Docs
Audiences
An audience is a list of customers, saved from a finding in a report. Affinsy builds the list, keeps it, and counts from your store’s own orders who on it goes on to buy. This page walks through the whole path: the rules behind a list, the settings, saving, getting the list to your email tool, and reading the result.
What an audience is
A market basket report ends in findings such as “customers who buy the grinder often buy the filters too”. An audience turns one finding into the customers it points at: here, everyone who bought the grinder and has not bought the filters yet.
- It is a list, not an automation. Affinsy does not send emails, does not tag customers and changes nothing in your store. You take the list to your email tool.
- It is fixed once saved. The customers on it never change. For a fresh list, save a new audience from a newer analysis.
- It is watched from the day you save it. Affinsy counts who on the list buys what was suggested to them.
Where to start one
Every audience starts from a completed market basket analysis, in one of these places:
- From a rule
- Open a rule in the report’s rules table and choose Build audience in the footer of its panel. The audience is the customers that one rule points at.
- From a product
- In the products table, use the audience icon on a product’s row, or open the product and choose Build audience. A dialog asks What should this audience be? with up to four answers (below).
- From the Audiences page
- Create audience offers From an analysis, which links to your latest reports, and Your own rule, for a pairing no analysis has found yet, such as a product you just launched.
- From the chat
- Ask the assistant to build an audience from a rule or a product. It counts the customers and can save the list to the Audiences page.
Starting from a product, the four answers are:
- Likely to buy the product
- Customers who bought what usually goes with it and have not bought it yet. Everyone on the list gets this one product as their suggestion.
- Cross-sell to its buyers
- People who bought the product, offered the partner products you tick. Each customer gets only the ones they have not bought yet.
- Reorder reminders
- People who bought it and are due to buy it again, on its usual cycle.
- Win back its buyers
- People who bought it and are past their usual time to buy it again.
An answer is greyed out when the report has nothing to build it from, for example a product with no reorder cycle.
The rules
A rule is one finding of the market basket analysis: the products on the left, an arrow, the product on the right. It reads “customers who bought the left side often buy the right side too”. Rules are found in your order history; nobody writes them, and they do nothing by themselves.
In the audience dialog the rules decide who is on the list and what each customer is offered:
- Who: customers who bought the left side of a ticked rule and have not bought the right side.
- What:the right side is that customer’s suggestion.
Under Likely to buy and Cross-sell the rules are listed with checkboxes. The columns are Rule, Watch, Strength, Lift, Also buy it and To reach. Untick the rules you do not want. A rule with a watch-out (for example a pairing that is fading) starts unticked.
Good to know
One audience has one centre. Either every rule leads to the same product, or every rule starts from the same product. A mix is refused with a short message, so each list has one clear offer.
Your own rule works the same way, except that you choose the products yourself: Customers who bought any of and Offer them. It is marked Not checked, because no analysis stands behind it.
Settings and preview
The second step, Settings & preview, has two switches:
- Apply substitute guard
- On by default. Leaves out customers who already bought a substitute of the suggested product: they picked the alternative, so the offer would likely miss. Customize lets you untick single substitutes.
- Include out-of-stock products
- Off by default, so nobody is offered something they cannot buy.
Generate audience then builds the list and shows how many customers are on it, what they are offered, and the first rows of the list. Nothing is saved yet. From here:
- Download CSV downloads the list as a file and saves nothing.
- Save audience goes on to the last step.
A reorder audience has one setting instead: Target reorder windows, with Before window, In window, Past window and Lapsed. Before the window is the heads-up list, in the window is the send-now list, past the window is the win-back list.
Saving and the holdout
The last step shows the list’s size and one choice, Holdout: none, 10%, 20%, 30% or 50%, with a recommendation for the list’s size.
A holdout is a part of the same list that you deliberately leave out of your campaign. Comparing the two groups afterwards shows what the campaign changed, instead of what would have happened anyway.
- With a holdout the list is split into Tagged (the customers to contact) and Holdout (the customers to leave alone), and a result is named when the window ends.
- Without one the list is watched for the same time, with no verdict.
- Under 200 customers a split is too small to read, so none is recommended.
Good to know
The holdout cannot change once the audience is saved, and neither can the list. Saving the same definition twice opens the audience you already have.
Reaching the customers
Affinsy sends nothing. The saved audience’s page has two buttons for taking the list out:
- Download
- The list as a CSV. With a holdout you can download the tagged customers, the held-out customers, or everyone with a group column. A store synced through the Shopify app also gets Shopify tags: Customer ID and Tags, to merge into your own customer export.
- Flow suggestion
- Step-by-step instructions for your email tool: Klaviyo, Mailchimp, MailerLite, Omnisend or Shopify Email. They cover making the list or segment, writing the email, an optional follow-up a week later, and keeping the held-out customers out of it. Copy steps copies them.
Tagged is the name of a group, not something Affinsy does. A customer only carries a tag in Shopify once you import it yourself, as the flow suggestion describes.
The Live switch in that dialog is for your own records: flip it once your flow is running. The result does not depend on it.
Reading the result
The result is counted from your store’s own orders, from the day the audience was saved. Nothing comes from your email tool.
- What counts
- A customer on the list who orders one of the products suggested to them, inside the window. For a reorder audience: who orders the product they were due for.
- The window
- 30 days for an audience built from rules. For reorder audiences, the product's own cycle, between 7 and 120 days. Orders after the window do not count. When it ends you can count on for 7, 14 or 30 more days with the same customers, up to 120 days in all.
- Without a holdout
- The status is Watching, then Ended. The page shows how many on the list bought and what they spent, with no verdict.
- With a holdout
- The status is Measuring until the window ends. Then the share of tagged customers who bought is compared with the share of held-out customers who bought.
The verdict is one of four words:
- Significant
- 90% confidence or more that the tagged group bought more often than the holdout.
- Promising
- 75 to 89% confidence.
- Insignificant
- Below 75%.
- Inconclusive
- A group had fewer than 50 customers, or fewer than 7 days had passed.
The chart, Bought a suggestion, by day, adds up the share of each group who bought, day by day. For audiences built from rules it is drawn over three bands, Above expected, As expected and Below expected. The bands come from the same list as it would have been on six earlier dates, when no campaign was running. The first seven days are shaded: it is too early to read anything.
Below the chart, two tables show what the result is made of: the rules behind the list, and the orders it is counted from.
What it does not prove
Confidence is how sure the comparison is, not the share of customers who bought. Without a holdout there is no comparison, so the figures say what happened on the list, not what your campaign caused.
The Audiences page
Audiences in the menu lists every saved audience with its size, the day it was saved, the report it came from, its progress through the window, and its result. The figures at the top add up the audiences that have a holdout: Extra buyers and Extra revenue are what the tagged groups did above the held-out rate.
- Filter by All, Rules or Reorder, or search by name.
- Hover a row to pin it to the top, archive it, delete it or open it.
- A win-back list is saved as a reorder audience, with the past window only.
Good to know
- Customer IDs are needed. An audience is a list of customers, so the orders must carry a customer ID. Orders synced from Shopify carry it wherever Shopify has a customer on the order.
- A list holds up to 50,000 customers.
- An empty list cannot be saved.
- Building and saving need an active plan, the trial included.
- The segments report does not build audiences. It has Export customers, a CSV of a segment for your email tool, which is not saved or measured.
- Open a market basket report and pick a rule or a product.
- Choose Build audience, tick the rules, generate the list.
- Save it, with a holdout when the list is large enough.
- Take the list to your email tool with Download and Flow suggestion.
- Come back when the window ends and read the result.