Marketing and product analytics: why combine them and how to connect the data in 2026

August 8, 2021

Tin Le

Updated on October 19, 2026: Adtek rewrote the entire article with the latest data, platform features and regulations for 2025 to 2026.

Marketing analytics and product analytics answer two halves of the same question. Marketing analytics tells you which channel and which campaign brought a customer in, and what that cost. Product analytics tells you what the customer did after arriving on your website, app or Zalo Mini App: which features they used, where they dropped off, whether they came back for a second purchase. Combining marketing and product analytics means joining these two datasets through one shared user ID, so you know which campaigns bring in customers who stay and buy again, not just which campaigns have the lowest cost per lead.

The original version of this article (2021) was written for mobile app teams. In 2026 the problem is broader, and cheaper to solve. GA4 exports raw event data to BigQuery, Amplitude and Mixpanel can sync data from a data warehouse, and Vietnamese customers move through Zalo (Vietnam’s dominant messaging app), Shopee, TikTok Shop and the brand website before they buy.

Marketing analytics vs product analytics: what is the difference?

	Marketing analytics	Product analytics
Core question	Where customers come from, what they cost, which channel converts	What customers do after arriving, how long they stay, which features keep them
Unit of data	Clicks, sessions, campaigns, cost, leads, first order	Events per user: product views, add to cart, feature use, repeat purchase
Typical metrics	CPC, CPL, CAC, ROAS, conversion rate by channel	Retention rate, activation rate, usage frequency, funnels, value by customer group
Common tools	GA4, Google Ads, Meta Ads Manager, AppsFlyer or Adjust for apps	Amplitude, Mixpanel, PostHog, Firebase
Who uses it	Performance, media and growth teams	Product managers, UX, CRM, customer care
Reporting rhythm	Daily, per campaign	Weekly, monthly, by cohort

Ad reports stop at the first conversion. Product analytics tools see in-product behavior clearly but usually have no idea which ad a user came from or how much that user cost.

A hypothetical example: campaign A has a CPL of 80,000 VND, campaign B 150,000 VND. Looking only at Ads Manager, the media team would move budget to A. If leads from A barely reply to Zalo messages while leads from B close twice as many orders, the decision should be reversed.

What goes wrong when the two data sources stay separate?

Awareness channels get undervalued

A YouTube video or a KOL may bring in few direct orders and still be the first touchpoint for your highest-spending customers. A last-click model will cut that channel’s budget. To see its real role, you need to look at the whole chain of touchpoints with multi-touch attribution and compare it against repeat purchase behavior inside the product.

The product team fixes a problem that does not exist

A high drop-off rate right after install or sign-up is usually blamed on onboarding. If most of those installs came from bots or fraudulent sources, the product team is optimizing for fake users. You can only separate those two explanations by attaching the traffic source to every session.

The ad’s promise does not match the landing page

The ad talks about a membership offer, and the user who taps it lands on the home page or a generic login screen. Deep linking, recommended in the original article, is still the right fix: send people straight to the content the ad promised. In 2026 you can measure the effect of that change with one shared funnel, split by campaign and by ad creative.

Questions only combined marketing and product analytics can answer

Mixpanel’s Warehouse Connectors documentation gives a very practical example: you spent 50,000 USD on a campaign, so are the users it brought in still around a month later? Answering that requires marketing cost data and product retention data in the same table. Similar questions Vietnamese businesses run into:

Which channel brings in customers with the highest rate of a second purchase within 90 days, not just the cheapest first order?
Do customers from TikTok use the product differently from customers who come from Google Search? Do they need different onboarding screens?
Between an ad about price and an ad about benefits, which one brings in customers with higher customer lifetime value (LTV)?
Which leads should go to sales for an immediate call, based on their source plus their behavior on the website (viewing the pricing page, downloading a document, opening a Zalo message)?

Every question above comes back to one technical requirement: the same person has to carry the same identifier in the ad platform, the analytics tool and the CRM.

How to connect the data in 2026: GA4, BigQuery, Amplitude, Mixpanel and CDPs

GA4 and BigQuery export

According to Google Analytics’ official documentation (checked in October 2026), GA4 exports raw event data to BigQuery, including to the free BigQuery sandbox. Standard GA4 properties have a limit of 1 million events per day for the daily export; if a property keeps going over that limit, the export is paused. The streaming export has no event limit but costs an extra 0.05 USD per GB, which is roughly 600,000 events.

There are two reasons to switch on BigQuery export early. First, the user-level data that Explorations reports in standard GA4 rely on is kept for at most 14 months, with a default of 2 months; data exported to BigQuery stays as long as you choose to keep it. Second, the setup documentation only describes exporting from the moment of linking onward (data arrives within 24 hours, one table per day for the previous day), so every month you delay is a month of history missing from your warehouse.

