Skip to main content
MixShift Intelligence

Why it moved,
not just what it did

Performance analysis for brands and agencies selling on Amazon

MixShift Intelligence turns your retail, advertising, and operational data into one connected explanation of performance. It quantifies the drivers behind every change, traces them to the product lines, products, campaigns, and targets responsible, and shows your team where to focus next.

Included with every MixShift subscription. Talk to us.

  • Seller Central and Vendor Central
  • Retail, advertising, and inventory
  • Account to product line to product
  • Campaign type to campaign to target
  • Month over month, year over year, custom
  • Workspace, reports, AI, and API
A worked result

A $51K gain, explained and reconciled

Scope

Account
Vendor Central, Manufacturer view, US
Periods
August 2026 vs July 2026
Metric
Ordered revenue
Measured change
+$51,000
$1.28M to $1.33M, +4.0%

What drove the change

DriverWhat movedImpact
TrafficMore people saw the listings. Views up 2.8%.+$35,400
ConversionMore of them bought. Conversion up 0.2 points.+$23,900
PriceThey paid slightly less on average. Price down 0.6%.-$8,300
Adds up to+$51,000

The three add up to the total change exactly, with nothing left over. Every result is checked this way before you see it.

Where it happened

  • Care and cleaning line
    Discounted 11% from August 1 to August 31, adding about 1,950 units. Last month ran a deeper discount, so the average selling price still landed 8% above last month.
    +$48,600
  • Legacy bottle line
    No advertising behind it and inventory drawn down: about 1,000 fewer units. Your team’s own notes have this line winding down.
    -$22,100
  • Sport bottle line
    In stock all month after 18 days out last month. About 1,250 more units, and an estimated $1,400 less lost to stockouts.
    +$20,700
  • Storage line
    Discounted 20% from August 13 to August 31, adding about 600 units across the back half of the month.
    +$18,800
  • Flask line
    Discounted 18%, but only from August 1 to August 4. Units down about 570, with an estimated $900 more lost to stockouts.
    -$16,100

The same +$51,000, now by product line. These five are the biggest movers and account for +$49,900 of it; every other line nets out to about +$1,100.

What else the data shows

  • Six product lines were on promotion, and 87% of the change sits in those six.
  • Stock fell 12%, from 6.6 weeks of cover down to 5.9.
  • Less was lost to running out of stock: an estimated $5,500 last month, $4,200 this month.
  • Ad spend fell 6% and ad sales fell 2%, so advertising got more efficient.

What explains the gain: promotional pricing on the care and cleaning line plus recovered availability, with traffic up on lower ad spend.

What to watch: whether the care and cleaning line holds its gain once the promotion ends.

What to check next: that line’s traffic and conversion in the first week at full price, and the flask line’s stockouts.

Every conclusion links back to the products, metrics, and periods behind it. Illustrative example based on an anonymized, rounded account result.

Attribution

Five connected analyses. One view of performance.

Operational Bridge

Which operational changes drove performance?

Breaks revenue, units, traffic, conversion, price, Buy Box, inventory, and availability into measurable drivers, then traces those drivers to the product lines and products behind them.

Advertising Bridge

What changed in advertising, and where?

Explains changes in spend, ad sales, traffic, cost, conversion, and efficiency across campaign types, campaigns, targets, and products. It also shows what an ad actually sold, separating sales of the advertised product from other products bought after the ad and from view-through sales, so you can see where a campaign really creates value.

Combined Read

How did retail and advertising move together?

Brings retail and advertising into one investigation, with TACOS, attributed share, paid pressure, and ad AOV against selling price, so a shift in spend, traffic, sales, or efficiency is not split across separate dashboards.

Lost Sales

What did running out of stock cost us?

Estimates the revenue missed during stockouts, identifies the products and periods affected, and connects the result back to inventory and demand.

Monthly Read

What changed this month, and against last year?

Compares month over month and year over year across retail and advertising, ranks what mattered most, and becomes the client-ready monthly report with the deeper analysis attached.

