right next move for your business — sometimes it isn't, and I'll say so.