Shopify Sidekick Pulse: Which Recommendations to Act On

Sidekick Pulse is a set of up to five recommendation cards on your Shopify admin Home page, drawn from your store data and market trends. Act on a card when it ties to a number you can see in your reports, costs little to try and could move revenue. Skip cards that read as generic advice.

That rule needs a method behind it, or the cards become either homework you never finish or advice you follow blindly. Give each card two quick scores, effort and revenue, check its claim against one report before you act, and dismiss the rest with a short reason. Shopify says Pulse is available only to some stores with recent sales activity, so your Home page may show fewer cards than described here, or none.

Follow me on Instagram@sagnikteaches

What Pulse actually shows you

Shopify's changelog post of 10 December 2025 says Pulse "proactively surfaces up to 5 tailored recommendations", so you do not have to write a prompt first, and that it rolled out over the following weeks. Sidekick researches your store's data, market trends and performance, and each card carries citations so you can check the underlying data. When you act on one, Pulse passes the context to Sidekick so the task can start straight away. Two changelog posts of 28 October 2026 widened the inputs: Sidekick can now use web search for market and industry context, and it can draw on data from installed apps that have supported Sidekick app extensions. Everything here was read on Shopify's help page and changelog in November 2026.

Connect on LinkedInSagnik Bhattacharya

From the help page:

Subscribe on YouTube@sagnikteaches

• Cards sit on the Home page of the admin.
• Selecting a card's action opens a Sidekick conversation and builds a to-do list. You review the list and follow the prompts, and the conversation stays in your Sidekick history.
• Some theme-change cards are tested first. Sidekick builds the change into an unpublished copy of your live theme, and SimGym compares it with the live theme in simulated shopping sessions. A higher add-to-cart rate earns a Tested by Sidekick label, and some cards show an Add-to-cart rate badge.
• A card may offer Review suggested changes without the tested label, which means the test found no improvement. Clicking it creates a draft theme named "[Sidekick] Suggested theme" plus a timestamp. Your live theme does not change until you publish it.
• To dismiss a card, click the X. You can say why it was not relevant, and Sidekick uses that to personalise later cards.

The caveats are worth reading twice. Sidekick tests only desktop changes aimed at add-to-cart behaviour, and only in stores that can use Canvas with their live theme. Shopify's changelog adds that these simulations are not tested on real shoppers and are not live A/B tests. The test result describes the tested copy, not any edits you make afterwards. The draft opens only in Canvas, so Canvas limits apply. And the help page does not say how often the cards refresh or exactly which data each card used. Those gaps matter for how much weight you give a card.

Two scores per card: effort and revenue

Before anything else, put each card on a one-line scorecard. Use two numbers from one to five.

Effort is how long it takes you or your team, including checking the result. A card that edits one product description scores a 1. A card that asks you to restructure your navigation, rewrite twenty collection pages or change shipping terms scores a 4 or 5.

Revenue potential is how much money the change could plausibly move, judged against your own numbers and not the card's wording. A change to a page that gets 40 visits a month scores a 1, however good the idea. A change to the product page behind half your sales scores a 5.

Then divide: revenue over effort. Anything with a ratio above 2 is a candidate to do this week. Anything under 1 waits or gets dismissed. The scoring is rough on purpose; the point is to stop ranking by how interesting a card sounds.

The verification step: every card meets a report

A card is a claim about your store. Treat it like a claim from a colleague: ask what number it rests on, and find that number yourself. Cards carry source citations, so the first move is to open them and see what kind of source each one is. A citation to your own sales or sessions is evidence about your store; a citation to a web page about your market, which Sidekick can now use, is evidence about other people's stores, and it only tells you what might be worth checking. The second move is to look at the matching report in the admin's analytics area, and to check it against your own sense of the business.

A short procedure works for most cards:

1. Write the claim in one sentence, in your own words.
2. Find the figure it depends on: sessions to a page, conversion rate, add-to-cart rate, average order value, return rate or stock level.
3. Pull that figure for the last 30 days and the same period a month earlier, from your reports.
4. Ask whether the card's claim is consistent. If it says a page converts poorly, does the number show that?
5. Decide what result would convince you the change worked, and by when.

For a theme card with a "Tested by Sidekick" label, add a step: remember that the test is a simulation, on desktop, aimed at add-to-cart. It says nothing about phones, and in many stores most visits come from phones. Look at your mobile share of sessions before you give a desktop test much weight. A card that raises desktop add-to-cart in a store with 80% mobile traffic is a smaller win than it looks.

Main example: a garage selling parts and accessories online

Take an illustrative car repair garage that runs a small online store beside the workshop, selling wiper blades, floor mats, bulbs and cleaning kits. About 900 sessions a month, a conversion rate around 1.4%, and 60% of traffic on phones. The owner opens the admin on a Monday and finds four Pulse cards.

Card one says to add a clearer "fits your car" message to the best-selling wiper blade page. Card two proposes a theme change tested by Sidekick, with an add-to-cart badge. Card three suggests writing a series of seasonal car-care guides. Card four recommends offering a discount code to first-time visitors.

