A materials supplier's ordering portal shows plenty of project teams starting an order and far fewer finishing one. A general contractor sees inquiries pile up while only a handful turn into awarded work. In both cases the overall conversion rate says something is wrong, and it says nothing about which step is failing.
Conversion funnel analysis maps each step a buyer takes toward a goal, such as a completed materials order or an awarded project, and measures how many people move from one step to the next. Done well, it shows where buyers drop off, which groups drop off most and which fix deserves a test first.
The five steps below follow one running example, a materials supplier's online checkout, with illustrative numbers. The later sections show how an AI agent automates the analysis and where Datagrid fits once the bottleneck sits in pursuit qualification instead of a web form.
How to Run a Conversion Funnel Analysis Step by Step
Funnel conversion analysis works best as a fixed sequence, because each step depends on definitions set in the step before it. The checkout figures illustrate the method, so benchmark against your own data.
Step 1. Define the Funnel Stages
Map the stages a buyer actually passes through, and tie each one to a recorded event with a clear entry and exit. For the checkout example, the stages are checkout start, shipping information, payment information and purchase. Decide whether buyers must enter at the first stage or can join later. Google Analytics calls these closed and open funnels, and the choice changes every rate downstream.
A pursuit funnel follows the same logic with different events: inquiry received, qualified, bid submitted and project awarded. Teams that scale business development agree on those definitions before anyone reports on them. A common mistake at this step is treating a page view as proof that an order or inquiry happened.
Step 2. Measure Drop-Off Between Stages
Calculate every adjacent transition, then the overall rate:
Stage conversion: Completions at the next stage divided by completions at the current stage.
Drop-off: One minus stage conversion, plus the absolute number of buyers lost.
Overall conversion: Purchases divided by checkout starts, tracked alongside the time between stages.
Suppose 200 project-team buyers start checkout and 140 submit shipping information. That is 70% progression and 60 buyers, or 30%, lost at the first transition, the same arithmetic Google uses in its checkout journey report. Count distinct buyers at every stage, because mixing sessions with buyers makes the rates meaningless.
Step 3. Segment the Drop-Off
Break a leaking transition down before changing the whole checkout. Compare traffic source, device, campaign, new versus repeat accounts and buyer type, such as general contractor, subcontractor or owner, wherever those fields exist. Show entrants, completions and losses for each segment.
In the example, split the 200 starts and 140 shipping completions by device and account type to see where the 60 losses concentrate. Check segment size first. A segment with 12 entrants can swing wildly without telling you anything.
Step 4. Diagnose the Cause
Pair the counts with evidence of what buyers experienced at that step: consented session recordings, heatmaps, form-field errors, short surveys and support calls. For the shipping step, look at whether delivery costs appear late, whether a required job number field rejects valid entries and whether buyers can see a delivery window for their site. Project teams that track material orders against a schedule may abandon an order that cannot confirm a date.
Write the finding as a testable explanation, such as "unclear delivery costs cause shipping-step exits on mobile." A pattern in the data is a lead to investigate, and it only becomes a cause once a test supports it.
Step 5. Test the Fix
Turn the diagnosis into a hypothesis and rank it by affected volume, likely impact and implementation effort. Run an A/B test where traffic supports one, with a set allocation and evaluation plan. Low-volume built-world funnels may need comparable before-and-after periods, which deserve more caution than a controlled test.
Measure shipping-step completion against the 70% baseline, then check purchases and overall conversion. A higher intermediate rate is a loss if final orders decline.
Common Time Sinks in Built-World Funnel Analysis
Manual analysis slows down in predictable places, and each delay pushes the fix further out:
Too many channels: Tracking every interaction buries the transition you need to fix. Start with the events the goal requires.
Reconstructing buying paths: Rebuilding journeys by hand takes hours, so rank paths by entrant count and outcome before tracing any of them.
Inconsistent records: Stage names, identifiers and required fields differ across systems, and cleaning them eats the time meant for analysis.
Favorite explanations: Teams fix what they believe is broken. Document competing explanations and let a test decide.
Pursuit funnels add their own delay. Bid teams answering requests for proposals (RFPs) who want to reduce RFP response time can often trace much of the delay to locating project information.
How an AI Agent for Funnel Analysis and Optimization Works
An AI agent for funnel analysis and optimization runs the five steps continuously instead of once a quarter. Configure it to handle these tasks against your own definitions:
Monitoring: Tracks stage conversion against its baseline on a cadence that matches how often the data refreshes.
Anomaly detection: Flags an abnormal drop with the absolute loss and entrant counts, so reviewers can judge volume.
Automatic segmentation: Breaks each change down by source, device, campaign or account type to find where it concentrates.
Likely causes: Attaches form errors, feedback and support notes from around the affected step.
Test hypotheses: Suggests a hypothesis and the stage metric it should move, ranked by volume, impact and effort.
Reporting: Shows results for both the target stage and the final goal, with a named reviewer for each flagged issue.
Datagrid connects to Google Analytics, HubSpot, Salesforce and Google Sheets. Keep funnel calculations in your analytics system and validate each agent task against known results before trusting it. Event definitions from Step 1 still come from your team, and a stale data source produces a confident but wrong alert, so check freshness before acting.
Datagrid for Built-World Pursuit Funnels
When funnel analysis shows that pursuits stall on qualification or missing project information, Datagrid's AI agents can execute the supporting work:
Prequalification responses: The Pre-Qualification Agent can review uploaded checklists and supporting project files to complete responses. Track unanswered questions and turnaround, and review every answer before submission.
Bid-package requirements: The Deep Search Agent can search specs, drawings, requests for information (RFIs) and submittals so a go/no-go decision rests on the actual requirement. It feeds the bid management work that follows.
Owner terms: During construction contract review, the Contract Review Agent can flag compliance gaps and conflicts before commitment. Responsible reviewers still decide what each finding means.
Win and loss records: Custom AI agents can capture go/no-go reasoning and owner feedback with each opportunity, which gives the RFP strategies your team uses a better record to learn from.
These agents depend on clean access to the right project files. Setup takes longer when documents sit across several systems with different permissions.
Get Started With Datagrid on One Funnel Bottleneck
Datagrid's AI agents, including custom agents built from a prompt, can keep funnel analysis running between reviews and handle the project-information work that stalls pursuits:
Stage monitoring: Agents can track stage conversion in connected analytics and customer relationship management (CRM) data and flag abnormal drops by segment.
Cause evidence: Agents can attach form errors, feedback and notes to each flagged drop.
Test planning: Agents can suggest hypotheses and rank them by affected volume and effort.
Prequalification responses: The Pre-Qualification Agent can complete checklist responses from your project files.
Requirement search: The Deep Search Agent can answer pursuit questions from specs, drawings and RFIs.
Your team still sets the stage definitions, approves each test and decides which fix ships. Get started with Datagrid by connecting one funnel and checking which stage loses the most buyers this month.
Frequently Asked Questions About Conversion Funnel Analysis
What Is Conversion Funnel Analysis?
Conversion funnel analysis maps the steps buyers take toward a goal and measures how many move from each step to the next. It shows where people drop off so you can investigate the cause and test a fix.
What Is a Good Conversion Rate for a Sales Funnel?
A good rate is one that beats your own baseline for the same counting unit and time window. Compare materials orders against past checkout starts, and compare awarded projects within the same project and owner types. The 70% figure in this guide is an illustration of one transition only.
What Are Lower Funnel Metrics?
Lower funnel metrics track buyers close to the goal. For a materials checkout, they include shipping-step completion, payment completion, purchases, overall conversion and the time between those stages.
