Fritz, Molecule’s AI agent, answers in plain English from your real sales in Shopify, GA4 or Mixpanel: which ads earn, which burn, and where the budget should go. On Pro it drafts the budget change too. No SQL, no exports, no afternoon lost in filters.
| Campaign | Spend | ROAS |
|---|---|---|
| Prospecting · Broad | €6,410 | 4.2 ↑ |
| Retargeting · Lookalike | €5,880 | 1.1 ↓ |
Every answer is built from real orders in Shopify, GA4 or Mixpanel, tied to the ads that earned them. With Shopify, each order is also counted once across Meta, Google, TikTok, Snapchat, Pinterest and ChatGPT, net of refunds and tax. So “what’s working” is measured in sales you actually made, not what each platform claims for itself.
Winners and losers down to a single creative: ROAS, CAC and sales per source, campaign, ad set and ad. With Shopify, net of refunds and tax.
Ask for reallocation suggestions and Fritz backs them with spend, ROAS and the budget engine’s view, then drafts the change for you to apply in one click.
Compare any two periods and Fritz breaks the swing down into the campaigns, countries and products that actually moved it.
Put Meta’s or Google’s reported numbers next to Molecule’s verified attribution. Fritz pulls both and always tells you which is which.
Revenue, profit and return rate per product, variant and SKU, including units still in stock, so you never scale an ad for a product about to sell out.
LTV curves, CAC payback and repeat-purchase retention per channel. See which source brings buyers who come back, not just buyers who click.
This isn’t a chatbot bolted onto a help center. Fritz queries the same attribution engine as your dashboard: every measure, every breakdown, in your display currency.
Total sales, ROAS, CAC, CPO, CPC, CPM, AOV, conversion rate, first-time-customer metrics and returns, from channel totals down to one ad, in any country. With Shopify, net sales and net ROAS too.
LTV curves, CAC payback windows, cohort sizes, retention and repeat-purchase behaviour, grouped by channel and cohort, over any horizon.
Revenue, profit, cost, ordered and refunded quantities, return rate and available stock per product, variant and SKU, straight from your store.
Today’s delivery pulled straight from Meta and Google Ads. Platform-reported and clearly labelled as such, so nothing gets silently mixed.
Fritz reads your data and explains it. On Pro it can also draft budget changes from the budget engine. Each one lands as a card with the evidence, and nothing changes in your ad accounts until you click Apply.
You choose how much say it gets: Guarded, Balanced or Pure AI. It never writes to your store. Answers come with the numbers behind them, in threads you can pick back up tomorrow. Or ask from Claude or ChatGPT through Molecule MCP.
Each card shows the plan it starts on:Starter+Pro+EnterpriseNo pill means every plan, Free included.Compare plans →
Ask Fritz instead of building a report. It answers from your real sales, explains a swing down to the campaign and drafts the fix for you to apply in one click.
| Campaign | Spend | Net ROASMolecule | ROASMeta | vs MondayMolecule |
|---|---|---|---|---|
| Prospecting · Broad | €1,240 | 1.9 ↓ | 3.0 | −1.3 (−41%) |
| DE · Prospecting | €610 | 1.4 ↓ | 2.7 | −1.6 (−53%) |
| Retargeting 7d | €880 | 3.8 ↑ | 4.6 | +0.2 (+6%) |
| Molecule | TikTok | |
|---|---|---|
| Net ROAS | 2.43 | 4.02 |
| Orders | 86 | 141 |
Fritz answers from your verified sales and ad data, with tables you can check.
Compare two periods and Fritz breaks the change into the campaigns, countries and products that drove it.
Fritz drafts budget changes, pauses and negatives as Apply cards that follow your budget policy.
Fritz pulls live Meta and Google detail and the state of your ads when the question needs it.
Conversations are saved, and Fritz says what it's doing while it works.
The biggest budget moves and the money at stake, applied in place.
Log in, open Fritz, and ask the question you’d normally spend an afternoon answering.
Molecule