↑↓ selectEnter openEsc close

    1. Elbuz
    2. AI in CRM

    Customer profile: segment, rating, purchases and pain points

    4 min readUpdated

    Elbuz builds each customer's profile from their purchases: segment, a 1–5 rating, average order value. AI adds interests and pain points from conversations. No need to track by hand who buys often and who has gone quiet.

    When you need this:

    • Find your most valuable customers: the “VIP customers” filter and rating stars in the “Contacts” grid.
    • Customers stopped buying: they turn “Dormant” after your business format's threshold.
    • A manager forgot what a customer complained about: the customer panel shows “Pain points and objections”.
    • Cross-sell after a purchase: AI picks related products using the profile.

    Where to start: “Contacts” → “Filters” → “VIP customers”.

    What the customer profile does

    • Segment: “New”, “Repeat”, “Regular”, “VIP”, “Dormant”, “Lost”, based on purchase documents, no AI involved.
    • 1–5 rating: recency, number and total of purchases in a single figure.
    • Interests and pain points: AI collects them from the customer's emails, chats and events.
    • Filters and columns: VIP, regular and dormant customers in the “Contacts” grid.
    • The profile in AI's work: replies to customers, cross-selling, finding customers by plain-language request.

    How the segment is determined

    Elbuz checks the rules in order; the first one that matches wins:

    1. Lost: the last purchase was more than a year ago (365 days);
    2. Dormant: the last purchase is older than the “dormant” threshold of your business format;
    3. VIP: at least as many purchases as the “regular” threshold, and a purchase total at least twice the average customer total in your store;
    4. Regular: at least as many purchases as the “regular” threshold;
    5. Repeat: two or more purchases;
    6. New: one purchase.

    With no purchases, there is no segment. Thresholds depend on the “Business format” in your company details:

    Format“Dormant” after“Regular” fromWhat counts as a purchase
    Online store90 days3 purchasescustomer order
    Store with a pickup point90 days3 purchasescustomer order, sales receipt
    Wholesale ()120 days3 purchasescustomer order, sales invoice
    Servicestwo average intervals between payments, but no less than 45 days2 purchasespaid invoice, signed service act

    If no format is set, the online store rules apply. Canceled orders and drafts don't count as purchases; returns are counted separately.

    Filter labels and segment hints in the app are written for an online store: 90 days and three orders. For wholesale and services, the segment is calculated with the thresholds from the table above, even if the label says otherwise.

    How the 1–5 customer rating is calculated

    The rating is the average of three scores, rounded to a whole number. All three carry equal weight.

    Score54321
    Days since last purchaseup to 30 daysup to 90up to 180up to 365longer
    Number of purchases10+5–93–421
    Purchase total compared with the store average5 times or more2 timesabout the samehalfless than half

    So the rating goes up with every repeat purchase and drops when a customer hasn't bought for a long time. The average total is recalculated across all your customers on every recalculation.

    When the profile is updated

    A background job recalculates the purchase part once an hour. It picks up:

    • customers whose purchase document, order, return or deal changed in the last two days;
    • customers who have never been calculated;
    • everyone else once a week.

    As a result, a customer who stops buying becomes “Dormant” with a delay of up to 7 days. Emails, chats and calls don't update the purchase part, only documents do.

    Together with the segment, the customer stage is updated if the matching stage is enabled in the directory (suppliers and competitors have no stage): “New” and “Repeat” set the “Customer” stage, “Regular” and “VIP” set “Regular”, “Dormant” and “Lost” set the stages of the same name. More about stages: Contact card.

    What AI writes about interests and pain points

    AI builds the second part of the profile together with the customer's “AI summary”. It receives the card, purchases, a few recent documents and deals, the last three emails without quoted text, feed events, chat messages and the scheduled follow-up, and returns:

    • a short summary: who the customer is, what they bought, what they wrote about, what to do next;
    • interests: categories and products they're interested in;
    • pain points and objections: complaints, dissatisfaction and doubts from emails, chats and calls;
    • customer type (retail, wholesale, supplier, partner, not a customer) and preferences: channel, delivery, payment. AI uses these two items in replies; they aren't shown on screen.

    Each list holds up to 6 items. AI is explicitly instructed not to make anything up: if a customer never complained, there are no pain points.

    The summary is built when you open the customer panel in “Contacts” or the customer side panel in a chat, and only if the customer has had new activity since the previous summary: an email, chat, document or event. Nothing new: the saved summary is shown and nothing is charged. The “Refresh summary” button rebuilds it on demand. One call in a test Elbuz account cost about $0.003 (measured from the charge log, October 13–14, 2026).

    Where to see segment, rating and pain points

    Where: Directories → Contacts → Contacts.

    • Grid: the “Customer segment” column; “Customer rating” is available in column settings.
    • Filters: “Filters” → “VIP customers”, “Regular customers”, “Dormant customers”; “New customers”, “Repeat customers”, “Lost customers” are in filter settings.
    • Customer panel: click a row to see the “AI summary” block and the figures below it.
    Contacts grid with the VIP customers filter and the Customer segment column
    1VIP customers filter
    2Customer segment column
    1. VIP customers — the filter under More that leaves only customers in the VIP segment.
    2. Customer segment — the column that shows each customer's segment.

