Finch Labs
Galton · from Finch Labs

The answers you need
to scale.

Honest measurement for Shopify operators spending $20k a month or more on paid.

Request a demo →

Connects to Shopify, Meta, Google Ads and GA4, then grades every platform's claim against your sales. Galton never writes to your accounts.

The stakes

What being wrong costs.

Every operator misallocates some share of paid budget. The question is how big the share is, and whether anything you own can show it to you. Platform-reported ROAS can't: it is the same system marking its own work.

The annual bill at 10% wrong

$20K / mo on paid

$24K a year

misallocated at 10% wrong

$50K / mo on paid

$60K a year

misallocated at 10% wrong

$100K / mo on paid

$120K a year

misallocated at 10% wrong

In one account, a shop owner doubted whether Meta was driving any incremental revenue at all. The read was that Meta showed no measurable effect on total revenue. But upon closer inspection, Galton noticed that new-customer count moved with Meta spend, day for day. Galton suggested a geo holdout to settle the question, and designed the test. The hypothesis was that even though Meta wasn't driving a large portion of revenue, it was indeed a crucial channel for new-customer acquisition: customers who would then repurchase through other channels, making Meta seem irrelevant when it was in fact the engine driving new-customer growth.

Turning Meta off on the revenue read alone is the decision a geo holdout is there to check. Your gap is your own, and finding it is the job.

The product

Ask what you'd ask a data scientist.

If the answer is in your data, Galton will surface it. If it isn't, it will design the experiment that settles it.

What moved overnight, and what's my highest-priority action? Home → How much does each channel over-claim? Channel Intelligence → Which channel earns the next dollar? Channel Intelligence → How are my paid channels affecting my unpaid channels? Channel Intelligence → When does a new customer pay us back? Customer Health → Is growth coming from new or repeat buyers? Performance → Do our free sample packs bring in long-term paying customers, and how do they differ from the rest? Customer Health → Which products deserve the budget? Product Insights → Which products are trending up, and which are trending down? Product Insights →

Every question routes to one of five modules. Below, what each one shows you.

Inside Galton · 01 · Home

Know what to look at first.

Better answers, better decisions, in the order they matter. When a number moves, Galton brings it to your attention immediately. When nothing moved, it stays quiet.

Needs your attention

Which channel actually creates demand?

Meta claims 2.1× the revenue GA4 grants it, and branded search co-moves with Meta spend. A holdout would settle it.

Recommendation waiting

What can you afford to pay for a customer?

Max CAC on the candle line fell to $22 as rebuys softened to 19%. Blended break-even sits at $27.

Changed overnight
New or moved this week
When does a cohort pay you back? Payback shortened to 81 days, from 94, as the post-purchase upsell took hold.
Where is the next dollar still working? Meta prospecting campaigns show signs they can keep scaling profitably.
Steady · no material change 7 re-checked, unchanged
New vs returning mix Returns and shipping Product mix Discount load

Illustrative. Your questions, your numbers.

Inside Galton · 02 · Performance

Get the full snapshot of your business' health in a single dashboard.

The numbers that decide whether the month is working, on one screen. Blended MER is revenue per ad dollar. CMER is the same figure after product costs. MER tells you what came in. CMER tells you what you kept.

Vital Signslast 30 days vs previous 30 days
Blended MER 3.2× 3.9% revenue per ad dollar
CMER 1.4× 5.2% contribution margin per ad dollar
Net revenue $412K 8.1% after returns and discounts
Ad spend $129K 4.0% every channel, one number
Contribution margin $181K 6.4% after product costs
Average order value $61 2.8% last 30 days
New-customer CAC $24 7.7% all paid spend, all new buyers
Returns 6.1% 0.3% share of gross
Galton's read Last 30 days Observational verified today

Revenue grew faster than spend, and the margin came with it.

  • MER rose to 3.2x while spend moved 4.0%, so revenue grew faster than the spend behind it.
  • CMER rose faster than MER, so more of each revenue dollar survived product costs.
  • 61% of revenue came from new customers, which is where the payback question starts.
Observational. What each channel contributed is the next question. Ask a follow-up →

Illustrative numbers.

