Every agency has had this conversation. A client opens their analytics, sees social sitting near the bottom of the channel report with almost no conversions next to it, and asks the reasonable question: why are we paying for this?

It is a fair question asked against unfair evidence. A meaningful share of the traffic that social generates does not arrive labelled as social. It arrives labelled as nothing, and most analytics setups quietly file it under direct, which is the bucket that means “we do not know”.

The useful thing is that this gap can be measured, as long as you have a system that knows the true source independently of what the browser reports.

A way to actually size the gap

Referral and affiliate platforms are one of the few places where you can see both numbers at once. When someone clicks a tracked partner link, the platform knows exactly which partner sent them, because the identity is carried in the link itself rather than in a referrer header that a browser may or may not pass along.

So you can compare what the platform knows against what a general analytics tool would have seen.

Across Shopify stores running referral programmes on Affilitrak, [a quarter of referral visits arrive with no identifiable source](https://affilitrak.com/blog/shopify-affiliate-program-statistics), 25.5% on a store-weighted basis. That makes unattributed traffic the second largest category in the dataset, behind Instagram at 28.8% and ahead of YouTube at 15% and Facebook at 14%.

The figures come from activity across 2,582 Shopify stores, 1,178 programmes and 8,695 partners. Store-weighted means each store counts once regardless of size, so one large retailer cannot pull the average around.

One point of precision, because it matters for what you can honestly tell a client. That 25.5% is traffic with no identifiable referrer, not traffic proven to have come from social. Some of it is dark social, links shared in messages or pasted in from a screenshot. Some of it is email, bookmarks, QR codes and people typing an address they remembered. The data cannot separate them. What it does establish is that the unattributed slice is large enough to change conclusions, not that every visit in it started on Instagram.

Where the source goes missing

None of these journeys is exotic. They are the ordinary ways people share things.

Direct messages: A link sent in a WhatsApp or Instagram DM usually arrives with no referrer attached. This is one of the most common routes a recommendation actually travels, and it is close to invisible.

In-app browsers: Opening a link inside the Instagram, Facebook or LinkedIn app rather than in Safari or Chrome frequently strips or alters the referrer.

Copy and paste: Someone sees a link in a story, types it into their browser hours later, and arrives looking like they always knew the address.

Privacy defaults: iOS and modern browser behaviour removes referrer data in more situations than it used to, and that trend runs one way.

Screenshots: A screenshot shared to a group chat carries no link at all. The recipient searches the brand name and arrives through organic search, or goes direct.

Every one of those started somewhere. None of them will be credited to that place in a default channel report.

Why this is a budget problem, not a reporting quibble

The first consequence is simple miscounting. Direct traffic is not a channel. It is a residue, and a large part of it is other channels with their labels removed.

The second consequence is what the miscounting does to decisions, and it is the expensive one. Budget conversations run on channel reports. If a client sees social at the bottom of the table, money moves to the channels that report cleanly, which usually means paid search. Not because paid search is doing the most work, but because it is the best-instrumented channel on the page. Channels that measure themselves well take budget from channels that do not, regardless of what either contributed.

There is a timing detail worth adding. Among referred orders where the timing could be reconstructed (N = 2,712), 51.4% were placed within five minutes of the referral visit and 79.3% within the hour. When the purchase happens in the same session as the arrival, there is no later touchpoint that might get attributed correctly instead. Lose the source at the door and the whole conversion is misfiled, not just the first step of a longer journey.

Four things that close the gap

Tag every link you control: UTM parameters survive most of the journeys above because they live in the URL itself rather than in a referrer header. A tagged link that gets copied, pasted and retyped still carries its own labelling. Highest return available, and it costs nothing.

Give partners and creators individually tracked links: If you run influencer, referral or affiliate activity for a client, per-partner tracking tells you who sent what regardless of whether the browser preserved anything. That is exactly why the platform in the data above could see traffic a general analytics tool could not. The mechanics of how that tracking works are covered in this guide to tracking affiliate sales on Shopify.

Check your attribution window before you blame it: Most people assume a longer window recovers lost credit. The evidence points the other way, since the overwhelming majority of referred orders land almost immediately, which is the reasoning laid out in this breakdown of how long an affiliate cookie should last. Widening the window is usually solving a problem you do not have while the real leak sits at the point of arrival.

Ask the customer: A single “how did you hear about us?” field at checkout or on an enquiry form is unfashionable, low-tech, and frequently more accurate than the model it is correcting. For an SME with modest volumes it is often the best data source available.

What this does and does not tell you

The measurements here come from Shopify stores running referral programmes, so they describe traffic sent by partners and creators. A service business in Guildford is not a DTC store, and the 25.5% is not a number to quote for your own account.

What carries across is the mechanism rather than the percentage. Referrers go missing because of how browsers, apps and people behave, not because of anything specific to Shopify. Any business whose customers recommend it to each other is losing source data the same way. The value of a measured figure is that it tells you the scale of the problem is material rather than marginal.

Conclusion

The next time a client points at a channel report and asks why social is not converting, the accurate answer is that the report cannot see a meaningful part of what social does. That is not a defence of every social campaign, and some campaigns genuinely are not working.

But a measurement system that structurally undercounts one channel should not be the only evidence in the room when the budget gets decided. Tag your links, track partners individually, ask your customers directly, and then have the conversation with numbers that reflect what actually happened.