Measuring sales from Turkish and Kurdish community ads

To measure whether advertising with Turkish and Kurdish community publishers generates extra UK sales, compare results with a credible group that did not receive the additional advertising. Use tracked links and offer codes to understand response, but do not mistake attributed purchases for sales that would not otherwise have happened.

The practical question: did this publisher buy add sales?

A community publisher reports strong engagement. Its tracked links generated orders. Your finance director asks whether those customers would have bought anyway.

That is an incrementality question, not a click-through-rate question. The useful test is whether adding a defined publisher placement to your existing marketing produces more sales than continuing without it.

For a UK brand buying Turkish-language media, Kurdish-language content or English-language community outlets, this distinction matters. Publisher audiences can overlap with your paid search, social, email and retail activity. A purchase credited to a publisher may involve an existing customer already responding to another channel.

Keep the question narrow: “Does adding this publisher package increase completed orders at an acceptable cost?” Do not turn one campaign into a verdict on whether an entire community is commercially valuable.

Define the media audience without assigning ethnicity

“Turkish and Kurdish audiences” is not a single media segment. Turkish language use does not establish Turkish ethnicity, and Kurdish people may use Kurdish languages, Turkish, English or other languages in different settings.

Media habits also vary by generation, faith, region and individual. A publisher’s editorial focus tells you something about its context, not the identity of every reader.

Before buying, ask each publisher for:

• Its UK audience evidence, including the source and reporting period.
• The languages used in the proposed placements.
• Its available regional delivery controls.
• How it distinguishes impressions, unique users and engagements.
• Whether reported reach includes overseas traffic or distribution partners.
• Whether placements can support a controlled experiment.

Treat each outlet as a documented media environment. Label reporting rows by publisher and placement, rather than labelling all purchasers “Turkish customers” or “Kurdish customers”.

Measure the performance of the advertising you bought, not an ethnicity you have inferred.

Choose one primary sales outcome

Agree the outcome before launch. For an ecommerce campaign, that might be completed orders delivered to eligible UK areas, net of cancellations and returns within a defined reporting window.

For a service business, it might be paid appointments rather than enquiry forms. For retail, it might be store-level transactions, provided store catchments and sales reporting support a credible comparison.

Set three things in writing:

1. Primary outcome: the business result that determines success.
2. Measurement window: the period covering advertising, purchase delay and relevant returns.
3. Commercial threshold: the cost per additional order or contribution needed to justify another buy.

Clicks, video views and landing-page visits remain useful diagnostic measures. They help explain delivery and response, but cannot substitute for the agreed business outcome.

If order volumes are too low to detect a commercially meaningful difference, lengthen the test, increase its scale or choose a more frequent outcome with a clear commercial connection. Do not rename engagement as sales impact.

Use the strongest feasible holdout design

A holdout is a comparable group deliberately excluded from the additional advertising. Both groups should otherwise continue receiving normal marketing.

Prefer a genuine randomised split

Where a publisher or buying platform supports a valid randomised experiment, split eligible users into treatment and control groups before exposure. The treatment group can receive the additional campaign; the control group cannot.

Compare outcomes across the assigned groups, not simply people who clicked against people who did not. Clickers are self-selecting and may already be more interested in buying.

Ask how assignment works, how cross-device exposure is handled and whether the experiment measures purchases beyond tracked clicks. A dashboard labelled “lift” is not sufficient methodological evidence.

Use geographic testing when delivery supports it

For some local or regionally targeted buys, comparable geographic areas can provide a practical alternative. Select multiple areas with similar pre-campaign sales patterns, then randomise treatment where feasible.

The added publisher advertising runs in treatment areas and is withheld from controls. Existing national marketing continues consistently across both.

This only works if geographic delivery is credible. A national editorial article, shared social post or broadly distributed newsletter may spill into control areas. Check readers’ actual locations rather than assuming a publisher’s office location defines its reach.

Avoid making one neighbourhood the treatment and another the control simply because both are associated with particular communities. Differences in retail access, income, customer base and local activity can undermine the comparison.

Be honest when no holdout is possible

A before-and-after sales comparison can describe what happened, but seasonality, promotions and other advertising may explain the change.

If the available buy cannot support a credible control, report attributed response, not proven incremental sales. That is a useful limitation to disclose, not a reporting failure.

Separate tracking from proof of impact

Give each publisher, placement and creative version consistent campaign parameters. Test links, redirects and checkout tracking before launch, including mobile journeys.

Offer codes can supplement this setup, but they have limitations: customers may forget them, share them or discover them on voucher sites. A discount can also change buying behaviour, so keep the offer consistent across an experiment unless the offer itself is what you are testing.

Report these measures separately:

• Delivery: where and how the advertising appeared.
• Attributed response: orders connected to links, codes or another stated attribution rule.
• Incremental outcome: additional orders estimated through the controlled comparison.

Do not add publisher-reported conversions to analytics conversions without checking overlap. Different systems may claim the same purchase under different attribution windows.

Consent choices, browser restrictions and cross-device journeys can also limit observed attribution. Missing tracking is not proof that advertising had no effect.

Calculate the result and show the uncertainty

For a randomised test, compare purchase rates between the assigned groups and scale the difference to the relevant treatment population. Account for unequal group sizes rather than comparing raw order totals.

For a geographic test, a common approach compares the treatment areas’ change with the control areas’ change. Check pre-campaign trends and use an analysis suited to area-level data; a simple spreadsheet difference does not establish statistical certainty.

Suppose a properly analysed test estimates 100 additional orders and the added campaign cost £4,000. The estimated cost per incremental order is £40. These are illustrative numbers, not a benchmark.

Include production, translation and publisher fees in the cost definition. Assess contribution after variable costs, not revenue alone, and avoid assuming future repeat purchases without supporting evidence.

Show an uncertainty interval where the method supports one. If the plausible range includes no improvement, describe the evidence as inconclusive rather than announcing a win.

Pre-launch measurement checklist

• One decision and one primary commercial outcome are agreed.
• Publisher audience claims have documented sources and dates.
• Treatment and control assignment is recorded before launch.
• Expected sales volume can support a useful test.
• Geographic spillover or cross-group exposure has been assessed.
• Pricing, availability and other marketing are comparable.
• Links, codes and purchase events have been tested.
• Costs, returns, attribution windows and stopping rules are defined.
• Data collection and sharing have passed privacy review.
• The report distinguishes attribution, incrementality and uncertainty.

Keep the measurement privacy-conscious

You do not need to identify customers’ ethnicity to assess a community media buy. Avoid creating ethnicity labels from names, postcodes, language preferences or browsing behaviour.

The ICO’s special category data guidance explains the additional rules for personal data revealing racial or ethnic origin, including relevant inferences. Where those rules apply, an Article 6 lawful basis alone is not enough; an Article 9 condition is also required.

Prefer appropriately aggregated reporting where it answers the question. Aggregation does not retrospectively remove obligations attached to collecting or sharing personal data. Have your privacy team check tracking technologies, consent requirements and any publisher-to-brand data transfer.

Review cultural risk before launch

A sound experiment can still test poor creative. Review language choice, translation, imagery, flags, maps and identity references with people who understand the specific audience and placement. Do not assume Turkish-language copy is appropriate for every Kurdish-focused outlet, or that one reviewer represents everyone.

Kulture adapts ad copy, images and video for UK communities and runs cultural risk reviews. Complete that review before the test, then log any creative changes: changing the message halfway through can change what your results actually measure.