Intent based targeting starts with pages, not job

Traditional marketing targets fixed segments, like a job title or an industry. Demandbase describes intent-based marketing as the opposite: it works from real-time behavior, not from a fixed audience list. The question changes from "who is this company on paper?" to "what is it doing right now?"

Realize lists the signals that count as high intent: searching "buy now", reading product reviews, adding to cart and requesting a demo. Each of those is an action. A company's size or sector tells you none of that.

Platforms collect and score these signals with NLP, machine learning and AI, according to Realize and Demandbase. The output is a ranked list of accounts. Your sales team spends its time on companies that have already shown interest, not on a cold list.

Picture two companies that both fit your ideal customer profile exactly. One read your pricing page and a comparison page this week. The other hasn't touched your site. On a demographic filter they look identical, and on behavior they're nowhere near each other.

It doesn't apply everywhere. If you sell into a few dozen named accounts you already know by heart, a behavior filter adds little, because your list is the targeting. It also adds little if your buyers rarely research online before they talk to someone.

Before you buy anything, write down the three actions on your own site you'd treat as active interest. If you can't name them, no tool will name them for you.

B2B intent data is either first-party or third-party. First-party data comes from your own website, CRM and email. Third-party data comes from external networks such as Bombora and Demandbase, as Demandbase and Jeluvi both describe.

The two answer different questions. First-party data tells you which companies are already looking at you. Third-party data tells you which companies are researching your category somewhere else, before they've found you.

HubSpot adds a useful split: active intent and passive intent. A demo request is active, a sign of high buying readiness. Reading blog posts is passive, a sign of early research. Treat them as two stages, not one pile.

The ad platforms sell intent too. Jeluvi points to Google's In-Market and Custom Segments and LinkedIn's Matched Audiences. They're useful for reach, but they hand you an audience, not a company that came to your door.

The trade-off is plain. First-party data only covers companies that have already found you, so it can't tell you about the rest of your market. Third-party data covers the market but has fewer details about what a company actually did.

This week, write two lists. One is the first-party signals you already collect. The other is the market you can't see. That gap tells you whether you need third-party data at all.

Here's the question every ranking guide skips: how do you tell a serious buyer from someone browsing? Most guides say "collect intent signals" and stop. The answer sits in your own visit data, read at company level, on three things: which companies came back, which pages they read, and in what order.

See, one visit means little. A company that reads one blog post is showing passive intent, likely early research. A company that reads your pricing page, returns within a week, and then opens a comparison page is doing something else. It's evaluating options, and the order is the tell: education first, then price, then alternatives.

You can run this on your own site this week.

1. Take the last 7 days of visits and group them by company, not by page.
2. Mark every company that read pricing, a comparison page or your docs.
3. Keep only the companies that came back after the first visit.
4. Check the order. Pricing then comparison is a stronger pattern than a blog post alone.
5. Rank what's left as a priority list, and have someone look at the top of it.

Some visits won't resolve to a company at all, and that's fine. You're reading the ones that do.

Watch for false positives. Jeluvi warns that false positives and stale signals are common failures. A competitor researching you can look exactly like a buyer. So can a student writing a paper. Check the company's name before you act, and drop the ones you recognise as neither buyer nor prospect.

Now the case where this fails. If your site gets little traffic, you won't have enough visits to see a pattern. Three companies on pricing in a month is an anecdote, not a rule. In that case, lean on third-party intent data or content syndication until your traffic grows.

The decision is simple. Pick the one page order you trust most, and treat it as your first priority signal.

The best argument against everything above is practical. First-party signals are only as good as your traffic. A site with a few visits a week can't show you which companies return, so reading its pages is guesswork.

For that site, third-party data is the honest answer. Bombora and Demandbase, both named by Demandbase and Jeluvi, watch research across many sites. They can tell you a company is looking at your category even if it has never seen your pricing page.

Be honest about the cost. Jeluvi calls stale signals and false positives common failures, and third-party signals are the hardest to check. You can't see what the company actually read, so you can't judge the order. You're trusting a score.

Privacy pressure makes this worse, not better. Realize and Jeluvi both say that privacy regulations, such as Apple's ATT, have caused signal loss and pushed marketers toward first-party data. The source most people buy first is the one getting shakier.

So the choice isn't ideology. If your site has enough traffic to show repeat visits, start with first-party data and add third-party data for the market you can't see. If it doesn't, buy third-party data or use content syndication, and plan to switch your weight later.

Either way, write down what you'd need to see to change your mind. A date to recheck, such as the end of next quarter, is enough.

Two things moved recently. The first is signal loss. Apple's ATT is the example both Realize and Jeluvi give, and the result they describe is a push toward first-party data you collect yourself.

The second is AI. HubSpot reports that 58% of marketers say AI referral traffic carries much higher intent. When someone asks an answer engine a detailed question and then clicks through, that click often arrives already informed. Treat referrals from answer engines as a signal to read, not just a source line in a report.

The research brief also notes that sources now put more weight on writing explicit rules and running holdout tests, rather than on collecting more signals. The shift is from "gather everything" to "prove which signal predicts buying."

The market matters. A US B2B team has to be careful with third-party data under CCPA, according to the research behind this piece. For an Indian team, none of the four sources I drew on says anything about India's rules, so take your own legal advice there rather than assuming the US position carries over.

Both teams can do the same thing with budget. First-party data costs you tracking and time, not a data licence. Whatever the price is in your market, quoted in USD or in rupees, check it against what you'd lose by flying blind.

The practical step: review what you collect and where it comes from, and note which signals depend on third-party tracking you can't control.

A signal without a rule is just a list. Jeluvi says a proper targeting rule has three parts: a qualifying signal, a freshness window and a named audience. Leave one out and the rule gets vague fast.

Here's one for a US team. Signal: visited the pricing page. Window: within the last 7 days. Audience: accounts on your target list in the US. An Indian team would write the same rule with its own named audience, such as accounts on its Indian target list.

The freshness window matters because intent decays. A company that read pricing 7 days ago is in a different place from one that read it months ago. The research brief leaves the best window per signal as an open question, so start with 7 days and adjust from what you see.

Then test it. Jeluvi recommends holdout tests: randomly leave a control group out of your targeting and compare conversions. If the targeted group doesn't convert better, your signal isn't predicting anything, however clean it looks.

The trade-off is real. A holdout means some companies showing genuine interest get no follow-up for a while. And with low volume, a holdout is noisy, so a small result proves little.

Run one rule, one window and one holdout first. Add a second rule only after the first has earned its place.

PageFox is the wrong choice for a site with little traffic: it tells you which companies are already visiting, it does not bring new visitors.

Decide which single page pattern you'll treat as a priority signal, then write it as a rule with a signal, a 7-day window and a named audience. This week, list the companies that visited in the last 7 days and mark who read pricing and came back.