How to tell if social media is actually making you sales

The short answer

Below: the study behind the opening number, exactly why Shopify's default hides the channel, a copy-and-paste tagging convention, which of the six relevant reports answers which question, and an honest account of what tagging still can't see.

Where the traffic actually went

"Direct" sounds like it means "people who typed your URL." It doesn't. It means we received no information about where this person came from — an error state that gets displayed as a channel. Rand Fishkin of SparkToro and Steve Lamar of Really Good Data measured how often that happens, by recruiting a panel of around 100 people to click tracking links posted across 11 social networks and recording what arrived. Out of 1,113 visits, this is what the referrer data looked like.

Read the top row again if you sell through short-form video. Every single TikTok visit in that experiment landed as Direct. Not most — all of them. SparkToro's own site, the study notes, shows 95% of its traffic as Direct in Google Analytics, which the authors describe plainly as untrue.

The cause is mundane. A referrer is an HTTP header the browser attaches to a request, and the in-app browsers inside social and messaging apps routinely don't send one. Nothing is broken on your store, and no app will "recover" the missing data after the fact. The information was never sent.

Why Shopify's default makes it worse

Losing the referrer would be survivable if untagged visits simply showed up as an unknown bucket you could estimate around. The problem is what Shopify's attribution does with them next.

Shopify's default model is last non-direct click, which its documentation defines as giving "100% of attribution credit to the last channel that the customer interacts with before making a purchase, excluding direct visits." A direct visit, in Shopify's words, is what you get when there is "no non-direct source to attribute to." Put those two sentences together and the mechanism is clear: an untagged TikTok visit becomes a direct visit, and a direct visit is then excluded from receiving credit. The sale gets handed to whatever else the customer touched — usually a branded Google search, or the email you sent — and TikTok shows a flat zero.

Tag every link. Here's the convention.

UTM parameters are five optional key-value pairs you append to a URL. Shopify's own guidance calls utm_source and utm_medium required if you use them at all, and advises "lowercase letters for all UTM codes, every time, no exceptions" — because TikTok and tiktok become two separate rows in your reports and neither shows the real total.

A worked example, linking straight to a product rather than to your homepage:

Assembled, the TikTok bio link becomes:

Three rules that prevent almost every mess we've seen: keep utm_medium identical across all organic social (social) so the channel rolls up as one line; change utm_campaign monthly so you can compare periods without building a date filter; and vary utm_content by placement so you can tell a bio click from a Story click without creating a new channel for each.

Where the tagged links go

Placement mechanics change more often than the analytics does, so treat this as current-as-of-writing rather than permanent. On TikTok, switching to a Business account removes the follower threshold on the profile website field, which is otherwise the main blocker for a new store. Instagram allows several links in the bio and a link sticker on Story slides, one per slide, with no follower minimum. Facebook Pages take a button destination and links in posts directly.

If you use a link-in-bio page rather than pointing at a product, tag the outbound links on that page instead of trusting the tool's own dashboard. You want the data to land in Shopify, next to the orders, rather than in a fourth analytics product that can't see revenue. Every extra hop between the video and the product page is also a place people drop out, so link straight to the product whenever there's only one thing you want them to see.

Now read the right report

Shopify splits this across two report families, and most merchants only ever open the wrong one. Acquisition reports tell you about sessions. Marketing reports tell you about money, and they only see traffic you've tagged — which is why tagging comes first.

Two limits worth knowing before you go looking for history: Shopify states that session-based metrics in the acquisition reports go back only to 1 October 2022, and attribution data in the Growth section only to 1 October 2021. Tagging is not retroactive either — nothing you do today recovers last quarter's TikTok traffic from the Direct bucket. The clock starts when you tag.

Then change the attribution model on purpose

When a report contains a sales metric alongside a marketing dimension, Shopify shows an attribution menu with five models. They are not alternative opinions about the same number — they answer genuinely different questions, and for social the gap between them is the whole story.

The practical habit: look at your social numbers under last non-direct click and again under first click, and treat the gap between them as the channel's discovery value. If first click shows meaningful revenue that the default hides, social is doing its actual job — starting journeys — and judging it on the default number will lead you to cut something that works.

What tagging still won't show you

Tagging converts the largest category of invisible social traffic into visible traffic. It does not make the channel fully measurable, and any tool promising that is selling something. Everything below survives a perfect UTM setup.

1. The journey that changes device. Someone watches a video on their phone, then buys on a laptop after searching your brand name. There is no click on your tagged link in that sequence, so the sale is recorded as organic search. This is extremely common and systematically flatters search at social's expense.
2. Influence without a click. A video that makes someone remember your brand three weeks later leaves no trace anywhere. The entire top of the funnel behaves this way.
3. Screenshots and word of mouth. A product screenshotted into a group chat, then searched for by name, arrives as direct or organic. The SparkToro data above is partly a measurement of exactly this behaviour.
4. Anything before you started tagging. UTM parameters are recorded at the moment of the visit. There is no backfill.

