This page is the follow-up to a test of pages you run on your own website about the SourceX referral program. SourceX's own pages are not part of it, and nothing in it involves collecting company records.
What a finished test leaves behind on your site
When a page test ends, the site still carries its scaffolding: two or more live pages that state the program's facts in different words, temporary redirects and canonical tags added to split traffic, outreach templates and email signatures pointing at each version, and a log of which version each CEO's request came through. Google's testing guidance says to run a test only as long as it is needed and, once it concludes, to update the site and remove test URLs, scripts and markup as soon as possible (our paraphrase).
A referral page has a second reason to tidy up. Every live version is one more place where an outdated sentence can outlive a change to SourceX's terms, and section 11 of those terms records five versions dated between 12 and 22 October 2026. Two pages need two corrections each time; one page needs one.
Intent first, result second
Two pages share an intent only when four things match. If any one differs, the pages are not competing for the same reader, and a gap between their numbers describes the readers, not the pages.
| Intent field | Same when | Different when (examples) |
|---|---|---|
| The reader who arrives | Both pages draw the same people from the same outreach | One serves portfolio CEOs, the other a regional association's members |
| The question they bring | Both answer "could this concern my company?" | One explains the program, the other answers a finance lead's question about who pays |
| The person who decides | Both name the company's CEO, as the program requires | Never different: a version that names anyone else as the decider fails before intent is weighed |
| The single action offered | Both end in "ask for an introduction, apply directly or decline" | One offers a route to the CEO, the other an application |
Look at intent before the numbers: once a smaller page has "lost", it is hard not to merge it, even when it was the only page written for readers the larger page never reached.
The intent-and-result consolidation decision
Three gates, always in this order, then a table.
- Has the reading date passed? Set it before launch, late enough that every 90-day attribution window in the test has closed. Before that date, change nothing: merging mid-count alters what readers see while you are still counting and blurs which version a request came through.
- Are the guards clean? A guard is a harm the test watches for: an application started without the CEO's yes, records or system details sent to you, readers asking "so we qualify?", complaints or requests not to be contacted. If any guard tripped on a page, fix or retire that page first and do not read its primary count as a win for anyone.
- What do intent and result say together?
| Intent | Result on the reading date | Decision | What happens to the retired URL |
|---|---|---|---|
| Same | Gap in qualified applications clears your preset threshold | Merge into the better page; port only factual corrections from the other | Temporary redirect to the survivor during the watch, permanent once it passes |
| Same | Gap inside the threshold, or an interval that includes zero | Keep the original page; retire the variant | As above |
| Same | Too few qualified applications to read, below the floor you set | Keep the original; the variant has not earned its upkeep | As above |
| Different | Both drew CEO requests, whatever the gap | Keep both; compare each with its own earlier period, never with the other | Stays live |
| Different | One page drew almost no CEO requests across the whole test | Keep it under a written minimum-use floor with a review date; retire it if it stays below | Decided at the review |
| Either | A guard tripped | Fix first, then restart from a comprehension check | The faulty page is paused or retired into a clean one |
The threshold is the one you wrote down before launch: the smallest gap that would have made you change pages. If you never wrote one, you cannot choose it now, because a threshold picked after seeing the numbers tends to fit them. Treat that test as showing no clear gap and keep the original. Small counts make this rule bite often; the arithmetic is worked through in how to avoid declaring a winner from five replies.
Reading the gap: qualified applications, understanding and side effects
The decision reads three things per page, all counted only after the reading date.
- Qualified applications. Applications from companies whose CEO asked for the introduction during that page's period, verified by SourceX and credited to your introduction. Visits, clicks and form starts are not counted.
- Understanding. In a short paraphrase check, did readers say the CEO decides, that SourceX decides fit, and who could be paid? A page that wins on applications while readers think their company "qualifies" has not won.
- Side effects. The guard tally above, page by page.
Report the gap with its uncertainty rather than as a single figure. Where a page persuaded fewer people to apply but more of those applications were qualified, the cohort reading in when better screening lowers applications but improves outcomes applies before you call either page better.
Merging without breaking links, credit or claims
- Choose the surviving URL. Usually the original, because outreach, signatures and archived newsletters already point to it.
- Redirect in two steps. Send retired URLs to the survivor with a temporary redirect while the post-merge watch runs, then switch to a permanent one. Google recommends a permanent redirect when a URL change is meant to stick and you are sure it will not be reverted, and notes that a temporary redirect keeps the source page showing (our paraphrase). During the watch you are not yet sure.
