Performance and CRM usually work with the same people and different data. The search and social team knows which words and products carry intent this month. The CRM team knows who opens, clicks and buys. In most companies those two lists live in different rooms, and the published evidence on joining them runs almost entirely one way.
The direction everyone measures: CRM into paid media
Google reports an average 26% increase in incremental ROAS for advertisers who connect offline and app data through its Data Manager, measured on Search campaigns bidding to conversion value, and email platforms such as Klaviyo, Mailchimp and ActiveCampaign now connect to it directly (Google). Meta reports 16% lower cost per quality lead for advertisers who linked their CRM through the Conversions API, from A/B tests on 68 advertisers in 2023 (Meta). Both are the platforms’ own averages, so read them as a direction, not a forecast for your account.
The direction almost nobody measures: paid media into email
We found no published, measured result for the other direction, so here is one of ours. A large international consumer-goods group asked its search and social teams for the keyword sets they wanted to push, and used them consistently in its brands’ newsletters for three months: in subject lines, headings, product names and the copy around the links. Over those three months the average bids on those keywords came down by around 27%, and the brands’ search positions rose on the days they sent newsletters carrying those keywords.
It is one programme and a correlation, not a controlled test: the same months also carried seasonality and other campaigns. Our reading, and it is a hypothesis, is that the newsletters sent engaged people to search for and click the same terms and pages the paid team was bidding on. The exact figure matters less than the method: two teams that already have the data can test this in a quarter.
And the answer engines
Scott Dodson, VP of Marketing at Holded, made a related point about AI search at AI Summit Barcelona on 23 September. A question to an AI assistant now runs to about 15 words, against 3.4 for a Google search in the US, so the content that gets cited is content that answers long questions in plain words. He estimated that 40% to 85% of what earns a brand those citations comes from third parties rather than its own site, and closed with one prediction: bet on Substack. A newsletter that is also published as an article on your site is an answer waiting for a long question.
Neil Patel reaches the same place from the content side in his video on GPT-6 Astra (Neil Patel, 9 minutes): when every competitor gets the same model on the same day, speed stops being an advantage, and the scarce thing becomes proof, meaning your own numbers, your own customer stories and a named person taking a position. That is why we share a case of ours here instead of another average.
Before you connect anything: the European check
Sending customer data to an ad platform is processing personal data, hashed or not. The Spanish data protection authority put it plainly in a 2015 report: hashing a customer database before sending it does not make it anonymous (AEPD, report 0364/2015). Google requires both consent signals, ad_user_data and ad_personalization, to be granted for users in the EEA, and does not support IP and timestamp matching in the EEA, the UK or Switzerland (Google). The first direction, keywords flowing into your newsletters, needs none of this: it moves words, not people.
Our perspective: a three-month test
Ask your search and social team for the ten keywords they most want to push this quarter, with their current average bid and position. Use them in the subject lines, headings and link copy of your regular newsletters, and keep a log of the send dates. After three months, compare bids and positions for those ten words against ten they did not hand you. In the other direction, start with one feed: send your customer list to the ad platform as an exclusion, with consent checked, so acquisition budget stops chasing people you already have. All policy-compliant Google Ads accounts can use Customer Match for exclusions.
The test plan puts both directions on one page, and Google’s own Ads Decoded episode on feeding first-party data to bidding (32 minutes) covers the paid side in detail.