How to handle brand effects in MMM, step by step

Usually in five moves. Split brand search from generic search in Google Ads. Collect weekly branded query volume. Give brand video a longer carryover than search. Decide whether branded search volume is a control or a first stage. Then read the baseline's trend and, if your spend allows, anchor brand channels with a lift test.

Pull the weekly sales and spend first; the data check has those menu paths. The steps below add what brand needs on top. The reasons behind them sit in the plain answer.

Step by step

1. Sort every campaign into brand, generic or performance. In Meta Ads Manager, campaigns on the Awareness objective go under brand; Meta's Awareness covers the old Brand awareness, Reach and Video views objectives. Robyn's guide says a model fed both jobs as one channel will likely estimate their average, with less accuracy. Path: Ads Manager, the objective of each campaign.
2. Make the brand search campaign buy only your name. In Google Ads, select Campaigns, then Search terms in the Insights and reports sub-menu. Move any generic terms out of the brand campaign before you export a single week. Path: Campaigns > Insights and reports > Search terms.
3. Model brand and generic search as two channels. Meridian's docs say campaigns on brand queries are very different from generic ones and belong in separate media channels. Brand search is often modeled on clicks, generic search on impressions. Path: Meridian docs, Paid search modeling.
4. Request weekly branded query volume. Google's MMM Data Platform delivers query volume labelled brand or generic, indexed rather than raw. New advertisers request an account through a form; in a new project, tick GQV under Supplemental data signals and choose a weekly time granularity. Path: MMM Data Platform > new project > Data format.
5. Build a stand-in from Google Ads while you wait. On the Campaigns page, segment the brand campaign by Time, then Week, and add Search impr. share from the columns icon under Competitive metrics. Impressions divided by impression share gives the impressions your brand ads were eligible for, a rough weekly gauge of brand searches. Path: Campaigns > segment icon > Time > Week, then columns icon > Competitive metrics.
6. Choose: control or first stage. As a control, branded query volume stops search taking unearned credit, but brand media loses the searches it caused. Meridian's docs say treating query volume as a confounder will more often be right. Its two-stage full-funnel model is the alternative when brand media clearly moves branded search. Path: Meridian docs, Control variables, Including query volume as a control variable.
7. Give brand video a longer carryover. Set binomial decay on brand channels, which Meridian recommends when much of a channel's effect lands in the second half of the window. Its docs suggest a lag of 4 to 20 periods for this decay. Raise max_lag within that range, or set an alpha prior with more mass near one. Path: Meridian docs, Set the adstock_decay_spec parameter.
8. Read the baseline after the first fit. Compare the baseline line in Meridian's model fit charts with the trend you expect for base demand. A baseline that climbs for years may hold brand building the model could not attribute, or plain category growth; the model cannot tell which. Path: Meridian docs, Assess the baseline.
9. Fit twice and compare brand search. Run the model with and without branded query volume as a control. If brand search's ROI drops sharply once the control is in, it was collecting demand that other channels created. Path: your two model runs, side by side.
10. Anchor one brand channel with a lift test. Meta's Conversion Lift guide asks for a campaign started in the past year, with $5,000 USD or more of spend and 500 conversions. Google Ads Conversion Lift needs a Google account representative, and Meridian can turn either result into an ROI prior. Path: Meta Experiments; Meridian docs, Set custom ROI priors using past experiments.

A worked example

Start with one store's export. On Store A's Channels sheet, Direct holds 57.7% of revenue in last click, first click and touched alike. In a mix model those sales have no spend line, so they fall into the baseline unless brand media explains their weekly swings.

On the Channels sheet, Organic Video holds 0.0% of this store's revenue in all three views. Unpaid brand videos can still enter Meridian as organic media, with carryover like paid media but no ROI. Click paths such as this sheet give them nothing to go on.

For illustration, take a different store with three years of weekly data. It runs Meta awareness video, Google brand search and Google generic search.

For illustration, say its brand campaign had 9,000 impressions in a week at a 90% Search impr. share: about 10,000 eligible (9,000 / 0.9). That weekly figure becomes its stand-in for branded query volume.

It then fits the model three ways. The pattern below is invented to show the mechanics, not a result:

Run one flatters brand search, as Meridian's paid search page warns. Run two is the fairer read on search but leaves the video starved. Run three gives the video credit for the searches it drove, if the first stage holds up.

Which run to trust depends on the stand-in series. If branded impressions rise in the weeks after each video burst, run three has something real to model. If the line stays flat through every burst, run two is the safer read, and a lift test on the video comes next.

What to check when the model looks wrong

• The baseline goes negative. Meridian's docs read that as treatment effects getting too much credit. They suggest more meaningful controls, such as query volume.
• Brand search beats every other channel. Meridian's paid search page describes exactly this risk: without query volume, paid search gets overestimated. Add the control before you move any budget.
• The video gets nothing. Check its spend first, since Robyn's guide warns that spend which never varies is hard for a model to read. Then check the decay: Meridian flags geometric decay as weak on long-term effects.
• A longer max_lag did not lift the video's ROI. That can be right. Meridian's docs say a larger max_lag does not necessarily raise ROI estimates.
• The two-stage model swings from run to run. Full-funnel needs measurable confounders for both stages. Meridian's docs say a single-stage model may otherwise be the more reliable choice.

What to do this week

1. Request branded query volume. Submit Google's MMM Data Platform form, then create a project with GQV ticked and a weekly granularity. Pass: the request is in; Google's page says final delivery can take up to three weeks. Fail: no account yet, so start the Google Ads stand-in from step 5 today.
2. Pull the stand-in for your whole modeling window. In Google Ads, segment the brand campaign by week with Search impr. share added. Pass: every week has impressions and a share. Fail: weeks with the campaign paused, which leave holes to fill or explain.
3. Check whether you can run a Meta lift test. Compare your biggest Meta campaign with Meta's Conversion Lift guide. Pass: it meets the guide, so create a test in Experiments. Fail: it does not, so read brand results from the model as a range, not a verdict.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: How to choose the right ad objective in Meta Ads Manager (Meta); An Analyst's Guide to MMM (Meta); About the search terms report (Google); Paid search modeling (Google); Use MMM Data Platform (Google); Use segments in your tables (Google); Get impression share data (Google); Control variables (Google); Full-funnel MMM: unifying upper-funnel and lower-funnel measurement (Google); Set the adstock_decay_spec parameter (Google); Set the max_lag parameter (Google); Assess the baseline (Google); Organic media and non-media treatment variables (Google); About Conversion Lift (Meta); About Conversion Lift (Google); Set custom ROI priors using past experiments (Google); Connect Search Console to Google Analytics (Google)

Related answers

• All MMM & measurement questions
• How does MMM handle brand effects?
• How to check if you have enough data for MMM, step by step
• Marketing mix modeling, explained
• Incrementality testing, explained