Inside Galton · 03 · Channel Intelligence

Does each channel earn what it claims?

Every platform claims as many sales as it can. Solving attribution is the wrong approach. Galton starts from your actual sales, the one number that can't lie, then applies a range of statistical methods to reach a verdict per channel, and marks where only an experiment can settle it.

Galton's read Full history Observational verified today

Meta likely earns close to two thirds of what it claims. Google's brand line is the soft spot.

  • Meta claims $268K. The model reads it incremental at roughly two thirds of that, and GA4 grants it $128K.
  • Google claims $181K. Brand search co-moves with demand you already had; nonbrand reads incremental.
  • Claimed together: $449K. Revenue that exists: $412K.
Observational. A holdout is what would settle the Meta gap. Ask a follow-up →
Blended · the clearest read in your data
Blended MER 3.2× revenue per ad dollar, all channels
CMER 1.4× contribution margin per ad dollar
Platform claim vs Galton verdict
Meta claims $268K Likely · 75 to 90% Reads incremental at near two thirds of the claim. GA4 grants it $128K. A holdout would settle the gap.
Google claims $181K Mixed · 40 to 60% Brand search co-moves with demand you already had. Nonbrand reads incremental.

Claimed together: $449K. Revenue that exists: $412K.

Observational verdicts cap at "likely". Only an experiment earns causal language.

Illustrative numbers.

Which channels bring in new customers, and at what cost?

A channel can read badly on margin and still be worth running if it keeps buying first-time buyers cheaply. Association, not attribution.

Full history

Google's contribution-margin read is unresolved, but about 44 new customers a day move with its spend at roughly $32 implied cost each, above your ~$27 break-even ceiling on acquiring a customer.

Channel Reads New customers/day Implied CAC Last 28d vs prior Last 28d vs last year
Meta$2.6K/day spend Likely · 75 to 90% ~118 ~131 last 28d ~$22 under ~$27 ceiling +12% +34%
Google$1.4K/day spend Mixed · 40 to 60% ~44 ~41 last 28d ~$32 above ~$27 ceiling −6% +9%

Scroll the table sideways for every column.

These are associations across your history, not measured acquisitions. A holdout is what turns them into a number you can bank on.

How implied CAC is computed

Implied CAC is the channel's mean daily spend divided by the new customers a day that move with it. It is not a measured cost per acquisition and not the platform's own reported number. All paid spend together implies ~$25 per new customer, and your store actually acquires ~183 first-time buyers a day across every source, paid and unpaid.

Illustrative numbers.

Inside Galton · 04 · Customer Health

Everything you need to know about the health of your customer base.

Lifetime value, retention, and payback, cohort by cohort, on contribution margin. You see when a cohort crosses break-even before the next campaign commits the cash.

Vital Signslast 30 days vs previous 30 days
New-Customer LTV $63 12-mo: first order + predicted
CAC $24 7.7% acquisition cost
LTV : CAC 2.6× health ratio
Repeat rate 28.4% 2+ orders
New customer % 61.2% 3.1% share of revenue, last 30 days
90-day churn 41.0% recent cohort
New customer AOV $58 5.5% first-order value
Days to reorder 47 days median 1st to 2nd
Why only three tiles carry a comparison

Repeat rate and days to reorder are taken over every customer you have, LTV:CAC inherits a cohort-age confound, and churn is anchored to the window's end date. A pill on those would be a period claim the number cannot make.

Galton's read Last 30 days Observational verified today

A new customer now pays you back on day 81, thirteen days sooner than last month.

  • First-order AOV rose to $58 from $55 as the post-purchase upsell took hold.
  • CAC eased to $24 from $26, so there is less to earn back in the first place.
  • Repeat rate held at 28.4%, so the gain is in the first order, not in retention.
Observational. Cohorts under 90 days are projected. Ask a follow-up →
Cohort retention · % with a second order

Open cells are months the cohort hasn't lived yet. Galton leaves them blank rather than guessing.

Blended payback crossed at day 81 this month, from day 94.

Illustrative numbers.

You asked

What can we afford to pay for a new customer?

Galton

Between $22 and $31, depending on the line.

