Marketing keeps bolting widgets onto the site. How do we decide what stays?

The install cost of a chat widget is one line of script and ten minutes of someone’s afternoon. The visitor cost is a second of delay, a covered call to action on mobile, and a bubble that slides in over the paragraph they were reading. Nobody signs off on the second list, because nobody is asked to.

So the question is not “is this tool useful?” Almost every tool is useful to somebody internally. The question is what the visitor pays for it, and whether anyone can show what came back.

In the redesigns we run, we consistently find three or four of these stacked on a site and one that can point to a result. The other two or three are there because a team asked, nobody objected, and no one has looked at the page as a whole since.

A designed page gets reviewed carefully once, in Figma, before launch. Add-ons arrive after launch, through a tag manager, and never trigger a second review. The page you approved and the page a visitor sees stopped being the same page months ago.

Each tool also has a different owner. Legal owns the consent banner. Sales owns chat. Growth owns the exit popup. Marketing owns the announcement bar. Every one of those decisions was reasonable on its own, and nobody in the company has seen the four of them together on a 390 pixel screen.

The cost is spread too thin to trigger anyone’s alarm. A second of delay here, a corner of the screen there, one covered button. Individually forgivable, which is exactly why it accumulates.

Across our redesigns, widgets are usually the first thing a client asks us to clean up, and the first thing quietly re-added within six months. That pattern tells you the problem is not any single tool. It is that nobody assigned them space.

The recurring collision is the bottom right corner of a phone screen. A consent bar sits along the bottom. A chat bubble sits above it. An announcement bar pushes the header down. On a small screen those three eat most of the first view, and the primary button ends up underneath the chat bubble.

The other recurring failure is timing. An add-on that fires before the page has settled does not read as a promotion. It reads as the site being broken. A popup that appears while text is still shifting looks like a bug, and visitors treat it like one: they hunt for the close button without reading a word.

You do not need a developer or a redesign to do this. You need a phone, a spreadsheet and about three hours. Run the steps in this order, because the cheap evidence comes before the expensive argument.

1. Inventory. Load the homepage, one product page and one blog post on an actual phone, not a desktop browser resized. Screenshot every state: arrival, three seconds in, after scrolling half a page, on exit intent. Count everything that appears without you clicking anything.
2. Attribute. Give each item an owner's name and the one decision it was meant to influence. Anything you cannot attribute to a person is already a candidate for removal.
3. Ask for the number, not the reason. "We added it to catch leaving visitors" is a reason. "It captured 41 emails last quarter, 6 of which replied" is a number. If the owner cannot produce a number, the tool has no defence.
4. Test removal instead of debating it. Turn the weakest one off for two weeks and watch the metric its owner claims it drives. Two weeks of data ends arguments that would otherwise run for two months.
5. Assign fixed real estate. Write down which zone each surviving tool owns: bottom right, bottom bar, top bar, inline. Two tools may never share a zone. If a new tool wants a taken zone, something leaves.
6. Set an interruption rule. Consent may interrupt, because it has to. Promotion may interrupt once per visitor, then never again. Chat appears only when the visitor asks for it.

Once you have the inventory, most items sort themselves against four tests. Apply them in order and you will usually be left with one or two survivors.

• It blocks loading or shifts the layout. Remove it, or load it after the page has settled. A widget that moves the text a reader is mid-sentence in costs you more than it can return.
• It duplicates a job the page already does. Remove it. A floating "20% off" badge on a page whose headline already says 20% off is asking the visitor to read the same offer twice.
• It triggers on arrival, before the visitor has read anything. Delay it or drop it. An offer shown before the pitch is a request for trust you have not earned yet.
• It is legally required, like a consent banner. Keep it, and design it properly rather than shipping the vendor's default. The default is built to be defensible in every jurisdiction, not to be small, quiet and legible on your site.

You do not need a chat widget if nobody staffs it. A bubble that answers in four hours, or answers with a bot that cannot help, is worse than a plain email link. The link is honest about what will happen next.

Review badges belong next to proof, not floating over every page. A trust badge earns its keep beside a pricing table or a signup form, where someone is deciding. Pinned to the corner of a blog post, it is furniture.

Most consent banners can be restyled, shrunk and moved without new legal advice. What the law requires is the choice and the record of it, not the vendor’s full-width dark box with two paragraphs of text. Check the wording with whoever approved it, then take the design back.

This is a maintenance problem, not a reason to buy a redesign. If your pages are fundamentally sound and only the add-ons are fighting each other, an afternoon of auditing and a few hours of implementation fixes it. Rebuilding the site would fix it too, and would also let the same pile-up start again in six months.

Removal is the easy part. The pile comes back because there is no rule about adding. Write one, keep it short, and put it where whoever edits the site will see it. In our handovers this goes into the CMS documentation alongside the content guidelines, because that is the page a marketer actually opens.

A version that has held up:

“The site has four interruption zones: top bar, bottom bar, bottom right, inline. One tool per zone. Nothing loads before the page has settled. Nothing appears on arrival except consent. Adding a tool means naming the zone it takes and the number it must produce in ninety days. If a zone is occupied, something is removed first.”

That last clause does the work. It turns every new request into a trade rather than an addition, and a trade forces someone to defend the tool already in place. The visitor cost that nobody was tracking becomes the thing the conversation is about.