He scores them. Card one: effort 1, revenue 4 (wipers are 38% of orders), ratio 4. Card two: effort 2 (review and publish a draft), revenue 2, because it is a desktop test and 60% of his traffic is on phones, ratio 1. Card three: effort 5, revenue 1, ratio 0.2. Card four: effort 1, revenue unclear, and it eats margin on a product that earns $9 after costs.

Now he verifies. For card one, he opens the product's page report and sees the wiper page gets 310 sessions a month, adds to cart at a lower rate than his other products, and has had returns tagged "wrong size". The claim holds. He acts: Sidekick opens a to-do list, he reviews the proposed text and edits it. For card two, he opens the draft theme in Canvas, and sees the change is cosmetic on desktop; he decides to look at it on his phone before publishing and leaves it for later. Cards three and four he dismisses, with short reasons: "too much effort for our traffic" and "discounts cut our margin on parts".

A month later he compares: wiper page add-to-cart rate, returns tagged wrong size, and orders. He writes the result in a one-line log next to the card, whichever way it went. After eight weeks the log shows two useful cards, one neutral and five dismissed, and he has a sense of what Pulse is good at for his store.

Using the AI to help you decide, not to decide for you

You can ask a general assistant such as ChatGPT or Claude to sanity-check your scoring, using numbers you supply. A prompt that works:

I run a small online store. Here are four recommendation cards from
my Shopify admin, my own effort and revenue scores for each, and
three figures from my reports (sessions, conversion rate, share of
mobile traffic). Challenge my scores: where might I be
overestimating revenue or underestimating effort? Ask me up to five
questions I should answer from my reports. Do not invent figures.
[paste cards, scores and figures]

An illustrative response: "Card two scores revenue 2, but you said 60% of sessions are on phones; does the test cover mobile? Card four: what is your margin after the discount? Have you looked at return reasons for the wiper blades?" Those questions are the ones worth answering, and they push you to the reports. What to fix in the output: if it offers a prediction such as "this will lift conversion by 15%", delete it. Nobody can know that from the cards.

Generic advice to skip, and how to spot it

Pulse works from store data and market trends, so some cards will be sound and some will read like content from any e-commerce guide. Signs a card is generic:

• It would apply to any store in your category without change. "Improve your product photos" fits everyone.
• It cites no store-specific figure, or only a trend.
• It asks for broad work, such as "grow your social presence", with no page or product named.
• It touches something you already decided on purpose, for example pricing you keep high to protect a brand.
• You cannot say what you would look at, a month on, to tell whether it worked.

Dismiss those with a reason. Writing "we already do this" or "does not fit our margins" takes seconds and, according to the help page, feeds the personalisation. If the same card returns three times, you may want to look at your own notes, as it may be pointing to a real gap you keep postponing.

When a trend card has no store evidence behind it

A card suggests adding a product category that is trending. The owner of a store selling cleaning supplies reads it on a Friday evening and spends the weekend adding 15 new products from a supplier, with descriptions written by an AI and photographs from the supplier's site. Six weeks later the new lines have sold twice, tied up $1,300 in stock, and diluted the collection pages that used to convert. The card was not wrong that the category was trending. It said nothing about whether this store's buyers would want it, and the owner never opened the report that would have said so, which was search terms on his own site.

The check he skipped would have taken ten minutes: look at on-site search terms for the category, and at how many existing customers bought anything related. Zero searches in 90 days is an answer. For anything that costs stock or a long set-up, ask for evidence from your own store before the first purchase order, and start with a small trial of three products.

Two more cards, walked through

These are further illustrations, so the scoring is not only shown on the easy cases.

A card about abandoned checkouts

Suppose a card says many shoppers leave at the shipping step and proposes a clearer delivery estimate. Effort is 1 or 2. Revenue depends on the evidence: open the checkout report and see where sessions drop. If the drop at the shipping step is large and you have had emails asking about delivery times, the claim stands and the ratio is high. If your drop-off is mostly at the payment step, the card has the wrong diagnosis and a clearer estimate will not help. The same card, two stores, two answers. That is why the report comes before the decision.

A card about stock

Another card warns that a best seller may run out. This one deserves a quick action, because the cost of being wrong is asymmetric: a stock-out loses sales and an unneeded check wastes five minutes. Check on-hand and incoming quantities yourself, look at the last 30 days of sales to estimate how long the stock lasts, and decide on the reorder. Purchase orders now live in the Shopify admin (Stocky can no longer be installed from the App Store), and the purchase order tutorial covers the reordering side. Do the arithmetic yourself: 46 units left and 9 sold a week is five weeks of cover, which is fine if your supplier takes three and tight if they take six.

How to tell whether a change worked

Decide before you start what you will measure and when you will look. For a product page change, the figures are sessions, add-to-cart rate and orders for that page, over four weeks against the four weeks before. Weekly numbers on a small store bounce about; a single good week is not proof. Two rules help. First, compare like for like: do not set a December month against a February one. Second, write down anything else you changed that month, such as a promotion or a new ad, because those muddy the result.

If a change appears to have helped, keep it. If the numbers are flat, that is a result too, and it should lower how much you trust the next card of the same kind. If something got worse, restore the old values, and tell Sidekick when you dismiss the next similar card.