A note for attribution: the streaming tables lack traffic source data for new users, and Google recommends taking that data from the daily export tables.

Bringing the data into Amplitude or Mixpanel

Once the data sits in a warehouse, a product analytics tool can pull from it instead of running a second, parallel tracking setup. Amplitude has a GA4 import via BigQuery: data moves from GA4 to BigQuery, is unloaded into a Google Cloud Storage bucket and is then loaded into Amplitude (Amplitude Docs, October 2026). Mixpanel’s Warehouse Connectors sync from Snowflake, BigQuery, Databricks, Redshift and Postgres; Mixpanel’s documentation lists the feature as a free add-on for organizations on paid event-based plans when they renew or upgrade.

Ad spend, marketplace orders and lead status from the CRM can go into the same warehouse and then sync to the analytics tool as user properties. A product manager who opens a retention report sees the channel and cost columns right there.

CDP or data warehouse?

A warehouse is where data is stored and computed. A CDP handles unifying customer profiles from many sources and pushing segments out to activation channels such as Zalo, email and ad audiences. Small and mid-sized businesses can usually start with GA4, BigQuery and a visualization tool. A CDP is worth the investment once you have several sources of identified customer data and people who actually use segments every week. Buying a CDP before you have agreed on a customer ID across your systems tends to create one more place where disconnected data piles up.

Vietnamese businesses: customer data lives in Zalo, marketplaces and the website

The original article’s model of “ads drive app installs, then measure behavior in the app” only fits some Vietnamese businesses. Most of the customer journey happens on platforms the business does not own:

According to VNG’s Q2 2026 business results (reported by VietnamPlus in July 2026), Zalo had 81.3 million monthly users and about 30,000 active Official Accounts as of June 2026.
According to Metric.vn (reported by VnEconomy in January 2026), 2025 sales on Shopee, TikTok Shop, Lazada and Tiki were estimated at 429.66 trillion VND, up 34.75% from 2024. Shopee held 56.04% of sales and TikTok Shop 41.31% (VnEconomy, January 2026).

For brands that sell mainly on marketplaces, “product data” in practice means order data, repeat purchase rate, reviews and in-store engagement, pulled from exports in the seller center or through the marketplace API. The marketplace decides how much buyer information it hands back. For real estate, education and automotive businesses, the real journey runs through Zalo: the customer leaves a phone number, a consultant messages them, the customer returns to the website to check prices, and only then meets in person.

That is why the phone number is often the most practical join key in Vietnam. From January 1, 2026, Vietnam’s Personal Data Protection Law No. 91/2025/QH15 is in force, recognizing customers’ rights to be informed, to consent, to withdraw consent and to request deletion of their data. If you want to match phone numbers from a form with behavioral data and send them back to ad platforms, you need consent for exactly those purposes, collected on the form itself.

Adtek’s view: 6 steps to set up combined analytics

When doing data-driven marketing for clients, Adtek usually works in the order below. Each step has a concrete output to check before moving to the next.

Pick 3 to 5 business questions, from the list above or from your most recent budget meeting. The questions decide which events you need to measure.
Write a one-page tracking plan. List the key events (sign-up, pricing page view, add to cart, purchase, repeat purchase, Zalo message opened), their parameters and consistent names across web, app and CRM.
Agree on one identifier. Assign a user_id to users who have logged in or left their details, and store the same ID in the CRM. For Zalo and marketplace channels, decide which field you will join on (phone number, order ID, CRM customer ID) and make sure you have customer consent for it.
Turn on GA4 BigQuery export now. Even if you are not using it yet, it preserves history. Check your daily event volume against the 1 million limit.
Bring outside data into the warehouse. Ad spend by day and campaign, marketplace orders, lead status from the CRM. Starting with scheduled file exports is fine; automate later.
Build one report for both teams. A table by monthly cohort: channel, cost, new customers, share still active after 30 and 90 days, cumulative revenue. Review it together every month and write down the decisions this table changed.

Questions to ask before buying another tool

Does the same customer carry the same ID in GA4, the CRM and the order system?
Is GA4 linked to BigQuery yet, and from what date is it exporting?
Can your current reports answer “which channel brings in repeat buyers”? If not, what data is missing?
Does your lead form ask for separate consent for analytics and advertising purposes?

When combining marketing and product analytics, most businesses get stuck at step 3, rarely at the tooling. Neither Amplitude nor Mixpanel can rescue data where every system calls the same customer by a different ID.

Adtek sets up GA4, BigQuery and reports that combine ad, marketplace, Zalo and CRM data. If you want a second look at how you measure today, contact the Adtek team.