Metrics

A/B testing indoor screen creative in 14 days

May 13, 20268 min

"I made 5 versions of the creative, which one do I choose?" — a typical question from a marketer launching an indoor campaign for the first time. The answer costs less than a starter campaign: run an A/B test for 14 days and let real data pick the winner. Let's break down how to do it properly.

Why an A/B test in indoor is simpler than in digital

Many marketers associate A/B testing with 200-page dashboards in Optimizely or Google Optimize. In reality, with indoor everything is much simpler:

Fewer variables — time of day, audience, location are already fixed. Only the creative changes.
Clean metrics — QR clicks and conversions, without noisy indicators like "view-time"
Fast results — 2 weeks is enough for a statistically significant sample in most niches
Low entry threshold — from ₴1,500 for a test package, not ₴50,000 as in performance digital

This is the best channel for a beginner advertiser who wants to learn to make decisions based on data.

What to test: 4 main hypotheses

Don't test "everything against everything" — that gives murky results. Test one variable at a time, in turn:

1. Message (the biggest impact)

Version A: "English in 4 months"
Version B: "Junior offer in 6 months"

The same course, two different "promises" — for two different audience motivations.

2. CTA (second by impact)

Version A: "Free lesson"
Version B: "20% discount"

Free usually drives more scans but worse downstream conversion. A discount — fewer scans, better quality.

3. Social proof

Version A: "1,300+ graduates"
Version B: "4.9 rating on Google"

The first option is about scale, the second about quality. Different audiences react differently.

4. Visual style

Version A: a static frame with a person
Version B: dynamic brand animation

Here there's often a surprise — static sometimes wins thanks to "readability" from 3 meters away.

How to properly structure an A/B test

Step 1. Choose 2 creatives with an identical budget

Don't run "a small budget for one, a large one for the other." That's not an A/B, that's A vs B with a skewed competition.

We take an equal number of impressions for each. The simplest approach — different coffee shops with the same audience. For example, creative A in Karamel', creative B in CAVA HOUSE (both in central Kyiv, similar clientele).

Step 2. A separate QR + UTM for each

This is critical. Without a unique UTM you won't be able to tell which creative delivered your 50 scans.

Creative A: utm_content=variant_a
Creative B: utm_content=variant_b

This small step takes 5 minutes but makes the whole experiment clean.

Step 3. Give it 14 days

Why 14 days and not 7?

A week is not representative — in coffee shops Tuesday and Saturday have different audiences
2 full weeks give 2 normal cycles (weekday + weekend)
Statistical significance for typical scan volumes (50–200 per creative) is reached in ~14 days

Less than 14 — you can catch a random anomaly week. More than 21 — a waste of time, the decision is already visible.

Step 4. Don't change anything mid-test

The most common mistake: you see that on day 2 variant A has 5 scans and B has 12, and you stop A. Don't do that. On day 14 B may have 80 scans and A — 120, because A revved up.

The first 3–5 days of data are noise. Look at the full picture.

Step 5. On day 14 — the decision

You simply look at these metrics:

Metric	Variant A	Variant B
Number of QR scans	87	142
Sessions on the site	79	130
Engagement rate	38%	52%
Conversions (leads/purchases)	4	12
CPA (at a budget of ₴3,000 per variant)	₴750	₴250

Here the winner is obvious: B outperforms A by 3 times on CPA. For the next month — we run B at all locations.

What to do if the results are ambiguous

Sometimes A and B give close figures (a difference of <20%). This is no reason to panic:

Check whether it's statistically significant — if the difference is <20% with <100 scans, that's noise, not signal
Look at conversion quality — perhaps A gives few scans but a high conversion to sale
Try a third option — the winner of both A and B against a third variant

Real examples of hypotheses that gave a 2x difference

From the experience of advertisers running indoor campaigns in Kyiv:

"Free trial lesson" vs "Course with a 30% discount" — the first option gave 2.5 times more scans, but a 1.8 times worse downstream conversion. The net CPA was roughly equal.
"Haircut ₴350" vs "Top barber of Podil" — the specific price won by 3 times.
QR with the motivator "Book now" vs QR with "See examples of our work" — the second option gave 2 times more scans and 4 times more real leads.
Logo + text without a person vs Logo + a customer photo — the face gave +30% scans (the effect of a human face).

These are not rules, but hypotheses. In your niche it may be different. That's exactly why you test.

What you should NOT A/B test in indoor

Not every variable is worth fussing over:

❌ Minor text differences ("Sign up" vs "Register" — the difference is irrelevant)
❌ Music/sound — many coffee shops play without sound
❌ Color accents at 5% — the human eye won't notice in 15 sec
❌ Logo variations — that's not an A/B for indoor, that's for brand guidelines

Test big hypotheses about messaging, benefits, CTA. Trifles save pennies at the expense of your time and focus.

HostAd for A/B testing

HostAd makes A/B tests convenient through several mechanics:

Quick uploading of different creatives. Changing the creative between weeks (without restarting the entire campaign) is literally a few clicks.
Coffee shops with a similar audience nearby. The map shows coordinates, so you can pick 2 locations in one district with the same demographic profile. This makes the test cleaner.
Transparent pricing. The same price for the same slots gives identical "entry conditions" for both variants — without the agency's "the price for you is special."
QR and UTM at the platform level. No need to re-upload the creative to change the UTM — you just change the URL in the campaign settings.

Summary

An A/B test on indoor is the cheapest and simplest way to learn to make decisions based on data instead of "we think." 14 days + a ₴6,000 budget (₴3,000 per variant) give you an insight that will then save tens of thousands on proper scaling.

What to do today:

Formulate 1 main hypothesis (message, CTA, proof, or visual)
Make 2 creative variants that differ only in that one point
Go to the HostAd map, choose 2 coffee shops with the same audience
Create 2 campaigns with different UTMs per variant
Run for 14 days
On day 14: a decision based on CPA, scaling the winner

More about metrics — in the articles QR code in indoor advertising and how to measure ROI.