Blog
Notes from the team.
Practical notes from the people who design packaging, build stores and apps, run paid media and social, and train AI agents at Boson Group.
Each article is a checklist you can use this week: what to set up, what to measure and the mistakes we see most often — in branding, web and mobile, performance marketing, social media and AI.

Measure before you spend: the tracking to set up before your first ad dollar
Most ad budgets are judged by numbers nobody agreed on: the platform says one thing, the CRM another, finance a third. A week or two of setup before launch fixes that — conversion events that mean money, UTM rules people actually follow, offline sales sent back to the platforms and one weekly dashboard the whole team reads the same way.

Rebrand without losing recognition: what to keep, what to change and how to test
Your regular buyers don’t read your pack. They find it by a color, a shape and one or two marks they have seen a hundred times, and a redesign that moves the wrong one makes them walk straight past you. Here is how to decide what stays, what changes and how to check it before the new pack ships.

What to prepare before you hire a web agency: a brief that gets accurate estimates
Send three agencies the line “we need a new website” and you will get three estimates that are miles apart — and all three can be honest. The gap comes from what the brief leaves out. Here is what to write down first, so the numbers that come back describe the same project.

Running a short-video content factory: 30+ creator videos a month, and which ones sell
One good creator video is luck. Thirty a month is a production line: a few formats that repeat, one-page briefs, a two-day approval loop and a rights log. Here is how to set it up so you can see which videos actually move product.

Your first AI agent: how a mid-size company picks the use case and runs the pilot
Most first AI projects fail at the choice, not the model: the job is too vague, too rare or impossible to measure. Here is how to pick a first use case with volume, clear rules and a number you can check, and how to run an eight-week pilot that ends in a decision.

Choosing influencers by data, not follower counts
A creator with 500,000 followers can sell less than one with 15,000, and the follower count won’t tell you which is which. The data that does is mostly free: who the audience is, how it responds and how past sponsored posts performed. Here is how to read it, how to brief the creator and how to measure what you got.

Packaging for a 200-pixel shelf: designing for marketplace thumbnails
On a marketplace or in a delivery app, your pack is a small tile in a grid of forty, seen on a phone for a second or less. Most packaging is still approved as a full-size print at arm’s length. Here is how to design a pack that survives being shrunk, and the main image that goes with it.

Landing pages that convert: one offer, a clear first screen and tests that mean something
Most landing pages lose visitors for ordinary reasons: two offers fighting over one button, a first screen about the company instead of the deal, a form that asks for a phone number nobody plans to call. Fix those first. Then run tests whose winner is still the winner a month later.

What to feed Performance Max and Advantage+: the inputs that decide results
Performance Max and Advantage+ took most of the levers away from the media buyer, and many teams concluded there is nothing left to manage. There is: these campaigns learn from what you give them. Here is what to put in — the feed, the values, the creative, the signals and the fences — and what to check every week.

Customer-support bots: what to automate and how to hand over without starting over
Customers don't mind a bot. They mind typing their order number three times and then explaining the whole problem again to a person. Here is how to decide what a support bot should handle, how to pass a conversation to a person with the context intact, and how to keep the answers right after launch.

One key visual, a hundred formats: building a KV system that survives every channel
A key visual is usually approved as one beautiful horizontal poster. Then it has to become a square post, a vertical video, a shelf wobbler and a tiny mobile banner, and somewhere around the twelfth format the idea falls apart. Design the system first and the picture second.

AI for retention: how to raise LTV without hiring a data team
Right now your database holds customers who are about to leave — and you almost certainly don’t know it yet. Retention is the cheapest way to grow revenue: a 5% increase adds 25–95% to profit, and winning a new customer costs 5–25 times more than keeping an existing one (Bain, Harvard Business Review). In 2026 you don’t need a data team for it: a CRM export, a no-code model and a list of fifty names are enough to start.

How to get into an AI’s recommendations: a GEO audit checklist for e-commerce
Shoppers increasingly choose not in Google or Yandex but in an answer from ChatGPT, Alice or Perplexity. Here is what happens to a store the AI simply doesn’t see — and what an e-commerce team can do to get its products into the recommendation.

Online store speed: what actually affects sales and what to fix first
A perfect PageSpeed score is not the goal; shoppers who reach the order confirmation are. In most online stores the speed that matters comes down to four things — images, scripts, third-party tags and the checkout — and they are not equally worth your time.

Computer vision for retail and brands: what a pilot really needs
Telling a genuine handbag from a fake by its photos, checking that your products are on the shelf, catching a crooked label on the line: these are practical jobs for a vision model now. The hard part is rarely the model. It's the images, the labels and the edge cases, and here is how to prepare them.

A creative-testing system for paid social that a small team can run
On Meta and TikTok the creative now does most of the targeting, which makes testing ads the main job of a paid social team. Most teams still test at random: new ads when the old ones tire, winners picked by gut. Here is a simpler system — one hypothesis per test, a set budget, a rule for calling the winner and a monthly calendar.

Brand guidelines people actually use: templates over rules
Most brandbooks are a long PDF that people open once, at the handover meeting, and never again. The people who make your brand every day — a marketer on deadline, a sales manager building a deck, a freelancer hired last week — need an answer in a minute, not a manifesto. Here is what a working brandbook contains and how to keep it in use.

Price, stock and review monitoring: what to track on marketplaces and partner sites
Your products sit on Amazon, Walmart, a dozen retailer sites and a few reseller stores you've never heard of, at prices you didn't set. Monitoring shows you where things go wrong, but only if you track the right fields, check at the right rhythm and know who acts on each alert.

Do you need a mobile app or a better mobile site?
An app feels like the serious option and a mobile site like the compromise. Often it is the other way around: an app earns its place only when people come back often, want to be notified, need it offline or use the phone itself — and it keeps costing money every month after launch.

A community-management playbook: response times, tone and the escalation ladder
Every reply under your posts is read by far more people than the one who wrote the comment, and an unanswered complaint is a public answer too. Community management works when it’s a written system rather than one person’s mood: response times, a voice, a ladder for complaints, moderation rules and a weekly habit of turning comments into ideas.

Limited editions and new flavors: a label that stands out and still belongs to the range
A limited edition has two jobs that pull in opposite directions: stop the shopper who walks past your core range every week, and still look like the brand they already trust. Lean too far one way and it disappears among the regulars; too far the other and it looks like someone else’s product. Here is how to set the rules before the designers start.

A website audit in one week: from analytics to a prioritized backlog
An audit that ends as a 90-page PDF changes nothing. A useful one takes a week, checks five things in a fixed order — analytics, funnel, UX, technical health, content — and ends with a short, ranked list of tasks your team can start on Monday.
Contact
Send a brief — we reply with a time and cost estimate within 24 hours.

