Join Venice

Become a member to get exclusive access to:

⁠A digital community of over 165 European eCommerce founders
⁠⁠Access to discounts from over 50 tools and vendors
Exclusive in-person and digital events
October 29, 2026
Marketing

Your Black Friday Game Plan: Leonardo Caracas on running 15 Black Fridays at once

Leonardo Caracas, Partner at Jump Ventures, runs Black Friday for a portfolio of 15 e-commerce brands. In this Venice Experts session he explains why a product launch beats a deeper discount in the ad auction, how to pick the offer with a silent promo test, how to set a COGS ceiling and model the margin before November, and how phone alerts and hourly media-buying agents run the campaign once it starts.

Leonardo Caracas, Partner at Jump Ventures · 28 October 2026 · Venice Experts

Overview

Leonardo Caracas runs Black Friday for a portfolio of 15 e-commerce brands with combined annual revenues of €250M+, and this was his 10th one. His playbook splits the season into three phases: pre-Black Friday (the whole of "Black November"), the moment itself, and the 1 to 15 December window up to carrier cutoffs. The pre-work is about choosing between a product launch and a promo, testing the offer silently, fixing a COGS ceiling and modeling the margin before you commit. The in-campaign work runs on real-time phone alerts and, in Jump's case, hourly media-buying agents with guardrails. Almost everything he showed runs on ABC Store Analytics, a free Shopify app Jump incubated, connected to Claude through MCP.

Five takeaways for builders:

  1. Treat Black Friday as a product-launch slot, not a deeper discount. A relevant launch wins the ad auction; a percentage sign is a remarketing play with high CPMs and few new customers.
  2. Pick the offer by running the silent promo hack in September and October: switch the site to a candidate promo for two or three days, tell no one, and read the lift in conversion, AOV and profit. Repeat every year.
  3. Set the maximum COGS you will accept and run the margin scenario with every cost in it before you set the discount. A very high revenue level at 2 to 3 percent profit is just processing orders.
  4. Expect buyers to wait until about ten days out, peak on the Friday, overspend on the weekend if you let Meta run, and come back for gifts from 1 to 15 December.
  5. Match sales to fulfillment and customer-support capacity, and set your own cutoff two or three days inside what the 3PL tells you, per country.

Three phases, one month

Caracas means the whole of November when he says Black Friday, "or Black November, as we like to call it." The pre-phase is the work of the previous two months, though he says you can still apply the learnings in the two weeks left. The post-phase runs from 1 December to your cutoff dates, which for most players fall from the 12th to the 15th of December. Black Friday was barely a thing in Europe when he started out in 2015-16; it has since grown into the big event for everyone in e-commerce.

Product or promo

The first pre-season question is product or promo. Caracas's warning comes from experience: if you lean on promo and get more aggressive each year, you reach a point where there is nothing left to say. After exhausting every merchandising mechanic for his main brand, he found that reserving the biggest product launch of the year for Black Friday week beat the competition on bidding, because the ad had something more relevant to say than another discount.

Patrick Crowley asked what changed. Caracas pointed to a Warner Brothers licensing collection his team held back until Black Friday week. They launched it, cut all discounts for the week, and saw much higher margins while beating competitors in the auction.

"all of a sudden, the CPMs drop by 5 times."

— Leonardo Caracas, 00:19:37

You can do promo, product, or both. Caracas usually treats it as either-or: when the launch is the message, there is no offer on top.

Test the offer silently, every year

If you are going promo and do not know which mechanic to run, do not guess. In September and October Caracas picks random two- or three-day windows, switches the whole site to a candidate promo, and keeps pumping traffic as normal without an email or an ad announcing it. With no context driving the lift, the change in conversion rate, AOV and profitability tells you which offer people actually want. Of three promos tested, one was a clear winner.

The test is site-wide, not per product: 50 percent off versus buy three get two versus 30 percent off over $100, with visitors getting whichever offer is live when they land. An ad-plus-landing-page-plus-coupon variant works if you want to compare advertised offers.

Protect COGS, protect profit

Two things Caracas says he has shot himself in the foot on. First, decide the maximum cost of goods sold you are willing to reach through discounts or bundles, so there is still room to advertise. Second, build a proper simulator with every cost in it. Without one, a very aggressive sale can post a very high revenue level at 2 to 3 percent profit, at which point you have complicated your operation by adding volume and made no money. His slide put it as: know your cost floor before setting any discount, and keep margin positive after ad spend. The simulator on the next slide models profit across discount and ROAS scenarios, then turns the analysis into targets.

Set the target, then ask what has to give

Caracas demonstrated the simulator live. ABC Store Analytics pulls Shopify, Meta, Google Analytics, Klaviyo, TikTok ads and Amazon together with COGS and transport costs, and the breakthrough came when the team connected all of it to Claude through MCP. The connector ships standard prompts: monthly review, quarterly review, profit review, Black Friday review. He asked for a profit review of one brand for September and got total sales, net profit, contribution margin, MER, where profit leaks, and a sensitivity analysis.

Then he played with it. Planning 1 million for November, net margin would be 24 percent if every other unit-economic held. Pushing COGS from 25 to 30 percent for a bigger discount drops it to 18. Trimming Meta by one or two points of spend brings it to 19. Flip it round and ask for 20 percent instead of 15, and the model lists which P&L categories have to move. The report also drills into markets and products, so you push the countries with the most margin headroom and the products with the lowest COGS.

Asked how far ahead he runs this, Caracas said the plan is never set in stone and he refreshes it constantly. The one thing you cannot update in real time is the stock purchase, so the goal per company is set months ahead to size how much stock risk to take.

