ABM stages sales will trust: conditions, windows and calibration
How to build ABM stages that sales actually uses: what each stage should mean, which signals and time windows to use, how CRM deals fit in, and how to calibrate stages against real pipeline.
An ABM stage tells sales where an account is in its buying journey, based on what the company did: ads it saw and engaged with, pages it read, AI answers it came from and deals in the CRM. Sales trusts stages when each one means something they'd act on, uses a short time window, and has been checked against which accounts actually turned into pipeline.
Most stage models fail for the same reason: they're built from what marketing can measure, not from what sales would do differently. This article walks through how to design stages, pick conditions and windows, and calibrate them.
Start from the action, not the signal
Before you write a single condition, write down what sales or marketing does differently at each stage. If two stages trigger the same action, merge them.
| Stage | What it should mean | Who acts, and how |
|---|---|---|
| Identified | We've seen this company at all (ad impressions or a visit) | Nobody yet. It's a coverage number |
| Aware | They've had enough exposure to recognise us | Marketing: keep frequency, rotate creative |
| Interested | Someone engaged: a click, comments, a ChatGPT ad click | Marketing: move to proof content and retargeting |
| Considering | Several visits or many pages read in a short window | SDR: research the account, warm outreach |
| Selecting | Pricing, demo or contact pages, or an open deal | AE: direct outreach, or active deal support |
| Won | A won deal in the CRM | Exclude from prospecting, move to expansion |
These are IntentBid's default stages, and they're a reasonable starting point. The names matter less than the rule: every stage has an owner and an action.
Pick conditions that reflect real behaviour
Each stage needs conditions built from signals you trust. IntentBid lets you combine up to six conditions per stage, each with a metric, a minimum and a time window, and choose whether the stage needs any condition or all of them. The available signals:
- LinkedIn impressions, clicks and engagements (by company, from LinkedIn's reporting)
- ChatGPT Ads clicks and visits from AI answers (ChatGPT, Perplexity and others, identified on your site)
- Website visits and pages read
- High-intent page visits: pricing, demo or contact
- Open deals and won deals in the CRM
A few principles help:
Impressions alone should never move an account past Aware
Impressions measure your spend, not their interest. A big account on a big budget will rack up impressions whether anyone cares or not. Use impressions for the first two stages only.
Engagement thresholds should scale with the account
Three LinkedIn engagements in 30 days is a meaningful signal at a 200-person company. At a 50,000-person bank, it might be three people who liked a post. If your list mixes very different sizes, either set stricter thresholds or lean on website behaviour, which is less affected by headcount.
High-intent pages beat volume
One visit to a pricing page usually says more than ten blog page views. Give high-intent pages their own condition and put it in the later stages.
Use "all" sparingly
"All conditions" makes a stage precise but rare. It works well for a stage like Selecting where you want both a high-intent visit and recent engagement. For early stages, "any" is better: there are many ways to become aware of a vendor.
Choose windows deliberately
Every condition has a window: 7, 30 or 90 days, or lifetime. The window decides whether a stage describes what's happening now or what happened once.
| Window | Good for | Risk |
|---|---|---|
| 7 days | Hot signals: pricing visits, demo pages, ad clicks | Very noisy for small accounts; stages flicker |
| 30 days | Most engagement and visit conditions | Misses slow, steady research |
| 90 days | Identified and Aware stages | Too slow to tell sales anything urgent |
| Lifetime | Won and deal-based stages | Never decays; accounts stay "hot" forever |
The usual mistake is long windows on late stages. If Selecting means "visited pricing in the last 90 days", sales will call accounts whose interest ended two months ago, and they'll stop trusting the stage after a week. Late stages need short windows. Early stages can use long ones.
Stages also need to go down. An account that met Considering last month and has gone quiet should drop back. With windowed conditions, it does so automatically, because the account is re-staged on every refresh and sits in the highest stage whose conditions it currently meets.
Bring the CRM in, both ways
Stages without CRM data have a blind spot: they don't know which accounts are already in a sales process, or already customers.
- Deals in. Use open deals as a condition for a late stage, and won deals for a Won stage. This stops an active opportunity from showing up as "Considering" just because the champion hasn't visited the site this month.
- Stages out. Sales lives in the CRM, not in your ABM tool. IntentBid writes the stage, score, 30-day LinkedIn impressions, clicks and engagements, engagement rate, intent tags and last campaign to company records in HubSpot every day, so reps see the stage where they work and can build views and workflows on it.
If sales has to log into a separate tool to see a stage, they mostly won't.
Calibrate against pipeline
A stage model is a hypothesis. Test it once you have a quarter of data.
- Pull every target account with its stage at the start of the quarter.
- Mark which ones created an opportunity during the quarter, from the CRM.
- Compute the opportunity rate per stage.
You're looking for a staircase: each stage should have a clearly higher opportunity rate than the one below it. Common patterns and fixes:
| What you see | Likely cause | Fix |
|---|---|---|
| Two adjacent stages have the same rate | The conditions don't separate behaviour | Merge them, or make the higher one stricter |
| Selecting has a lower rate than Considering | High-intent condition is too loose (careers page counted as "contact"?) | Tighten the page rules |
| Almost every account is in one stage | Thresholds too low or windows too long | Raise minimums, shorten windows |
| Opportunities come from Identified accounts | Sales is sourcing deals the program never touched | Fine, but don't claim them as influenced |
With small lists, the numbers will be noisy. Look at direction, not decimals, and recalibrate each quarter rather than each week.
Use intent tags to say why, not just how much
A stage says how far along an account is. It doesn't say what it's interested in. Intent tags fill that gap: they come from the themes of the campaigns an account engaged with and the pages it read. A campaign named "ABM | [Security] | Q4" tags every account that engaged with it as "Security"; a page rule tags accounts reading pages with "/soc2" in the address the same way.
Sales then gets a sentence they can use: "Considering, interested in Security and Pricing", not just "score 64". For the trade-offs between this kind of first-party intent and bought intent data, see Intent from your own ads vs. third-party intent data.
Agree the model with sales before launch
The last step is organisational. Run a 30-minute session with two AEs and an SDR:
- Show the stages and conditions in plain language.
- Show five real accounts in each stage and ask if they'd agree.
- Agree what happens at each stage (who, and within how long).
- Agree that you'll review the model quarterly with the calibration table.
That session is the difference between stages that live in a dashboard and stages that change what reps do.
How IntentBid helps
Stages, conditions, windows and intent tags are yours to set in Stages & intent settings, and every account is re-staged with each refresh. See Set up ABM stages and intent tags. The Account engagement report filters by stage and intent, and saves any filtered set of accounts as an audience for a LinkedIn campaign.
FAQ
How many stages should we have?
Five or six, including Won. Fewer can work. More than seven usually means two stages share the same action.
Should stages and scores both exist?
Yes, for different jobs. A score ranks accounts within a stage, so sales knows which Considering account to call first. The stage decides who owns the account.
Can an account skip stages?
Yes. An account sits in the highest stage whose conditions it meets. A company that books a demo straight from a ChatGPT ad lands in Selecting without passing through Aware.
What if sales disagrees with a stage?
Ask for examples, then change the conditions, not the account. Hand-edited stages break the model and make calibration impossible.
How often should accounts be re-staged?
On every data refresh. Stages that update weekly are already out of date for the late stages.