- Strip the test markup. Remove the canonical tags that pointed variants at the original, the scripts that split traffic and any experiment parameters in links.
- Update the links you control. Outreach templates, signatures, slide decks, newsletter archives.
- Leave credit alone. SourceX decides credit from dated records, and section 2 of its terms says "an earlier link click alone does not decide credit." A redirect neither preserves nor endangers credit, because credit never rested on the click. Do not tell anyone otherwise.
- Recheck the survivor against the rules every version had to follow. Section 4.5 of the sample Channel Partner Agreement says: "Neither Party may publicly identify the other as a partner or disclose non-public deal details without written authorization, except as legally required." Section 4.6 bars telling a company that an estimate, positive interest, a signature or a data upload assures a purchase or a reward, and section 7.1 bars committing SourceX to prices, licenses or payouts. Keep the line that SourceX pays any reward separately, never out of the company's payout, and keep "may be a fit; SourceX decides after its review" wherever fit comes up. Date-stamp the page with the terms version you read.
- Keep the archive. The retired page's text, the dates it was live and the test log. Deleting them deletes your record of what a reader was told.
Do not stitch the best paragraphs of both versions into one. A hybrid is a new page nobody has tested.
A post-merge watch with tripwires
Run the watch for one full quarter of normal outreach, then wait for the 90-day windows of requests made in that quarter to close before reading qualified applications. Read the guard measures monthly.
| Tripwire on the surviving page | Action |
|---|---|
| An application started without the CEO's yes | Pause the page, rewrite its call to action, restart the watch |
| Records, files or system details sent to you | Pause the page, keep nothing that arrived, fix the wording |
| More "so we qualify?" replies than before the merge | Rewrite the fit paragraph and recheck comprehension |
| More complaints or do-not-contact requests | Review outreach volume and timing before the page itself |
| Qualified applications per CEO request below the lower bound you set | Revert to the pre-merge pages, which the temporary redirect still allows |
When the watch passes with no tripwire, switch the redirects to permanent and close the log with one dated line: which page survived, why, and which threshold decided it.
Three consolidation calls (Illustrative)
Illustrative. Generic partners and pages; no outcomes from any real test.
A consultancy's short and long pages. Both versions were written for CEOs of owner-run logistics firms and reached them through the same newsletter. On the reading date the gap in qualified applications sat inside the threshold the firm had set, and no guard tripped. Same intent, no clear gap: the original short page stays, the long page redirects to it, and two corrections found in the long version are ported across.
A fund platform's portfolio page and LP-network page. The portfolio page drew far more CEO requests; the LP-network page drew few. The readers differ, the relationships differ and the route to each CEO differs, so the pages keep separate lives. The LP page gets a minimum-use floor and a review date instead of a merge.
An advisory firm's finance-lead page. A routing page for finance leads produced two applications started without the CEO's agreement. Guard tripped: the page is retired into the CEO page before anyone looks at its primary count, and the failure is written into the log. The design that produced it is described in the referral landing page test for different decision-makers.
Which companies the surviving page speaks to
Consolidation changes how many pages you keep, not what they may say about fit. The surviving page restates SourceX's published screen in its own words: at least 50 people on the full-time payroll at the company's high point (contractors do not count), several years of records the company produced itself, the legal right to license that material, and a chief executive who has asked for the conversation. Relevance to SourceX's current needs and the rights picture are settled only in SourceX's review. The page never asks readers for records, systems or headcounts; a contact form needs a name, role, email and an optional question. Anyone on your team who submits an introduction signs the SourceX Channel Partner Agreement before the first one goes in.
When consolidating is the wrong call
- The terms changed during the test. Reissue every version with the new date and restart; the old counts compare two descriptions of superseded terms.
- The pages differ by language or region. A page in another language, or one built around a region's location questions, has a different reader, so the intent gate keeps it.
- The test never reached its floor. "Keep the original" is the answer, not "merge into whichever looked ahead".
- Search traffic is the reason. Search visibility is not a referral outcome; let redirects follow the decision, not lead it.
- The merged page grows too long. If the survivor needs paragraphs from the retired page to stay accurate, run the comprehension check again before the watch begins.
A site-level version of a page test, on SourceX's own headline, is laid out in the maximum-payout headline study, and choosing between a personal link and a resource page is covered in how to compare personal links and partner resource pages. More designs sit in the guides hub.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
Once the decision leaves you with one surviving page and a clean watch, check its fit wording against who qualifies and the program terms, and make sure everyone who will submit introductions has registered as a partner first.