Lifetimely: LTV Drivers
LTV Drivers
At a glance
LTV Drivers identifies first-order features associated with higher or lower customer value and lets you compare historical value across first-order groups. Use it to generate acquisition, merchandising, and retention hypotheses; an association in this report does not prove that the feature caused the result.
Availability and data freshness
The insights views show a no-data message when there is not enough historical data to identify meaningful drivers.
Freshness and history: Lifetimely does not currently publish the source delay, model refresh schedule, model training cutoff, or earliest available history. LTV Driver Breakdown falls back to the six months before today through yesterday in the store timezone; a date range saved from this or another report can override that default. The two insights sections do not expose a date range.
On this page
- Choose an insights or breakdown view
- Use LTV Driver Insights
- Use LTV Driver Breakdown
- Understand the data contract
- Interpret the results
- Reconcile differences and missing data
Open the report
Open Lifetime value > LTV Drivers (/lifetime_value/ltv_drivers). LTV Driver Insights is selected initially.
Choose a view
- LTV Driver Insights contains a positive/negative LTV-driver table and a separate Return customer insights table. These are precomputed, store-level results: the current page sends only pagination to these APIs and offers no date, dimension, or report-filter controls for them.
- LTV Driver Breakdown compares customer count, historical LTV, and average order count by a selected first-order dimension. Its date, dimension, and filters do not change either insights table.
The page does not provide a comparison-period control or a way to apply the Breakdown filters to the insights models.
LTV Driver Insights
Positive and negative LTV drivers
Use Display to switch between Positive LTV drivers and Negative LTV drivers. Each view initially requests up to 10 rows. The table can display Product, Order tag, Customer tag, Discount code, Marketing source, Marketing medium, Marketing campaign, First touch marketing channel, and Last touch marketing channel features.
| Column | Definition |
|---|---|
| Name | Feature type and feature value associated with customers' first orders. |
| Delta | A visual bar based on the absolute Lift values in the displayed positive or negative list. It is normalized within that list, has no independent unit, and should not be compared as a shared scale between the positive and negative views. |
| Lift | Percentage difference from overall average customer lifetime value, as defined by the current tooltip. |
| Historical LTV | Average total sales per customer in the last 24 months, including taxes and after discount codes and refunds are deducted. The currency shown is the active store-context currency. |
Export downloads the selected positive or negative list and requests up to 1,000 rows. The CSV contains Name, Lift, and Historical LTV; it does not include the Delta visualization, hidden model ranges, or the opposite driver type.
Return customer insights
This table states whether customers with a feature in their first order are shown as more or less likely to return than average customers. It shows five rows per page and supports pagination.
| Display | Definition |
|---|---|
| Feature | Product, order tag, customer tag, discount code, or marketing feature identified by the insights result. |
| Retention ratio | The API-provided odds value, displayed as x times more or less likely according to the sign of a separate model coefficient. Do not read 2x as a two-times increase in repurchase probability until those semantics are confirmed. |
The section's Export requests up to 1,000 rows and writes Type, Name, and Retention Ratio. It does not include the coefficient used to choose “more” or “less.”
LTV Driver Breakdown
Use the controls
The fallback state is Product, Last 6 months through yesterday, no complex filters, Show color scale on, 10 rows per page, and Customers descending. A stored report range can replace the fallback dates.
- Breakdown by: Choose Product, SKU, Product Type, Country, State, Customer Tag, First Order Tag, or First Order Discount Code. Changing the dimension changes row membership, not the metrics.
- First order date: Choose custom dates or Last 6 months, Last 12 months, This year, or a Since 2020 through Since 2025 preset. Boundaries use the store timezone and end at yesterday for presets.
- Show color scale: Adds relative color shading to Historical LTV and No. of Orders. The current UI calculates each scale from only the rows on the loaded page, so a color can change after sorting or pagination. This changes presentation only.
- Filters: Open Filters to narrow the qualifying population. Available groups include marketing attribution, tags, products, channels, subscriptions, first-order value/discount, geography, discount code, and surveys; Shopify-only groups are hidden for Amazon contexts. Saved cohorts can also populate these filters.
- Sorting and pagination: Click Name, Customers, Historical LTV, or No. of Orders to sort; clicking the active column reverses direction, while selecting another column starts descending. Change rows per page or navigate between pages.
