Why TikTok’s MMM Study Matters for Organic Growth.

TikTok has spent the past year publishing marketing mix modelling research, and every version reaches the same conclusion: last-click attribution undercounts what TikTok contributes.

Marketing mix modelling, usually shortened to MMM, is a statistical method that estimates how much each marketing input contributed to sales, working from aggregate history rather than individual tracking.

It is ad measurement research, and TikTok commissioned it.

Read it anyway.

The useful part is not the ad numbers. It is what sits underneath them. Real money is going into models built to correct for last-click attribution crediting the final touch and ignoring everything that created the intent. Organic social has been on the wrong side of that arithmetic for a decade.

What marketing mix modelling actually is

MMM takes a history of everything a business did and everything that happened to it, then estimates how much each input contributed to the result. Spend by channel, price, promotions, distribution, seasonality, competitor activity, the wider economy. The output is an estimated contribution per channel with a margin of error attached.

It runs on aggregate data, so it needs no cookies, no pixels and no device identifiers. That is why it came back into fashion as user-level tracking got harder.

It is also old, out of packaged goods econometrics, and it answers a different question from a conversion report: not who bought, but what moved the total.

What TikTok’s research says, and who commissioned it

In December 2025 TikTok published a study run with NIQ that modelled two beauty brands in Romania’s face care category, using store-level sales data from January 2023 to December 2024. It found TikTok delivered 29% to 33% of all media-driven incremental sales for both brands, and concluded both were investing at around half the level the model treated as optimal.

A Kochava analysis of major Android and iOS apps in North America across the first quarter of 2025 found MMM credited TikTok campaigns with an average of 35% more incremental impact than last-touch reporting. A Monks meta-analysis with TikTok’s marketing science team, covering ecommerce and retail brands over 2023 and 2024, put the gap at 50% or more.

Notice the pattern. A platform funds research into how its own channel is measured, and the research finds the channel undervalued. Kochava holds a badge in TikTok’s own MMM partner programme.

Notice the samples too. Two brands, one category, one country. Mobile apps in one region over one quarter. Monks does not publish how many brands, or which markets. Directional findings from narrow slices, none of which tell you anything about your business.

What last-click and MMM can each see

Neither measures what your platform dashboard shows you, which is why the metrics that actually matter for brand growth are rarely the ones it puts first.

Last-click attribution is precise about the wrong thing. It follows identifiable people and hands the whole conversion to the final touch.

• It sees the click that closed, never the exposure that created the intent.
• It is blind to word of mouth, group chats and anything a privacy setting hides.
• It gets more confident the shorter the journey, which flatters channels that catch demand and penalises channels that create it.

MMM has the opposite shape, and sees everything at a blurry resolution.

• It can include offline and untrackable activity, because it works on totals.
• It cannot tell you who bought, so it settles no argument about a single campaign.
• Two modellers given the same data can return different answers.

Why organic gets undercredited by both

Someone watches your content for eight months. They never click a link, because there was nothing to click. When they finally decide to buy, they type your brand name into Google, land on your homepage and convert.

The report credits branded search, or a paid search line bidding on your own name. The eight months that caused the search get nothing.

MMM does not automatically fix that. Most models are built from media spend, because spend is the variable a business controls. Organic social has no spend line, so unless someone deliberately adds an activity variable, posting volume, organic reach, watch time, it never appears as a channel at all. It gets absorbed into the baseline, the sales the model assumes you would have made anyway.

Read that TikTok figure again with this in mind. 29% to 33% is a share of media-driven incremental sales, not a share of the business.

Notice where the measurement lives, too. Inside an ads manager, keyed to a pixel, available to accounts running campaigns. TikTok’s Attribution Portfolio, announced in May 2026, put assisted conversion and time to conversion reporting on exactly those terms.

Nobody is building the equivalent for your organic feed. If you want organic in the numbers, someone on your side has to put it there, which is a forecasting decision before it is a measurement one, and the work behind NBK’s revenue and performance forecasting.

What an organic team can measure without a modelling budget

You do not need an econometrician to make the case. You need a handful of signals tracked over months rather than weeks, and the discipline to keep reporting them when they are flat.

• Branded search volume. If the content is working, more people look for you by name, and Search Console shows it for nothing.
• Direct traffic. People who type your name or open a saved link were sold somewhere else first.
• Assisted conversions. Path reports show which channels appeared before the closing touch. Not proof, but the cheapest correction to last-click you have.
• Share of voice where customers ask questions. Reddit threads, forums, comment sections and review sites. Count the mentions and read the wording.
• The self-reported source. Put “how did you hear about us” on checkout, on the enquiry form and in the first sales call, and log it as a field, not a note.

That last one is the most undervalued instrument in marketing. It is imprecise and biased towards recent memory, and still tells you more than any conversion report you own.

Change one thing on purpose and watch

MMM’s serious cousin is the holdout: change one input deliberately, hold everything else steady, and watch what the outcome does.

Almost nobody in organic social does it, because everything changes at once. A new format launches the same month the team lifts cadence and a product goes on promotion, and nobody can separate the causes afterwards.

The organic version is not complicated. Stop a format for four weeks, or double one format in one market and leave another alone. Then watch branded search, direct traffic and enquiries for the following eight weeks, not the following three days.

It will not be clean. It will still be the strongest evidence you have, because you caused it on purpose.

What belongs in the monthly report

The report you send upward decides what gets funded. If it carries only reach and engagement, organic keeps being judged as an awareness cost, and the case has to be rebuilt from scratch every quarter.

Put these in every month, each with a twelve month trailing line behind it:

• Branded search impressions and clicks.
• Direct traffic sessions.
• Self-reported source, as a share of new customers.
• Assisted conversions where social appeared in the path.
• Every deliberate change you made, dated, so a shift three months later has something to point at.

None of that needs a modelling budget or an invented figure. It needs the same numbers, in the same order, for long enough that the trend becomes the argument. That is what turns aligning social with business goals into a habit rather than a slide in a strategy deck.

How NBK thinks about measuring organic

Measurement is an operations problem before it is an analytics one. The teams that can prove organic works are rarely the ones with better tools. They are the ones whose reporting has been consistent long enough for a trend to exist.

Our background is publisher-side, where content decisions were defended with numbers every week: more than 600 million views a month, 46,000 posts shipped, 25 years of combined senior experience across UNILAD, LADbible Group, SPORF and Social Chain. The lesson is dull. Decide what you count, count it the same way every month, and never let a good month rewrite the definition.

Next step

If your social output looks healthy but its value is still an argument every quarter, the constraint is usually the reporting, not the content. NBK can help build the system that settles it.