TikTok Ads CPM benchmarks, France: build yours in 14 days

In France there is no single TikTok Ads CPM but a wide range, driven by six variables: campaign objective, placement, audience size, seasonality, format (Spark or not) and creative quality. An external benchmark is only usable when the objective and placement are identical. CPM is rarely something you simply endure: you steer it through the retention rate of the first three seconds. The best practice is to build your own reference index over 14 days, then compare yourself against it week after week. This article gives you the normalisation grid and the protocol, not yet another table. According to the SRI and UDECAM's 35th Observatoire de l'e-pub, published in February 2026, the French digital advertising market reached €12.4 billion in 2025 (+11%), driven in particular by Social and Video.

Is there really an average TikTok Ads CPM in France in 2026?

No. There is a distribution, not an average, and almost all the figures circulating about “the TikTok CPM in France” come from foreign panels quoted in dollars. When you type “CPM benchmark 2026” into a search engine, the first results are Anglo-Saxon aggregators: analytics tools that compile their own customers' accounts, UGC platforms that publish the figures from their campaigns, e-commerce suites whose panel is mostly made up of American merchants. None of these pages says how many accounts it is calculating on, over what period, or with what campaign objective.

This is not a methodological detail. A “beauty sector” CPM calculated on 300 accounts of which 80% are American doesn't hold up for a French advertiser: bid density, inventory fill rate and the retail calendar have nothing in common. Not to mention the currency. Carrying a dollar figure into a media plan in euros means a 5 to 10% error before you've even started.

You also need to distinguish three notions that everyone confuses.

• Displayed CPM: what you enter, or what the tool estimates for you before delivery. A forecast, nothing more.
• Billed CPM: what TikTok actually charges you per thousand impressions served. It is the line in your Ads Manager.
• Useful CPM: the cost of a thousand impressions that have a chance of producing something. An audience network impression and a TikTok In-Feed impression are not bought on the same market: the gap in price per thousand is commonly a factor of 3 to 4 on our accounts.

On our French client accounts, the gap between the lowest and highest CPM recorded in the same month regularly exceeds a factor of 4. Same country, same quarter, sometimes similar verticals. Looking for “the” average CPM in that cloud of points is like asking for the average price per square metre in France.

Why do two French advertisers pay CPMs that differ by up to four times on the same day?

Because six variables move the CPM, and five of them are under your control. As long as these variables aren't aligned between two accounts, the comparison is meaningless. Here is the grid we apply before agreeing to compare anything.

1. The campaign objective

Reach, Traffic, Video Views, Conversions, Catalogue: these are distinct auction markets. With a Reach objective, you buy raw impressions and the algorithm looks for the cheapest contact. With a Conversions objective, it looks for a likely buyer, and the CPM absorbs the extra cost of that selection. Comparing your “Conversions” CPM to a “Reach” benchmark is mistake number one, and it is made every day. On a single account, switching an ad set from Traffic to Conversions is enough to multiply the CPM by two to three, with nothing else having moved.

2. The placement

Leaving placement on automatic opens the inventory beyond the TikTok feed, wherever the advertising network (formerly Pangle) is eligible: impressions there are far cheaper, and your overall CPM collapses, which delights whoever looks at the CPM column and often degrades the CPA. A caution, though: the TikTok advertising network is not available on France targeting. In that case, a very low CPM is more likely explained by the objective chosen, a very broad audience or a high-hold-rate creative that sells nothing. Check in Ads Manager which placements are actually open before explaining a low CPM by off-TikTok inventory. To benchmark cleanly, isolate TikTok alone. Always.

3. Audience size and exclusions

The smaller the audience gets, the tighter the auction. Targeting 400,000 people in France with three exclusion lists is structurally more expensive than broad. Pure retargeting, on audiences of a few tens of thousands of profiles, often shows the highest CPMs in the account, and that's normal, it's not a warning sign. What matters is the cost per acquisition that comes out of it.

