01
Separate funnel stage from rep activity
Also called Lifecycle stage vs lead status
Funnel stage says where a contact sits on the way to becoming a customer and only moves forward. Lead status says what the rep is doing with them. Keeping the two apart fixes reporting, routing and most CRM arguments.
Why it matters. When one field tries to do both jobs, reports double-count and leads fall through gaps.
How to do it
- Use a standard set of stages: lead, marketing qualified, sales qualified, opportunity, customer.
- Use lead status for activity: new, attempting, connected, qualified, nurture, do not contact.
- Move fit and ICP labels into their own calculated fields.
- Let a company take the most advanced stage of any of its contacts.
What goes wrong
- Moving a stage backwards to mean "disqualified". Use lead status for that.
02
Track the numbers that move before revenue
Also called Leading-indicator dashboards
Revenue arrives months after the work. Track what moves first: new leads nobody has touched, how long priority accounts wait, how leads progress, how many top accounts have fresh signals and how many people you know at each.
Why it matters. By the time revenue drops, the cause is a quarter old. Early numbers let you fix it this week.
How to do it
- Rep view: untouched leads trending to zero, oldest priority accounts first.
- Manager view: progress between statuses, accounts with only one contact.
- Leadership view: coverage of top accounts and signal activity.
- Write one line under every chart on how to read it.
What goes wrong
- Dashboards nobody owns. Every chart needs an owner and a decision it informs.
03
See which channel every deal came from
Also called Channel attribution: first touch, latest touch and pipeline by source
Record the source of every lead when it first arrives and keep it on the contact, the company and the deal. Then compare outbound, paid ads, events, organic search, AI search and referrals on qualified pipeline, revenue and full cost, not on clicks or form fills.
Why it matters. A channel can look cheap per lead and be expensive per customer. Ad platforms count clicks; only the CRM can show which channel produced revenue.
How to do it
- Agree one short list of channels everyone uses: outbound, paid search, paid social, events, organic search, AI search, referral, partner.
- Capture the source on every form with hidden fields for campaign tags (UTMs) and the ad click ID, and set it for outbound and event leads at import.
- Keep two fields: first touch (what created the lead) and latest touch (what brought it back). Copy both onto the deal.
- Report qualified pipeline, won revenue and full cost per channel for the same period, including media, tools, agency and creative.
- Cost per qualified lead = total channel cost ÷ qualified leads. Compare channels on this and on pipeline, never on clicks.
- Review monthly and move budget from channels that fill the funnel to channels that fill the pipeline.
What goes wrong
- An empty source field. If most deals say "unknown", fix capture before reading any report.
- Counting newsletter sign-ups next to demo requests as if they were equal leads.
- Comparing a one-month campaign with an all-time CRM count, or adding costs in different currencies.
- Giving one channel all the credit. B2B buyers touch several channels, so read first and latest touch side by side.
04
Find the ad spend that produced nothing
Also called Google Ads wasted spend audit and negative keywords
List every search term that cost money and brought no conversions, group them by theme and put a price on each theme. Then add negative keywords with a weekly 15-minute review, with a person approving every change.
Why it matters. Search ads often show for words that sound like your product but attract students, job seekers or developers.
How to do it
- Pull search terms for the last 30 to 90 days.
- Flag terms with spend and zero conversions, sorted by cost.
- Group them by theme: jobs, free, how-to, unrelated software, competitors.
- Keep a "do not block" list of buying words such as pricing, cost and alternatives.
- Review weekly and add negatives only with approval.
What goes wrong
- Blocking "pricing" or "cost" because they look cheap.
- Adding negatives from less than two weeks of data.
05
Measure whether AI assistants recommend you
Also called AI visibility and generative engine optimisation (GEO)
Ask ChatGPT, Claude, Gemini and Perplexity the questions your buyers ask, many times each, and score every answer: not mentioned, mentioned in passing, listed or top pick. Track your mention rate and share of voice against competitors over time.
Why it matters. More buyers ask AI assistants for recommendations before they visit any website. If you are not in the answer, you are not on the shortlist.
How to do it
- Collect real buyer questions from sales calls and your customer profile.
- Tag each question by stage: learning, comparing, deciding.
- Run them on several assistants with web search on, several times each.
- Score every answer with one rubric and check facts about you are correct.
- Report mention rate and share of voice for questions that do not name you.
What goes wrong
- Trusting one run. Answers change every time.
- Counting a mention that describes a different company.
SparkToro (January 2026) found under a 1 in 100 chance of getting the same brand list twice from AI tools for the same question. [independent research]
Run it with the AI visibility check agent
06
Find out who AI recommends instead of you
Also called AI citation and lost-mention analysis
For every question where AI named a competitor and not you, collect the sources it cited and sort them: competitor pages, review sites, "best of" lists, articles. The pattern becomes your content and PR plan.
Why it matters. AI assistants repeat what the web says. The sources they cite show exactly where you need to be mentioned.
How to do it
- Separate who got mentioned from which pages were cited.
- Rank the cited domains by how many questions they appear in.
- Sort pages by type: lists, comparisons, reviews, definitions.
- Plan: get listed, publish the missing comparison, fix pages that are cited but broken.
What goes wrong
- Only looking at links. Some assistants mention sources in text without a link.
07
Let results tune your system
Also called Signal performance loop
Your scoring is a hypothesis. Tag every meeting and deal with the signal that started it, log results after each campaign and adjust the weights every quarter, so the system gets smarter each cycle.
Why it matters. Without a feedback loop, scores and targeting drift away from what actually converts.
How to do it
- Tag each meeting and deal with its starting signal and campaign.
- After each campaign, log replies and meetings by signal.
- Each quarter, raise the weight of signals that convert and drop the ones that do not.
- Keep a dated log of every change and why.
What goes wrong
- Changing five things at once, so you cannot tell what helped.
Questions people ask
What is the difference between lifecycle stage and lead status?
Lifecycle stage shows where a contact is in the funnel, such as lead or customer, and only moves forward. Lead status shows what the sales rep is doing with them, such as attempting contact or qualified.
How do I calculate cost per qualified lead?
Add all costs for the period, including media, tools, agency and creative, and divide by the number of qualified leads in your CRM for the same period.
How do I know if ChatGPT recommends my company?
Ask it the questions your buyers ask, several times each, with web search on, and record whether you are mentioned. Compare the same questions across other assistants and over time.