  • Serum line: $31, on a 34% rebuy rate.
  • Candle line: $22. Rebuys at 19%.
  • Blended break-even sits at $27.

Contribution margin, not revenue. Cohorts under 90 days are projected.

Illustrative conversation. Your questions, your data.

Inside Galton · 05 · Product Insights

Which products deserve the budget.

Every product ranked on the metrics that decide budget, with the customers it recruits and what they go on to be worth. Click a product and the row opens into its SKUs.

Galton's read Last 30 days Observational verified today

Skincare recruits your most valuable customers, on 34% of the budget.

  • Customers whose first order contained a Skincare product go on to be worth $84 on average, 1.9× the $44 from Home fragrance.
  • Home fragrance takes 51% of allocated spend against Skincare's 34%, and its CMER sits at 0.8×.
  • Inside the Serum line, the refill SKU is the rebuy engine: up 31% while the lead SKU grew 9%.
Observational. Lifetime value looks backwards, and spend follows available demand as well as value. Ask a follow-up →
Leaderboard · last 30 days · click a product for its SKUs
Product Revenue Margin Margin % Δ Units Orders COGS coverage
Serum DuoSkincare $88.4K $46.9K 53.1% ▲ 13.0% 2,140 1,880 96.4%
SKU / variant Revenue Margin Units Orders
SRM-DUO-30 30 ml $41.2K $22.1K 980 910
SRM-DUO-50 50 ml $33.6K $18.1K 690 640
SRM-DUO-RF refill pouch $13.6K $6.7K 470 330
Candle TrioHome fragrance $61.0K $22.6K 37.0% ▼ 8.0% 1,910 1,640 91.2%
Wick + Trim KitAccessories $24.1K $10.1K 41.9% ▲ 3.0% 760 700 88.5%

Scroll the table sideways for every column.

Top 3 of 24 products. Margin is contribution after product costs, from the COGS you keep in Shopify; COGS coverage is how much of the revenue carries a measured cost rather than your assumed rate.

Where the ad dollars go · campaign spend classified by category
Category Revenue Margin Direct spend Brand Catch-all Total spend ROAS CMER
Skincare $188.0K $96.5K $38.2K $6.1K $0 $44.3K 4.2× 2.2×
Home fragrance $142.5K $51.3K $61.8K $4.6K $0 $66.4K 2.1× 0.8×
Accessories $52.4K $21.5K $9.4K $1.7K $3.1K $14.2K 3.7× 1.5×
Gift sets $29.1K $11.3K $3.1K $0.6K $0.4K $4.1K 7.1× 2.8×

Scroll the table sideways for every column.

How spend is allocated across categories

Brand spend spreads across all categories, catch-all only across categories without a dedicated campaign, both revenue-weighted. ROAS and CMER here are blended and observational: they say what a category's ad dollar coincided with, not what it caused.

Illustrative numbers.

What you get

Decisions you can defend.

What actually changes once the numbers are honest.

You decide on margin, not revenue.

Galton reads the COGS you keep in Shopify, so max CAC, payback, and LTV are all contribution margin. Top-line ROAS can look healthy while the P&L says otherwise.

You find the ceiling before you spend into it.

A max CAC per product line and a payback date per cohort. You know when to push, and when the next dollar stops working.

You stop paying for orders you'd have gotten anyway.

Harvesting demand versus creating it, platform claims versus GA4, and a holdout to settle it when the stakes are big enough.

Why you can believe it

Every answer shows its work.

Observational, and says so.

When the data only shows correlation, Galton says "associated" and "co-moves". It never dresses a model up as proof.

Causal, when earned.

Causal language arrives only after an experiment: holdouts Galton designs and reads. When it says "drives", there was a test.

Fresh or silent.

Answers are dated and re-verified nightly. When cohorts are too young to read, Galton waits rather than guessing.

Also from Finch Labs

Finch, the creative strategist: which ads win with which buyers, and what to make next. See Finch →

Status

Currently in closed beta.

We onboard hand-invited brands and agencies, a few at a time, and we work the first weeks with you. Write to us and this is what follows: a call to check that your spend and order history can carry the reads, then a connect that takes minutes, then your first answers while the backfill is still running.

hello@finchlabs.ai →