    The column shows a segment badge and stars; hover over the stars to see “Customer rating (recency, frequency and spend): N/5”. The columns are read-only: you can't set a segment or rating by hand.

    Preview panel of a customer with the AI summary and order statistics
    1AI summary
    2Orders and recent purchases
    1. AI summary — a short text built automatically from the customer's orders and contacts.
    2. Orders and recent purchases — order count, average order, last order and interests.

    Below the summary, the panel shows “Orders” (count · total), “Average order”, “Last order”, “Returns” (if any), “Recent purchases” (up to 5 “date — product” lines), “Interests” and “Pain points and objections”. The customer card header has a single stage badge with “ · VIP” appended and stars; the hint on the stage explains how it was calculated.

    How AI uses the profile

    In replies to emails and chats. Before replying to a customer, AI receives their profile: stage, number and total of orders, segment and rating, recent purchases and categories, type, interests, pain points, preferences and the summary. So a reply can mention “your order from August 13”, and a draft for a customer who has already complained can take the complaint into account. How replies work: Replies to customers.

    In cross-selling. The “Cross-sell after purchase” flow runs daily at 11:00 and picks up to 10 customers who meet all of these conditions:

    • rating of 1 or higher, and an order completed 1–3 days ago;
    • no returns and no complaint emails in 90 days: a complaint is an email that AI classified as a complaint or negative; a complaint in a chat or call doesn't stop cross-selling;
    • AI hasn't suggested an email to them for at least 14 days.

    For products bought in the last 30 days, AI picks related items: first the “Related products” from the product card (how to fill them in), then items frequently bought together, then the same category, only in stock and at the selling price. The email waits for approval in “Pulse”. Cross-selling looks only at customer orders, so invoices and service acts in the “Services” format don't trigger it. Flow settings: Automations (CRM flows).

    In search and flows. The AI consultant and builder flows filter customers by segment, rating and stage; the consultant can also filter by customer type, interests, pain points and preferences from the profile. These fields are visible to anyone who can see “Customer segment”. The profile has no purchase prices: totals are calculated from the selling amounts in documents, and recent purchases are stored without prices.

    Fill in “Related products” in the cards of popular products: cross-selling uses them first, so the picks are more accurate than “from the same category”.

    What to ask the AI consultant

    • “Show VIP customers with no follow-up for more than 30 days”;
    • “How many dormant customers do we have, and which of them have the most orders?”: the consultant counts orders; purchase totals aren't available in its search;
    • “Show regular customers with no scheduled follow-up”;
    • “What are customer Petro Myroshnyk's pain points and interests?”: the consultant answers from the customer profile;
    • “Find customers interested in charging stations”;
    • “Which regular customers complained about delivery?”;
    • in the customer card, the “✦ AI” button → “Summarize the customer”: who they are, purchases, open deals and orders, recent tickets;
    • “Ask Sales to suggest follow-ups for dormant customers”: an assignment for an AI employee; details: Delegating to specialists from chat.

    More phrases and the cost of a reply: AI consultant: what to ask.

    Scenarios

    Find the customers who bring in the money

    Before: sorting by turnover and counting by hand who buys often.

    After: the “VIP customers” filter: customers with three or more orders and a total of at least twice the store average; the stars show how recent the last purchase is.

    A customer who has already complained

    Before: the manager doesn't remember that the customer wrote about headphones that pinch.

    After: the panel shows “Pain points and objections: The headphones pinch and didn't fit”; AI sees the same when preparing the next reply, and after a complaint email cross-selling won't prepare emails for this customer for 90 days.

    Cross-selling after a purchase

    Before: after buying a smartphone, the customer gets nothing else.

    What to do: fill in “Related products” for the smartphone and keep the “Cross-sell after purchase” flow enabled.

    After: one to three days after the order is completed, “Pulse” has an email with in-stock cases and chargers, sent once you approve it.

    FAQ

    Why doesn't a customer have a segment?

    They have no purchases yet, or the hourly recalculation hasn't reached them. Segments are assigned only to buyers.

    Why isn't a customer with high turnover VIP?

    VIP requires both a number of purchases (three or more, two or more for services) and a total at least twice the average. A single large purchase gives “New” with a high score for the total.

    Where did AI get the customer's pain points?

    From the customer's emails, chats and events when it built the AI summary. If the customer never complained, the item is empty.

    Why did wholesale customers who buy once a quarter become “dormant”?

    Check the “Business format” in your company details: the threshold is 120 days for wholesale and 90 for an online store. If no format is set, 90 applies.

    Can I change a segment by hand?

    No. Elbuz calculates the segment and rating from purchases; the columns are read-only.

    Where it is and who has access

    Menu
    Directories → Contacts → Contacts
    Access right
    access to the “Contacts” window; customers are visible according to the “Data scope” rule
    Prepare
    customer orders in Elbuz: with no purchases there is no segment or rating

    Related topics

    Was this article helpful?

    Missing a feature? Tell us.

    We will add it for free if other users need it too.

    If only you need it, we will agree on it separately.

    Article contents

    Suggest an improvement

    Try Elbuz free