The only practical instrument for that layer is asking. A post-purchase survey — one question, "How did you hear about us?", on the thank-you page — captures the answer at the one moment the customer is engaged, identified and still remembers. Shopify has no native field for this, so it takes a survey app (Fairing, Zigpoll and Grapevine all build for Shopify's checkout extensions). Ask "how did you first hear about us" if you want discovery data, which is the question social answers; ask "what brought you here today" and you'll get last-touch answers you already have.

Expect the survey and the analytics to disagree, sometimes badly. That disagreement is not a failure — it's the size of the gap the analytics can't see, and it's the most useful number in this whole exercise.

One warning about benchmarks

Once you have numbers, you'll want to know whether they're good, and the internet will offer you a conversion rate to compare against — social converts at 0.7%, or 1.1%, or 1.5%, depending which page you land on. Those figures circulate between blogs without a disclosed sample or method, much like the contradictory posting-time studies we've written about before. Compare your channels against each other, in your own store, over the same period. That comparison is real, and it's the one Shopify's reports are built to give you.

A measurement setup that's proportionate

You do not need an attribution platform to run a store. Five things, in this order, get you from "no idea" to "defensible answer."

1. Tag every social link today. Bio links on all three networks, using the table above. Twenty minutes, and it's the step every later number depends on.
2. Wait a full month before judging anything. You need enough orders for the difference between channels to mean something, and the data only starts accumulating from the day you tag.
3. Read two reports, not eight. Sessions attributed to marketing campaigns for traffic, Sales attributed to marketing for revenue. Ignore the rest until those two are telling you something stable.
4. Check both attribution models. Default, then first click. Write both numbers down. The gap is the discovery value you'd otherwise never see.
5. Add the post-purchase survey once you're posting regularly. Not before — with low order volume the answers are noise. Once you have a steady flow, it's the only thing that sees the part UTMs can't.

And then the harder question

Measurement tells you whether social is working. It doesn't tell you why it isn't. In our experience the answer is usually upstream of analytics entirely: the account posts twice a week, or posts only product photos, or posts nothing at all for a fortnight when the shop gets busy. Attribution can't rescue a channel that isn't running, and a clean UTM setup measuring four posts a month will confidently report that social doesn't work for you.

That's the gap Magik is built for — it generates posts from your Shopify catalogue as creator-style videos and styled product images and publishes them to TikTok, Instagram and Facebook on a schedule, so the channel actually has volume to measure. It doesn't do attribution, and it isn't a substitute for tagging your links: put the UTM parameters in your bio links yourself, using the convention above, and the posts Magik publishes will drive traffic that Shopify can actually see. If you'd rather compare tools first, the comparison page covers Outfy, Xyla AI, IDEQO, Minta and Buffer honestly, including where each beats us.

Related reading while you wait for a month of data: what filling a content calendar actually costs, what to post beyond product shots, and the setup guides for TikTok and Instagram. If you're generating content with AI, the labelling rules are worth ten minutes too.

Frequently asked

Sources

The attribution definitions and report names come from Shopify's own Help Center; the referrer data from the original published study rather than any summary of it. Shopify revises its documentation and admin regularly, so check against your own store if a name doesn't match.

1. SparkToro — New research: dark social falsely attributes significant percentages of web traffic as "direct", Rand Fishkin with Steve Lamar of Really Good Data, April 2023. 1,113 visits from ~100 panel participants across 11 networks and 16 tracking URLs; source of every percentage in the misattribution table.
2. Shopify Help Center — Measuring marketing performance. Source of the five attribution models and their verbatim definitions, the statement that last non-direct click is the default, the rule that traffic carrying utm_campaign is attributed to marketing, the definition of a direct visit, and the 1 October 2021 Growth data boundary.
3. Shopify Help Center — Marketing reports. Source of the report names: Sales attributed to marketing, Sessions attributed to marketing campaigns, Performance by referring channel, Performance by marketing activity, Performance by UTM campaign, Marketing top channel performance.
4. Shopify Help Center — Acquisition reports. Source of the Sessions by referrer description and the 1 October 2022 limit on session-based historical data.
5. Shopify — How to use UTM parameters (and when to use them). Source of the five parameters, the requirement that utm_source and utm_medium be present, and the lowercase-everything naming guidance.
6. Placement mechanics (TikTok Business account profile links, Instagram bio links and Story link stickers) reflect the platforms' current behaviour as of September 2026 and change more often than anything else on this page. Verify in the app before building a process on them.
7. Post-purchase survey tools are named as examples of the category, not as endorsements. We have no commercial relationship with any of them.
8. Published conversion-rate-by-channel benchmarks are deliberately not cited, for the reason given in the text: the figures in circulation have no disclosed sample or methodology.

Keep reading