How to Get Inquiry Attribution Right: First Touch, Last Touch, and 12 Click IDs

Bottom line first: Attribution isn’t about “picking a model” — it’s about “figuring out where the data comes from first”. Most overseas sites are working with attribution data that’s simply wrong: same-domain redirects aren’t filtered out, direct visits take the first-touch slot, and Click IDs expire and get lost.Build a model on bad data and your conclusions will always be wrong.

1. The First-Principles Problem: Many Sites Have Attribution Data That’s Dirty from the Source

Here are four common real-world errors, all of which occur before you “pick a model”:

Mistake	Performance	Consequence
Same-domain redirects not filtered out	A customer jumps from an article page to a product page and it’s recorded as a “new source”	Organic traffic gets inflated and double-counted
Direct visits take the first-touch slot	A customer’s first visit was through Google, but first touch is recorded as “direct visit”	All first-touch channels are distorted
Click IDs expire and get lost	A conversion happens 30 days after an ad click, and the gclid is already gone	Ad performance gets underestimated
Treating visit environment as a channel	WeChat’s in-app browser is treated as “channel = WeChat”	Paid channels and WeChat get mixed together, making it impossible to tell what’s actually working

Together, these four problems make an attribution report “look complete but actually be unusable”. So the correct order is:Fix the data source first, then pick a model. Below we break it down into three layers.

2. Layer 1: Identify “Source” Accurately

2.1 Priority for Determining Source

The source of a visit should be determined in the following order,and the order cannot be reversed:

Priority	Basis for determination	Description
①	Click ID parameters	Most reliable (gclid / fbclid, etc.) — attribute directly when present
②	UTM parameters	Manually tagged, second most reliable
③	Referrer	Exclusion rules required, see below
④	Direct visit	Fallback

Why does Click ID take priority over UTM? Because UTMs are added manually and are easy to mistype, miss, or tamper with; Click IDs are appended automatically by the ad platform, so they are unique and cannot be forged.

2.2 Three referrer exclusion rules

Referrer is the most error-prone part. Three exclusion rules you must apply:

Rule	How to do it	Why
Exclude same domain	When the referrer domain == your own domain, treat it as an internal navigation and do not overwrite the source	Otherwise internal clicks will keep “washing away” the real source
Normalize search engines	Group Google domains from different countries into one source	`google.de`/`google.co.uk`/`google.com` Should be unified as Google
Classify invalid referrers as direct visits	Empty referrer, from your own app, from an email client	Email client referrers are usually empty and easily misjudged

Same-domain exclusion is the most important of the three. When a customer clicks from your blog post to a product page, if it is recorded as “new source = blog post”, their original source (say, Google search) is lost. Over time, reports show a false picture like “the blog is the biggest source”.

2.3 Rules for writing the first-touch slot

First Touch is written only when it actually brings back a source. This rule looks like a detail, but it determines the accuracy of the entire attribution system.

Scenario	Write first touch?	Reason
First visit comes from Google search	✅ Write	Has a clear source
First visit comes from an ad click (with gclid)	✅ Write	Has a clear source
First visit is a direct visit (typing the URL / bookmark)	❌ Do not write	No source information, cannot occupy the first-touch slot
First visit comes from internal navigation	❌ Do not write	Not a real source

Why doesn’t a direct visit occupy the first-touch slot? Because if it were allowed to occupy the slot, a customer who arrives via a brand keyword search (likely first recorded as a direct visit) would permanently lose their real first touch. In reality, a “direct visit” is often the effect of some earlier marketing effort— it’s just that the source of that effort wasn’t recorded.

The correct approach: leave the first-touch slot empty and wait for the first visit that brings back a real source to fill it.

3. 12 Click IDs and the 90-day retention period

A Click ID is a unique identifier appended by the ad platform to the landing page URL. They are the only reliable means of cross-platform attribution, because only they can tell you “which platform, which click, which ad this visit came from”.

#	Parameter name	Platform	Typical validity period	Purpose
1	`gclid`	Google Ads	30–90 days	Search ad attribution
2	`wbraid`	Google Ads (iOS)	30–90 days	Alternative under iOS privacy restrictions
3	`gbraid`	Google Ads (iOS App)	30–90 days	In-app WebView scenario
4	`fbclid`	Meta	90 days	Facebook / Instagram ads
5	`ttclid`	TikTok	90 days	TikTok ads
6	`li_fat_id`	LinkedIn	30 days	LinkedIn ads
7	`msclkid`	Microsoft Ads	90 days	Bing ads
8	`twclid`	X (Twitter)	30 days	X ads
9	`yclid`	Yandex	90 days	Yandex ads
10	`epik`	Pinterest	30 days	Pinterest ads
11	`s_kwcid`	Adobe / some platforms	30–90 days	Enterprise Ad Tracking
12	`irclickid`	Affiliate Marketing	30–90 days	Affiliate Channel Attribution

Three key design decisions:

Decision	How to do it	Why
Store separately	Each of the 12 ID types gets its own field, never merged into a single “click_id”	One user may carry several IDs at once, and each needs its own attribution
Keep for 90 days	Save for 90 days after first capture	Covers long B2B decision cycles (30–60 days is common)
First capture wins	For each ID type, only the first is recorded; later ones don’t overwrite it	Keeps attribution stable and avoids last-click bias

The 90-day retention decision matters especially for B2B. For industrial equipment, 45–70 days commonly pass between the first ad click and the final inquiry submission. If Click IDs are kept for only 30 days, more than half of conversions fail to attribute, and ad performance is badly understated.