- Export: Downloads Name, Customers, Historical LTV, and No. of Orders for the active dimension, dates, filters, sorting, and page position, requesting up to 1,000 rows. If more than 1,000 rows qualify—or if exporting after moving beyond page 1—verify the CSV scope before treating it as a complete result.
The Breakdown request automatically carries store-wide prefilters, shop calculation settings, and Shopify/Amazon context. The store timezone sets date boundaries, and the active store-context currency formats monetary results. Change prefilters and calculation choices outside this page; see shop calculation profiles.
Ask AI is available from the page header. Its page context includes the Breakdown first-order dates and selected dimension, and the loaded insights and Breakdown responses are made available as report data. Confirm the active tab and controls before relying on its summary.
Data contract: what goes into each view
The insights views and Breakdown are separate calculations and should not be expected to rank or total identically.
| View | Request and population behavior verified from the current implementation |
|---|---|
| Positive/negative drivers | Precomputed results for the active shop. The request contains pagination only; it does not carry the visible Breakdown dates, filters, prefilters, shop calculation settings, or Amazon switch. |
| Return customer insights | Precomputed results for the active shop, also requested with pagination only. The model returns a feature, coefficient, and odds value; the UI uses coefficient sign for direction and displays odds as the ratio. |
| LTV Driver Breakdown | Customers selected by first-order date and grouped by the selected dimension after the request's store prefilters and optional complex filters. The request also carries shop calculation settings and store-platform context. |
Breakdown metrics
| Column | Definition |
|---|---|
| Name | The selected dimension value. |
| Customers | Number of new customers assigned to the row after the selected first-order dates, prefilters, and complex filters are applied. A customer may potentially appear in multiple product or tag rows; do not sum rows until assignment rules are confirmed. |
| Historical LTV | Average total sales per customer for the row, including taxes and net of discount codes and refunds, displayed in the active store-context currency. |
| No. of Orders | Average number of orders per customer in the row. |
Check a Breakdown value
- Confirm the active shop, First order date, Breakdown by, and numbered Filters badge.
- Verify store-wide prefilters, timezone, and active currency outside the report.
- Sort by Customers and inspect whether a product, order, or customer can carry more than one selected dimension value before adding rows together.
- Export from page 1 and compare the row's customer count and sales with an equal-date, equal-filter cohort in Cohort Breakdown. These reports can still differ because Cohort Breakdown groups value by cohort age while this table returns one historical aggregate per dimension.
- For retention, compare the same first-order attribute in Repurchase Rate Breakdown, accounting for its selected repurchase window and potentially overlapping rows.
Interpretation example
Suppose Positive LTV drivers shows Product: Starter Kit, Lift of +15%, and Historical LTV of $230. This means the report associates that first-order feature with customer value 15% above its overall-average baseline and reports $230 average sales per customer over the stated 24-month historical window. It does not prove that placing the Starter Kit in more first orders would increase an individual customer's LTV by 15%.
To investigate, select Product in Breakdown, choose a mature first-order range, and inspect the Starter Kit's customer count, Historical LTV, and No. of Orders. Then compare it with other products at equal cohort ages in Cohort Breakdown and compare repurchasing in Repurchase Rate Breakdown. Control for acquisition source, discount, geography, and cohort age; run a controlled experiment before making a causal claim.
Resolve differences and missing data
- Insights do not change with Breakdown controls: This is expected in the current UI. The model endpoints receive only shop identity and pagination; Breakdown dates, filters, and settings are sent only to the Breakdown endpoint.
- Breakdown does not match Cohort Breakdown: LTV Driver Breakdown returns one historical aggregate per dimension row. Cohort Breakdown organizes accumulated metrics by cohort age and offers different dimensions and metric choices. Match first-order dates, prefilters, complex filters, currency, and dimension semantics before comparing.
- Rows do not add to the store total: Products, tags, and possibly other multi-value dimensions can overlap.
- Colors change between pages: Color bounds are recalculated from the currently loaded page. Compare the numbers, not color intensity, across pages or sort orders.
- No insights data: The UI identifies insufficient historical data as one cause, but eligibility thresholds, model refresh timing, and other no-output states are not published. A Breakdown result does not guarantee model output.
Next actions
- Compare driver groups at equal cohort ages.
- Check whether the same first-order attribute predicts repurchasing.
- Review store-wide prefilters before reconciling the Breakdown.