What your staff should know

If more than one person uses the admin, agree who looks at Pulse and who may act on it, because acting on a card starts a conversation that can lead to published changes. Shopify says Sidekick works within staff permissions, so a person with narrower access sees and can do less. A simple house rule is enough: one named person reviews cards weekly, anyone can suggest a change, and nothing is published without a second look at the draft. Keep the log in a shared sheet so the reasoning does not sit in one person's head.

Twenty minutes each week with the Pulse cards

1. Open Home on the same day each week and note the cards. Two minutes.
2. Score each card for effort and revenue. Five minutes.
3. For the top one or two, find the underlying figure in your reports and compare with the previous month. Eight minutes.
4. Act on at most one thing a week. Start the conversation from the card, read the to-do list, and edit anything wrong.
5. Dismiss the rest with a reason.
6. Add a line to your log: date, card, score, action, and the figure to recheck in four weeks.

Limiting yourself to one change a week is what makes the result readable. Two changes to the same page at once and you cannot tell which did anything. If you also use Sidekick for other jobs, using Shopify Sidekick to run your store faster covers the day-to-day tasks, and Sidekick skills for repeat tasks shows how to save the check as a routine.

When a card involves a change you cannot easily undo

Shopify's help pages say store changes made through Sidekick are drafts until you review them and publish, and that published changes cannot be undone automatically: to reverse one, you ask Sidekick to change the values back and publish again. So a card that proposes pricing, policy or navigation changes deserves a gentler path. Take a note of the current values before you start, either in a spreadsheet or by screenshot. For theme work, the unpublished draft is a safety net: preview it on phone and desktop, and publish only when you would be comfortable leaving it live for a week.

The reliability of what Sidekick says and does is a separate topic, covered in whether Sidekick is reliable enough for store decisions. The short version for Pulse is the one at the top: a card is a lead, and your reports decide whether to follow it.

If you want help building that weekly check around your own store's numbers, and deciding what to automate, that is part of my AI implementation consultation, and you can get in touch here.

What to do with the first set of cards

Do not try to clear all five. Pick the highest ratio, find its figure in your reports, and run it for four weeks. Dismiss the rest with reasons. At the end of the month look at your log. If most cards are generic, that is useful to know, and the time Pulse costs you is the two minutes a week it takes to dismiss them. If two of five were genuinely good, you have a source of ideas that costs nothing but attention.

Questions about Pulse cards

Why can I not see Pulse in my admin?

Shopify's help page says only some stores with recent sales activity have access, you must be the owner or a staff member with admin access, and Shopify Network Intelligence must be activated for the store. A quiet store, a restricted staff account or a switched-off setting could each explain it.

Does Pulse change my live store by itself?

No. Shopify says Sidekick does not change your live store without your approval. Acting on a card opens a Sidekick conversation with a to-do list, and you review changes. Theme suggestions arrive as an unpublished draft that stays separate until you publish it.

What does Tested by Sidekick mean?

For some theme changes, Sidekick builds the change into an unpublished copy of your theme and runs simulated shopping sessions through SimGym. If the copy has a higher add-to-cart rate than the live theme, the card shows the label. It is a simulation, not a record of real shoppers.

How often do the cards refresh?

Shopify's help page does not say how often recommendations refresh or exactly which store data they use. If timing matters to you, note when a card appears, dismiss it, and see when its replacement arrives. Do not rely on a figure from a third-party post.

Does dismissing a card do anything?

Yes. You click the X and can say why the card was not relevant. Shopify says Sidekick uses that feedback to personalise future recommendations. Giving a short, honest reason is worth the ten seconds if you want fewer generic suggestions.

Further reads

• Best AI Assistant for Shopify: Sidekick, ChatGPT or Claude by Job — Pick the right assistant for each store job.
• How Much Does It Cost to Add AI to a Shopify Store? — What AI add-ons cost on a Shopify store.
• AI for Ecommerce Analytics: Shopify Reports, GA4 or ChatGPT First? — Where to look first when you check a recommendation.
• Shopify Magic vs ChatGPT for Product Descriptions — Compare the built-in writer with ChatGPT for product copy.
• AI Product Recommendations on Shopify: Built-In vs Paid Apps — Compare on-site product recommendation options.
• Ecommerce Product Recommendation AI: Quiz, Widget or Chat Assistant? — A quiz, an on-page widget and a chat assistant costed for a small store, with a rule by catalogue size and a method to measure add-to-cart lift.
• Does Shopify Have AI? What Is Built In and What Costs Extra — Which Shopify AI features come with Basic, Grow, Advanced and Plus, at what price, and where the extra costs begin.
• Shopify Agentic Storefronts Explained: What You Still Control — A control-by-control walk through Agentic Storefronts: what the admin lets you manage, what stays on by default, and what the data decides.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Shopify's Sidekick Pulse help page; Shopify changelog posts of 10 December 2025 and 28 October 2026 (web search context, supported app data, AI-tested storefront recommendations); Shopify Sidekick help pages on store changes and limits; Shopify's purchase orders and Stocky help pages. All read 5 November 2026.