4. Layer Two: Four Attribution Models and How to Choose

Fix the data sources first, then worry about models.

Model	Logic	Use cases	Limitation
First Touch	100% credit to the first touch	Evaluate “who brought in new customers”	Understates the channels that close the deal
Last Touch	100% credit to the last touch	Evaluate “who drove the conversion”	Gives direct visits and brand terms a huge advantage
Last Non-Direct	Credit to the last non-direct visit	Recommended default(same as GA4 default)	Requires handling direct visits correctly
Linear / Position-Based	Split credit across multiple touchpoints	Fine-tuning once you have many touchpoints	Requires a longer data history

Why is Last Non-Direct the recommended default?

Because pure Last Touch has a serious flaw:the last step before many conversions is a “direct visit”(the customer has already decided to contact you and opens your site to fill in the form). Counted as Last Touch, all those conversions go to “direct visit,” making every channel look useless.

The right approach: trace back to the last touch that carried a source. This is the standard used by GA4 and most modern analytics tools.last_non_direct standard.

One practical tip: look at both at once.

Which one to look at	What question it answers
First Touch	Which channel brings in new customers? (In B2B, usually content and organic search)
Last Non-Direct	Which channel drives conversions? (Usually brand terms and retargeting)

Looking at only one will always mislead you. The classic mistake: looking only at Last Touch and concluding “SEO doesn’t work, ads work best” — because ads are often the last click before conversion. In reality,SEO content built awareness first; the ad just got a free ride.

5. Layer Three: WeChat Is a “Visit Environment,” Not a “Channel”

This is often confused in Chinese go-global scenarios and deserves its own explanation.

The built-in browser in WeChat (including WeCom) is a “visit environment,” which is an orthogonal dimension to “channel.”

Dimension	Example values	Questions answered
Channel	Organic search / paid ads / social / direct visit / email	Where did they come from?
Visit environment	WeChat / WeCom / regular browser / in-app WebView	What environment are they visiting from?

Why keep them separate? Here’s a concrete example:

Scenario	Channel	Visit environment	Correct reading
A customer opens a link a friend shared in WeChat	Social (share)	WeChat	WeChat is the carrier; the channel is social
A customer opens your ad in WeChat	Paid advertising	WeChat	Paid gets top priority and can’t be overridden by WeChat
A customer types your URL directly into WeChat	Direct visit	WeChat	Environment is WeChat, channel is direct
A customer opens the document you sent in WeCom	Email / IM send	WeCom	Environment is WeCom

Mix them together and you get two kinds of errors:

What not to do	Consequence
Using “WeChat” as the channel value	Paid ad results get filed under “WeChat,” so you can’t measure your ad spend
Ignoring the visit environment dimension	You don’t know the customer’s visit environment, so you can’t improve the share experience or adapt the landing page

The right approach: store channel and visit environment in separate fields, give paid channels top priority (never overridden by environment), and keep WeChat as its own field alongside them. That way you can measure both “how many inquiries did ads bring in” and “how many of those came from WeChat.”

6. Visitor attribution: how to determine country and city

Attribution isn’t just about knowing where they came from, but which country they came from — that’s baseline data for any overseas business. There are two modes here, each with trade-offs:

Model	Rationale	Accuracy	External request	Best for
Lightweight mode (default)	Country header from the CDN + browser timezone inference	Country (timezone can help)	zero	Most overseas sites
Precise mode	IP lookup API configured by the site owner	Country + city	Requires API configuration	When you need city-level data

Why is lightweight mode the default? Because one visitor session triggers multiple resource requests, and if every request hits the IP database, you run into three problems:Slower responses, an external dependency (if the API goes down, everything goes down), and privacy compliance risk (sending visitor IPs to a third party). Lightweight mode uses the country info the CDN already puts in the response headers, at zero extra cost.

When do you need precise mode? When you need to answer questions like “which city sends the most inquiries, and should we do local outreach or attend a trade show there?” That’s when configuring the API is worth it.

