Is there an official source for Turkey ad benchmarks by industry?

There is no single official table. IAB Türkiye publishes market level advertising investment reporting and TÜİK publishes national internet and technology use statistics, but neither breaks the market into per sector cost per click or cost per acquisition figures. Sector tables you find online almost always come from a publisher's own account panel, so check the methodology note for the market it actually measured.

Can I use US or global benchmark tables for a Turkish account?

Use them for direction, never for targets. Relationships tend to travel between markets, for example that considered purchases convert more slowly than impulse ones. Absolute costs do not travel, because auction competition, currency and the publisher's sector definitions all differ. If you set a target from an imported table you are optimising against someone else's account.

How many conversions do I need before a baseline means anything?

Decide the number before you spend rather than after, and size the budget from it. If your target cost per purchase is 400 TL and you want 50 purchases, the window costs 50 × 400 = 20,000 TL at target. If the observed cost turns out to be 625 TL, completing the same 50 purchase sample costs 50 × 625 = 31,250 TL. Ending a test early because the calendar month closed is the most common way baselines get built on noise.

Which metric should I benchmark first?

The one closest to money that you can measure reliably. Cost per acquisition and return on ad spend are the decision metrics, while cost per click and cost per mille are diagnostic: they explain why the money metric moved. Benchmark the money metric against your own margin, and use the auction metrics only to investigate a change you have already spotted.

Turkey ad benchmarks by industry: what exists

When advertisers ask for Turkey ad benchmarks by industry, they usually want one specific thing: a table of cost per click, cost per thousand impressions and conversion rate, broken out by sector, for the Turkish market. That table does not exist as a single authoritative published source. What does exist is a set of benchmark reports built from whichever accounts the publisher happens to see, a small number of regional breakouts that fold Turkey into a wider EMEA bucket, and platform tools that will quote a forecast for the targeting you enter. Telling those apart is most of the job. The rest is building the one benchmark that is genuinely yours.

Where Turkey ad benchmarks by industry actually come from

Nearly every sector table circulating online traces back to one of three families of source, and the three answer completely different questions.

Publisher panels. Companies that manage or supply tooling to a large number of advertiser accounts periodically publish sector averages drawn from that book of business. WordStream and LOCALiQ are the best known publishers of recurring search and social benchmark posts of this kind. The figures describe real accounts, but the sample is the publisher's own customer base and the geographic mix is whatever that customer base happens to be. Open the methodology note before you quote anything from it. If the note does not name the market, assume the market is not Turkey.

Platform planning tools. Google Keyword Planner returns forecast ranges for the keywords, locations and budgets you actually enter. Meta's campaign setup shows estimated results for the audience you actually built. These are the only cost signals here generated from live auction data for the geography you selected, which makes them the closest thing to a current Turkey specific reference available before you spend. Read them as forecast ranges, not as commitments.

Market level industry bodies. IAB Türkiye publishes a periodic advertising investment report, and TÜİK publishes national statistics on household internet and technology use. These tell you how large the market is and who is online, which helps you size demand. They will not hand you a cost per acquisition target, and any article presenting them as though they did has changed the subject.

Four provenance questions before you trust a number

A benchmark is only as good as the answer to four questions. Ask them in this order, and stop at the first one the source cannot answer.

Whose accounts are in the sample? A panel drawn from one agency's clients is self selected: those advertisers chose that agency, at that budget level, in those verticals. That describes a customer base, not a market.
Which geography, and in which currency was it recorded? A global average containing Turkey as a thin slice is dominated by whichever markets contributed most of the spend, and a figure converted at an unstated date carries that date inside it.
What counted as a conversion, and under which attribution setting? A conversion rate measured on a click plus view window is a different number from the same account measured on last click. If you have not internalised why, the piece on ad campaign attribution models is the prerequisite for reading any benchmark table at all.
Which metric is on the table, and over what period? Cost per click, cost per mille and cost per acquisition move for different reasons, so a table that reports one of them tells you very little about the others. The breakdown in CPA vs CPC vs CPM is worth keeping open beside any sector list.

A source that cannot answer all four is not a benchmark but decoration, and it gets expensive once you set targets against it.

What the sector label hides

Even a well documented table has a structural problem: the sector label is invented by the publisher. Buckets like retail, services or home improvement are editorial groupings, and a business that sells one high value product to a narrow professional audience can sit in the same row as a mass market catalogue store. Many businesses straddle two rows and belong properly to neither.

Bundling does the rest of the damage. A search row that mixes branded queries with non branded ones will always look cheaper and better converting than your prospecting campaigns, because people searching your own name were never really being acquired. A social row that mixes retargeting with cold traffic has the same defect. Unless the methodology says the two were separated, assume they were not.

Currency drift compounds all of this, but it is a distinct problem with its own answer, so treat any lira denominated figure older than your last planning cycle as a historical record rather than a target. The practical conclusion is narrower than most people expect: directional statements survive the trip between markets, absolute costs do not. That a considered purchase converts more slowly than an impulse one is portable. That it costs a particular amount to acquire is not.

Building a Turkey baseline when you have no history

If you are entering the market, you are not benchmarking yet. You are buying the data you will later benchmark against, and the quality of that purchase depends on decisions you make before the first impression serves.

Fix the definitions before you spend

Write down one conversion event, one attribution setting, one currency and one reporting window convention, then leave all four alone for the whole baseline period. Any mid flight change resets the comparison, and you will not notice, because the reports keep rendering.

Size the window in conversions, not in days

A calendar month is a budgeting unit, not a statistical one. Decide how many conversions you need before drawing a conclusion, and let that set the spend. Suppose your target cost per purchase is 400 TL and you settle on 50 purchases as the smallest sample you will act on. At target, that window costs 50 × 400 = 20,000 TL.

Now spend it and count honestly. If those 20,000 TL produced 32 purchases, your observed cost per purchase is 20,000 ÷ 32 = 625 TL. Be precise about what that means: 625 ÷ 400 = 1.56, so you are running at roughly 1.6 times your target. Rounding that up to a factor of 2 in your head is a small slip with a large consequence, so carry the ratio you actually computed. Completing the full 50 purchase sample at the observed rate would take 50 × 625 = 31,250 TL, which is the real budget question in front of you.

Test the result against your margin, not against a table

With an average order value of 800 TL, that account returned 800 ÷ 625 = 1.28 lira of revenue for every lira of spend. Whether 1.28 is good has nothing to do with any published sector figure and everything to do with your contribution margin. At a 40 percent contribution margin, break even sits at 1 ÷ 0.40 = 2.5, so 1.28 is not a slightly weak result, it is a loss making one. Running the same test with your own margin in the break even ROAS calculator takes a minute and replaces an entire afternoon of table hunting.

Record the spread, not just the total

One blended figure hides everything interesting. Alongside the total derived number, keep the weekly values and look at the middle one rather than the average, because a single unusual week drags an average and leaves the median roughly where it was. Two accounts with the same blended cost per purchase can have very different spreads, and the spread is what tells you whether the number is stable enough to plan against.

Reading the baseline once you have one

From this point the useful comparison is your account against itself: the same window length, the same conversion definition, the same attribution setting, period over period. External sector tables keep one legitimate job: telling you the direction of travel in a vertical you have never advertised in. They are a poor source of targets and a worse source of ceilings.

Re baseline on a schedule rather than after a bad week, and verify your sources again each time, since methodology notes change quietly between editions. Once the baseline is stable, the question stops being what the market costs and becomes how fast you can add budget without unpicking what you measured, which is a scaling problem rather than a benchmarking one.