How deep it goes

From the account result to the products and levers behind it

Start with revenue, or any supported operational or advertising metric, and follow the change through its measurable drivers. Intelligence separates traffic, conversion, price, product mix, advertising, and availability effects, then moves through the account to show where each effect happened.

Compare this month with last month, the same period last year, or any two ranges the data covers. Investigate the account, a product line, a sub-brand, or a single product without starting the analysis over at every level.

Operational

Revenue, units, traffic, conversion, average selling price, Buy Box, sellable inventory, weeks of cover, and estimated lost sales.

Advertising

Spend, ad sales, impressions, clicks, CTR, CPC, orders, CPA, ad AOV, conversion, ACOS, ROAS, and the split between the advertised product, other products, and view-through.

Customer outcome

From reporting call to working session

A monthly report flagged one product as a meaningful negative driver. In the Workspace the team ruled out stockouts, saw inventory still building, and found traffic, ad spend, and clicks had all fallen by roughly half. Their own business context confirmed a known listing issue.

Instead of spending the call assembling numbers, the team left with the issue identified, its impact quantified, and the next action clear.

Where it shows up

Use Intelligence where your team already works

Intelligence Workspace

Explore a finding from the account total down to the products, metrics, and supporting data behind it, and test it against the brand context your team keeps.

Monthly Performance Report Max

A client-ready monthly brief and a deeper internal analysis behind it, both built from the same calculated findings. See Monthly Report Max.

MixShift AI plugin

Bring findings into Claude and other supported AI tools without asking the model to recreate the analysis from raw data. See the plugin.

Builder Platform

Use the same outputs in your own applications, agents, and workflows over MCP, REST, and the command line.

Trust

Built to show its work

Intelligence calculates the change, reconciles the drivers, and keeps the supporting data attached to the conclusion.

Numbers that reconcile

The driver contributions add back to the measured change, and rates and ratios are rebuilt from the numbers underneath them.

Evidence at every level

Move from the account result to the products, campaigns, metrics, and periods behind it.

Consistent everywhere

The same calculated result powers the Workspace, monthly reports, AI tools, and the Builder Platform.

The engine

The attribution engine inside MixShift Intelligence

At the core is HCAM, MixShift’s patent-pending Hierarchical Causal Attribution Model. HCAM breaks a performance change into measurable drivers and traces those contributions through the account, product line, product, and advertising hierarchies.

MixShift Intelligence is the layer around it, connecting that attribution with your brand context, operational and advertising evidence, inventory and promotion signals, forecasting, monthly reporting, the Workspace, AI delivery, and Builder Platform access. HCAM does the math. Intelligence turns it into something your team can act on.

What feeds a result

  • HCAM attribution

    Runs today

    Business data becomes reconciled evidence

  • Context and mechanisms

    Runs today

    Goals, events, and the brand context your team keeps

  • Expectations and actions

    Coming next

    Forecast, intervention, and counterfactual models

Governed AI execution

  1. Intelligence Service

    Identity, scope, retrieval

  2. Governed skill

    Claims, workflow, refusal

  3. AI synthesis

    Interpretation, not re-math

  4. Validated support

    Checks and an accountable human

The Intelligence layer. Attribution and brand context run today. Forecasting is coming next, and the intervention and counterfactual inputs use separately supported models.
Coming next

Forecast against actual

Set an expectation from the brand’s own history, then use Intelligence to measure the beat or miss and identify the products and drivers behind the variance.

Licensing

Bring the method to your own data

The architecture exists so a business can take the efficiency AI offers on analysis it is still able to audit, and that problem is not specific to Amazon. If you want to adopt the technology in your own business, we will consider licensing it.

Built for Seller Central and Vendor Central

The two report different things, so Intelligence applies the right definitions for each: traffic, conversion, advertising attribution, and inventory are each measured the way that account actually reports them. Vendor Central runs on the Manufacturer view.

FAQ

Common questions

An AI summary turns whatever it is given into sentences. Intelligence does the work first: it calculates what changed, checks that the parts add up, finds the products behind it, and gathers the supporting data. Only then does AI explain the finished result.

See why it moved

Start with an explanation your team can follow, and act on.