The November curve

A Black Friday review report builds a base, low and high case from your own history, which gives you a plan to start from even if you have never run one. On the pattern across his European brands: things improve from 1 November, and improve more if you have activations or a launch live from day one, but shoppers are skeptical early in the month and wait for a better deal. The only way to start strong on 1 November is to say this is the best promo of the year and it will not get better. Otherwise the window opens ten, at most twelve, days before Black Friday.

Peak lands on the Friday if you have done everything right. Then watch Saturday and Sunday: Meta sees the strong Friday signals and will spend fast over the weekend unless you keep a tight hold on budgets. Cyber Monday is usually weaker but is another window. From 1 December to the 15th you can still make relevant sales, and for some of his giftable brands that December window is bigger than November.

Run the campaign from your phone

The in-campaign hack is the KPI Alerts panel in ABC Store Analytics, installed on his phone. Caracas sets alerts on sales, ad spend and MER, for example sales 30 percent down or 50 percent up with a minimum of $1,000 a day, and gets notified in real time across brands moving a lot of volume. He says he has seen nothing else that alerts him this way when things shift. If things go bad, pause and stop bleeding cash. If they go well, put more wood on the fire.

Agents with guardrails

One of Jump's main projects this year is a 24/7 Meta agent, trained by the team, working across time zones while Caracas sleeps. Each store's agent posts an hourly update: sales so far, pace to the day's target, current MER, what is wrong and what it is about to do. It pauses ads, moves budgets and creates ads, on a rules platform Jump built, with every action logged and its outcome fed back so the agent decides better next time. Each brand has its own decision metrics.

Proposed actions arrive in a Slack channel where Caracas approves or rejects each one with a reason, such as a product nearly out of stock or a country whose margin collapses above a certain MER, so the agent learns the team's thinking.

For founders not there yet, Caracas said the starting point is simpler: connect your data, then ask Claude for your hourly sales and MER patterns, the thresholds at which an ad should be switched off or a budget raised, and rules you can act on by hand. A second agent reads the last 30 days of ad formats and best-selling products, writes new ads daily, and passes them through a critic agent that checks brand, offer and persona. AI picture ads built this way are one of Jump's top-performing campaigns. On a brand with 1,000 products, it finds high-converting SKUs that have never had ad spend.

Sell only what you can ship

On 1 or 2 December you are excited about the sales. Caracas's final warning is to check whether fulfillment and customer support can actually handle them, and to slow down before the cutoff if not. He learned it by burning himself: a customer whose Christmas order does not arrive asks for a refund and does not come back, and the money spent acquiring them is gone.

Carriers and 3PLs announce delays every year, so do not take their cutoff date on trust. Set your own with a two- to three-day buffer, know it per country, and keep checking how many orders a day the warehouse can really ship.

Q&A highlights

Q (Luke Hersheson): We have a 2.0 version of a product coming back after six months out of stock, planned for just after Black Friday. Should it launch during Black Friday instead?

Yes, if it is a hero product you can build the whole campaign around. A secondary category or accessory is not relevant enough to carry the Black Friday message. Caracas usually runs a launch without an offer: "It's about the product launch, it's not about the offer." (00:42:06) You can stack a launch and a discount, but he treats it as either-or. Luke said moving the launch into the Black Friday window probably made sense.

Q (Jennifer Bailey): ABC looks like it connects one Google Analytics account at a time. How do you run many stores, and what about email platforms other than Klaviyo?

You create organizations, each with multiple Shopify stores, and each store has its own Meta, Google and other connections. On email, it is Klaviyo only for now. Caracas uses the Klaviyo data daily to explain good days the ads do not account for: ask Claude what popped yesterday and it points to the email.

Q (Paul Kalo, Zipster): How do costs get into ABC? Can it integrate with accounting software like Exact so profit is real rather than assumed, and can Claude write back to it?

The Cost Settings tab pulls product COGS straight from Shopify, and Amazon costs if you have Amazon. Gateway costs, custom costs and shipping are set by hand, as fixed or variable expenses by month with start and end dates, or as a lump sum per month. On writing back through Claude, Caracas hedged: "I think so. I'm almost sure." ABC is an app Jump incubated, and its team will help with integration issues. Before it, the portfolio ran managerial reports in Looker Studio, which he described as slow and prone to breaking.

Resources Leonardo referenced

  • ABC Store Analytics — the free Shopify analytics app, incubated by Jump Ventures, that unifies Shopify, Meta, Google, Klaviyo and Amazon data; the KPI alerts, profit review and Black Friday review prompts he demonstrated all run on it.
  • Claude — the standard reports, sensitivity analysis and agent rules he showed are Claude conversations running on ABC data connected through MCP.
  • Model Context Protocol (MCP) — the connector standard that links ABC Store Analytics to Claude.
  • Klaviyo — the only email platform ABC connects to for now.
  • Jump Ventures — Caracas's growth-venture firm.

About Leonardo Caracas

Leonardo Caracas is a Partner at Jump Ventures, where he oversees a portfolio of 15 companies with combined annual revenues of €250M+ and is running 15 Black Fridays this year. He started his first e-commerce company in 2015 and scaled it to nine figures before he and his partners moved into acquiring e-commerce brands across Brazil, the UK, the US, Australia and Spain. He is based in Amsterdam and was one of the first ten members to join Venice.

Ready to connect?

Start exploring the Venice community and discover new opportunities to grow your business.

Browse Members

Join

Apply today to join Europe's most valuable network of Consumer focused founders.

Apply Now