Buying signals in ecommerce and retail

A signal is work with a deadline, not news

Most intent data is a list of companies that did something interesting. That is not the same as a list of companies about to spend money, and the gap between the two is where trigger-based outbound quietly fails.

A useful signal has three properties. It creates work that somebody now has to do. The work has a date attached, imposed from outside rather than chosen. And the company cannot easily do it with what it already has.

Test any candidate signal against those three. A funding round creates optimism and a spending intention, but no specific task with a specific deadline, which is why the reply rates are so much worse than the hype suggests. A migration to a new ecommerce platform creates thousands of concrete tasks with a launch date, and the company is by definition short of hands, since it just decided to change something structural.

This is also why the same signal has different value to different sellers. A replatform is a strong signal if you sell anything that touches product data, and irrelevant if you sell warehouse racking. Signals are not universally strong. They are strong relative to what you sell.

The signals that hold up

Four changes in retail reliably precede spending.

A platform migration. Moving between ecommerce platforms means every product, image, description and attribute is touched. The project has a public launch date, an internal owner and a budget already approved. It is detectable from the technology on the site, and it is the strongest single signal in the category for anything data-adjacent.

A marketplace or country launch. Listing on a new marketplace or opening a new country means meeting somebody else’s data requirements, usually stricter than the ones the company applies to itself, and often in another language. Detectable from a new domain, a new language on the site, a new marketplace storefront or a job posting for that market.

Catalogue expansion. A sudden increase in the number of products creates a proportional increase in the work of describing, photographing and listing them. Detectable by counting the product sitemap on a schedule and watching the delta, which almost nobody does and which is entirely free.

A first specialist hire. Not headcount growth in general. The first ecommerce manager, the first catalogue coordinator, the first performance marketer. A first hire in a function means the company has decided that function matters and has just given somebody a mandate and no tooling. That person is the easiest buyer in retail to reach and the most likely to answer.

The signals that look good and are not

Three popular triggers underperform badly enough to be worth naming.

Funding announcements. Every sales team in the market subscribes to the same funding feeds, so a funded company receives dozens of near-identical emails within days. The signal is real and the channel is saturated, which nets out negative. If you use funding at all, use it four to six months late, when the money is being spent rather than announced and the inbox has cleared.

General hiring growth. Retail hires constantly, mostly for warehouse and store roles that say nothing about tooling budget. A count of open roles is noise. A specific role is a signal. Filter on the title, not the volume.

Website visits from a target account. Anonymous visitor identification tells you a company has an internet connection and an employee with curiosity. Where it works at all, it works for large accounts with a long buying cycle. For mid-market retail it produces confident-looking lists with no predictive power.

The common thread is that all three are easy to buy and easy for everyone else to buy. The signals that work take a little effort to observe, which is precisely why they still work.

Every signal has a decay window

A signal is a claim about what a company is doing right now, and every one of them expires. Sending on an expired signal is worse than sending on none, because the message names something the reader has already moved past and marks the sender as automated.

The first-hire window is the one most often mistimed. Writing in somebody’s first week reaches a person still finding the coffee machine. Writing in week five reaches somebody who has now seen the mess, has been asked what they intend to do about it, and has a mandate to spend. The difference in reply rate between those two weeks is larger than the difference between most subject lines.

Where each signal is observed

None of this requires a paid intent platform. Every signal above is observable from public sources.

• Platform migration. Technology fingerprints on the site, changes in URL structure, a new checkout domain, or the platform’s own public customer directory.
• Marketplace launch. Marketplace seller pages, a new country domain, hreflang tags appearing on the site, or a shipping page listing a new destination.
• Catalogue expansion. The product sitemap, fetched monthly and diffed. A single scheduled script covers an entire market.
• First specialist hire. Job boards and company career pages, filtered on title rather than count.
• Trade fairs. Published exhibitor lists, which are usually free and complete.

The work is in the scheduling, not the access. A signal observed once a quarter is a historical fact. The same observation run weekly is a pipeline, and the difference is a cron job.

Ranking signals instead of collecting them

Teams that adopt signal-based prospecting usually end up with too many signals and no way to choose between two companies that both have one. Scoring solves this, and it only needs three inputs.

Strength. How directly does this signal imply the work you sell? A replatform for a product-data vendor scores high. The same replatform for a logistics vendor scores low.

Freshness. Where in the decay window is this company? A score that halves as the window closes keeps the list honest without anyone having to prune it by hand.

Fit. Does the company match your qualifying criteria anyway? A perfect signal at a company too small to buy is still a company too small to buy, and this is where signal-led lists most often drift off target.

Multiply the three and work the list from the top. The value is not precision, it is that the sequence of who gets contacted becomes a decision rather than a queue order.

Writing on a signal without sounding like surveillance

There is a line between observant and unsettling, and it is easy to cross with public information.

The rule that keeps a message on the right side is simple. Reference things the company published on purpose, and reference them the way a person in the industry would, not the way a database would.

A new marketplace storefront was published deliberately, so mentioning it is normal. A named individual’s start date, pulled from a profile and merged into a template, was not published as an invitation to be tracked, and reads badly even when technically public.

The second rule is to state the observation and move on. One sentence naming what you noticed, one sentence on why it creates the specific problem you solve, one question. Messages that spend three paragraphs proving how much research went in read as effort spent on the sender rather than on the reader.

And never mention a signal you cannot substantiate. Guessing that a company is replatforming, when it is not, ends the conversation in one line and does it in a way that is remembered.

Building this without an intent data subscription

A working signal pipeline for a retail market needs four parts, and none of them is expensive.

A base list of companies with domains, built from registers and place data. A scheduled observer that fetches the public artefacts above and records what changed since last time. A scoring step that combines strength, freshness and fit. And a writing step where a person, not a template, turns the observation into one sentence.

We run all four for the markets we cover, which is why our campaigns lead with what a company just did rather than with what it is. If you would rather run it yourself, start with the sitemap diff. It costs nothing, it is the signal nobody bothers to watch, and in a catalogue business it is the one that most reliably precedes a purchase.