7. Common mistakes and acceptance checklist

Seven common mistakes

#	Mistake	Consequence	Fix
1	Same-domain referrer not excluded	Internal navigation overwrites the real source	Add a same-domain exclusion rule
2	Direct visits take the first-touch slot	First-touch channel distorted	First touch only records visits with a source
3	Click ID stored for only 30 days	Long B2B conversion cycles lose attribution	Extend to 90 days
4	Click ID merged into one field	Attribution breaks when multiple channels coexist	Store 12 separate fields
5	Looking only at Last Touch	Underestimates the role of content and SEO	Look at both First and Last Non-Direct
6	Treating WeChat as a channel	Paid results get buried	Store channel and environment orthogonally
7	Recording attribution with client-side scripts	Privacy settings / ad blockers cause missing data	Record key data server-side

Launch Acceptance Checklist

Checklist	How to verify	Pass criteria
Same-domain navigation does not overwrite the source	Jump from inner page A to inner page B and check whether the source changes	Source stays the same
First touch is only written when a source is present	Visit as a new visitor with no source and check the first-touch field	First touch is empty, then written after a second visit with a source
Click ID is captured correctly	Visit using a link with `?gclid=test123` link visit	test123 is visible in the database
Click ID retention period	Simulate a visit 60 days later	Attribution still matches
Channel and environment fields are kept separate	Visit inside WeChat with a link containing `?gclid=` link	Channel = paid, environment = WeChat
First visit is linked to conversion	Submit a test inquiry	The inquiry record contains the full source chain
Data can be exported	Export inquiry data from the last 30 days	Includes source, Click ID, country and environment fields

FAQ

Attribution data can only ever be “best effort” — will it never be accurate?

It won’t be, and you have to accept that.Attribution can never be 100% accurate. The reasons include: cross-device behavior can’t be fully tracked, privacy settings cause signal loss, and offline touchpoints (trade shows, phone calls) can’t be recorded automatically. The achievable goal isn’t “absolute accuracy” but “consistent and comparable definitions”— as long as every channel is measured by the same rules, trend analysis is reliable.The real danger is inconsistent definitions: rule A this month, rule B next month, and the data can no longer be read as a continuous series.

A customer’s first touchpoint was an offline trade show. How do you record that?

Offline touchpoints can’t be captured automatically, but they can be entered manually. We recommend keeping a “How did you hear about us?” field in the inquiry form (optional) and having sales staff confirm the source during phone calls and enter it into the CRM by hand.Coverage for this kind of data is usually only 30–50%, but its value lies in correcting the bias in online attribution— if you find that many customers actually came from trade shows, you won’t mistakenly credit all the results to online channels.

Why not just use Google Analytics attribution?

GA is powerful, but it has two specific shortcomings in the overseas B2B context. First,it mainly serves “website analytics,” not “inquiry attribution”— the sales follow-up and closed deals after an inquiry is submitted aren’t in GA, so the loop can’t be closed. Second,the data sits in your GA account, while inquiry data needs to connect to your CRM. In practice, the sensible approach is: use GA to look at traffic and behavior, and use a separate tracking system to record inquiry attribution (written server-side to your own database), with both aligned to the same UTM definitions.

Issue	Which one to look at
Where do new customers come from?	First Touch
Which channel drives the conversion?	Last Non-Direct
Which channels appear at both ends?	Compare the two — channels with high overlap have the most reliable value

When First Touch and Last Touch conflict, which one should you trust?

Don’t “pick one to trust” — look at what question each one answers:Overlap is a very useful criterion. If a channel shows up frequently as both First and Last, it’s bringing in new customers and closing them — a core channel worth more investment.

Where is the Click ID stored? Are there privacy compliance issues?

A Click ID is an anonymous identifier generated by the ad platform. It contains no personal information, so it counts as technical data at the parameter level, which carries low compliance risk. But two things need attention: first,where it’s stored(a server-side database is more reliable than a front-end cookie and makes access control easier); second,how it’s linked to personal data(if the Click ID sits in the same table as customer names and emails, that’s a personal-data link, so you need to comply with GDPR and similar regulations and provide a deletion mechanism). Our recommendation:store technical data and personal data in separate tables, linked by an internal ID.

Channel type	Recommended attribution window
Paid search (brand terms)	30 days
Paid search (generic terms)	60–90 days
Social ads	30–90 days
Organic search / content	90+ days (better to skip a hard window and keep it long-term)
Trade shows / offline	Manual entry, no window limit

B2B sales cycles are long — how long should the attribution window be?

Based on real industry data, we suggest:The core principle: the window should be close to “the typical time from first touch to conversion.” For categories like industrial equipment with a 6–12 month decision cycle, a 90-day window is still too short — so we suggest also keeping a permanent record of the first touch, so that even beyond the attribution window it can still be used to analyze a channel’s long-term value.

Final thoughts

The value of attribution isn’t in “how precisely you calculate it” — it’s that it decides where your next budget goes.

If the data is dirty, the conclusions are wrong; wrong conclusions send budget to channels that “look effective but are really just taking a free ride.” Meanwhile the path that’s actually bringing in new customers (often content and organic search) gets cut because it shows no immediate return —this is the most common misallocation of resources for companies going global.

So the right order for attribution is:clean up the data sources first (same-domain exclusion, first-touch rules, Click ID retention period, separating channel from environment), then talk about model choice. The former is an engineering problem — get it right once and you benefit long-term; the latter is an analysis problem — you can keep adjusting it as your understanding evolves.