4. French seasonality

The French retail calendar dictates competitive pressure on inventory. The Black Friday sequence, on 28 November in 2025, through to mid-December concentrates e-commerce budgets, and the price of attention rises. The first week of January, with the start of the winter sales, causes a second, shorter rise. Conversely, the second half of July and the month of August are the cheapest windows of the year on our accounts. We don't publish a quantified seasonal coefficient: it varies too much by vertical, and we don't have a long enough series to pin it down honestly. According to Service-public.fr, the winter sales begin on the second Wednesday of January at 8 a.m., or on the first Wednesday if the second falls after the 12th; in 2026, they therefore began on Wednesday 7 January.

5. The format: standard In-Feed or Spark Ads

A Spark Ad runs from a real account, with its comments and interactions visible. Engagement is generally higher, and the algorithm rewards it. But not systematically: on accounts with a thin organic history, we have seen Spark Ads do worse than In-Feed. The format is something you test, not something you decree. According to TikTok's “About Spark Ads” page, Spark Ads are built on posts from organic TikTok accounts, so all the views, comments, shares, likes and follows gained during the promotion are attributed to the original organic post.

6. Creative quality

It is the heaviest variable, and the only one nobody puts in their benchmark tables. It deserves its own chapter.

How is CPM formed in the TikTok auction, and why does creative bring it down?

CPM is not a price you pay, it is a score TikTok gives you. The auction is played on an eCPM: your bid multiplied by the probability that the user does what the platform expects of them. Long view, click, purchase, depending on the objective. Two advertisers bidding the same amount therefore don't win the same impression at the same price: the one whose creative holds attention better wins the slot while paying less, because their impression earns the platform more in engagement and time spent.

In concrete terms, the signal that weighs most upstream is the retention of the first three seconds, the hold rate. A user who scrolls before the third second destroys two things: your chance of converting, and your quality score on the next auction. The loop is short. A creative that starts badly sees its CPM degrade within 48 hours, and then its delivery choked off.

Across a portfolio of French campaigns in 2025, creatives in the top quartile for 3 s hold rate cost us 20 to 35% less per thousand than creatives in the bottom quartile, with strictly identical objective, placement and audience. It is the only variable that produces that order of magnitude for us without touching the media.

This mechanism has an uncomfortable consequence for many media teams: raising the bid when the CPM climbs amounts to paying more for a bad score. You are forcibly buying an impression that the algorithm doesn't want to give you at the market price. The lever isn't in Ads Manager, it's in the rushes folder. If you want the detail of the measurement, we have documented it in how a video hook is measured and what it really costs to produce.

A nuance, because it was costly to learn: the lowest CPM is not always the best business. A bait creative, a shock shot, a provocative question, can show an excellent hold rate, crush the CPM, and sell nothing. In 2025 we stopped optimising campaigns on CPM alone. CPM is a creative diagnostic indicator, not a management KPI.

Which figures can you compare yourself with, and which should you ignore?

A CPM benchmark only holds up for your account if it specifies six things: the geography, the currency, the campaign objective, the placement, the collection period and the sample size. If one of these six elements is missing, the figure is an illustration, not a reference. Apply the test to the pages you consult and you'll clear things out very fast.

We also publish a sector grid: if you are looking for orders of magnitude by vertical, go and see our CPM, CPC and ROAS benchmarks by sector in France. The page you are reading answers a different question: why your figure differs from that grid, and how to tell whether the gap is a problem or noise.

The last row is the only one to earn a clear yes. That is not a rhetorical pirouette: it is the logical conclusion of the table. French sources in euros are useful for framing an order of magnitude before a first media plan, provided you cite them with their collection date and not the date the article was updated. American sources are useful for observing relative trends between sectors. Neither tells you whether your €7.30 CPM on a Conversions campaign targeting France 18-34 is good or bad.

An honest note on what we don't know: we have no reliable public data, in France, that isolates CPM by campaign objective on a large, dated sample. Nobody publishes it. Ad sales houses don't disclose that level of granularity, and the aggregators don't segment. Any statement of the type “the Conversions CPM in France is €X” should be treated with suspicion, including when it comes from us.

How do you build your own CPM benchmark in 14 days?

Fourteen days, a dedicated campaign, a single objective, a broad audience and four creatives from different families are enough to produce an in-house reference index more useful than any sector table. The protocol below is the one we run when we open an account. It isn't complicated. It just takes discipline on the variables.

The set-up

• One campaign, one objective. The one that matches your real use. If you buy on Conversions day to day, measure on Conversions. A Reach benchmark will be of no use to you.
• Manual placement, TikTok only. We exclude the audience network for the duration of the test. You can reactivate it afterwards if your figures justify it.
• Three ad sets at most, with the same broad France audience, with no exclusion other than recent buyers. No stacked interests, no 1% lookalike.
• Four creatives from distinct families: a product demonstration, a to-camera testimonial, a problem/solution angle creative, a list or comparison format creative. Four variants of the same script measure nothing.
• Lowest cost bid strategy, with no cost cap, for the first seven days. A cost cap distorts the reading of the CPM.
• Minimum budget held without interruption. Switching off and relaunching resets the learning phase to zero and pollutes the series. On the budget level, see the real entry ticket for testing TikTok Ads. According to the TikTok Ads Manager help centre, performance volatility generally begins to decrease after about 25 campaign results or 7 days from when the campaign enters the learning phase.

The reading

First reading at D+7, second at D+14. You export the CPM per day and per creative, not the aggregated campaign CPM. Then you calculate the median and the interquartile range, never the mean. A single day at a €14 CPM drags the mean upwards and makes you believe in a structural problem that doesn't exist. The median gives your reference level; the interquartile range gives your normal amplitude, the one below which a variation deserves no reaction.

The deliverable fits on one page: an in-house “reference CPM” sheet, by objective and by format, with its date window. Example of a line: Conversions objective, TikTok only, broad France audience, 1 to 14 February 2026, median CPM €6.10, interquartile range €5.20 to €7.40. From there, you know how to read your account. A CPM of €7.10 is no longer an alert, it is the top of your range. A CPM of €11 is.

Redo the exercise every quarter. Your reference moves with your catalogue, your brand awareness and the calendar. A sheet from March 2026 is worth little by November.

What should you do when your CPM climbs: 4 real causes and 3 false culprits?

In the vast majority of cases we audit, “my CPM has exploded” comes down to four causes identifiable in under an hour of analysis. None requires contacting the ad sales house.

• Audience saturation. Frequency rises, the pool of new profiles empties, the auction tightens to reach the last ones. Look at incremental reach day by day: if it flattens, that's where you are.
• Creative wear-out. The hold rate of your winning creative falls, its engagement score degrades, the CPM follows mechanically. We won't go back over the subject here: the thresholds, curves and misreadings are covered in ad fatigue thresholds and false diagnoses.
• Seasonal competition. If your CPM rises by 30% in the third week of November, the problem isn't on your side. Check that the rise affects all your ad sets simultaneously: that is the signature of a market effect.
• Unintentional narrowing of targeting. An exclusion added, a lookalike tightened, an age bracket cut “to improve quality”. The CPM shoots up within 72 hours. Look at the ad set's change history before looking elsewhere.

Now the three false culprits. “The algorithm has changed”: perhaps, but you have no way of checking and no action you can take on it. It doesn't deserve a meeting. “TikTok has become expensive”: the absolute level of the CPM only matters relative to your CPA and your margin. A platform at a €9 CPM that delivers a €14 cost per acquisition is cheaper than a platform at a €4 CPM that delivers €31. “We need to raise the bid”: it is the costliest reflex in the market. Raising the bid on a fatigued creative amounts to paying full price for an impression that the algorithm is already billing you with a penalty.

The order of intervention we apply: first broaden the audience, then inject new creatives, and only then touch the bids. In that order. We rarely reverse it.

What is your real cost per 1,000 useful impressions, media and creative included?

The only figure that commits your P&L is the loaded CPM: media spend plus production cost amortised over the asset's real lifespan, divided by the impressions served. Formula: (media spend + amortised production cost) / impressions × 1,000. Nobody calculates it, and that's a shame, because it often overturns the “let's produce cheaper” trade-off.

Here is a simulation built on our production orders of magnitude and CPMs observed on French e-commerce accounts. These are not published data, they are working assumptions consistent with one another.

Two lessons come out of this table. The first: the creative that is most expensive per unit produces the lowest cost per thousand, because it lives longer and buys its inventory more cheaply. The second, less intuitive: the same creative becomes the worst of the four if it is cut too early. Production cost is only amortised over served volume. Killing an asset at 10 days “because the week's CPA was average” destroys the profitability of the production.

To feed the “production cost” column with your own figures, the grid is here: cost per creative by volume. And to size the number of creatives needed for the throughput your media budget demands, the calculation of the creative budget by media tier gives you the rule of three.

Is CPM still the right steering indicator in 2026?

CPM remains the best thermometer of an account's creative health, and the worst objective to optimise. It measures the price of the attention bought, not the value of that attention. A serious media buyer reads it as a diagnostic, alongside three other metrics.

• Cost per 2- or 6-second view, which corrects the illusion of the low CPM: a thousand impressions scrolled past in 0.8 seconds are worth nothing.
• The 3 s hold rate per creative, which is the upstream variable of the CPM and the only short-term lever.
• CPA and blended ROAS, which settle things in the last resort. If the blended figure holds, a rising CPM is not a problem.

We still see management committees where the TikTok CPM in France is presented as a monthly KPI, with a reduction target. That is a bad incentive: the media team will get the reduction by reactivating the audience network and broadening to the extreme, and you will pay the bill in sales volume three months later. Put the CPM in the “diagnostic” column, next to frequency. Not in the “objective” column.

What we would say to an e-commerce director opening their first account in 2026: spend half a day building your reference sheet following the 14-day protocol, then stop looking for TikTok Ads CPM benchmarks for France on the internet. The figure that matters to you isn't published, it is in your Ads Manager, and it will move the day your third second of video becomes better than your competitor's. Everything else is reporting decoration.

Frequently asked questions

What is the average TikTok Ads CPM in France in 2026?

No French public source publishes an average CPM segmented by objective on a large, dated sample. The figures in circulation come from Anglo-Saxon tools, in dollars, on mostly American panels. Build your own reference over 14 days, as a median and interquartile range, at a constant objective and placement.

Why did my TikTok CPM suddenly rise?

Four causes cover almost every case: audience saturation, creative wear-out, seasonal competition and unintentional narrowing of targeting. First check the ad set's change history, then the 3 s hold rate of your creatives. Raising the bid is the costliest and least effective reflex.

Is a low CPM always a good sign?

No. A very low CPM often signals delivery outside the TikTok feed, on the audience network, where impressions cost little and convert poorly. It can also come from a high-hold-rate creative that sells nothing. Always read the CPM alongside the cost per acquisition and blended ROAS.

Can you compare your CPM with that of another French advertiser?

Only if the six variables are aligned: campaign objective, placement, audience size, period, Spark or In-Feed format, and creative quality level. Two accounts in the same sector with different objectives commonly show a gap of one to three. Without that alignment, the comparison has no operational value.

How is the loaded CPM calculated?

Loaded CPM = (media spend + amortised production cost) divided by impressions, multiplied by 1,000. Amortisation is spread over the asset's real lifespan, not a theoretical one. A more expensive creative that runs for six weeks often shows a lower cost per thousand than a cheap creative killed in ten days.

Should you exclude Pangle and the audience network to benchmark?

During the measurement phase, yes, without hesitation. Mixing TikTok inventory and the external network makes your CPM unreadable and incomparable from one week to the next. Once your reference is established on manual TikTok-only placement, you can test reactivating it and measure its real effect on the cost per acquisition.