Audit the stack and the organization

A growth plan is only worth what the team can ship, so this part of the audit measures the machine that will carry it out. You are looking for four things: tools that answer the questions you ask, automation someone actually watches, a named owner for every metric, and a pace you can put a number on. The result sets the capacity that everything in the prioritization has to fit into.

Throughout this chapter, say a B2B SaaS at 2M ARR has 25 people, including a growth team of three and two developers shared with product. The numbers are invented to show the method.

Inventory the stack from the invoices

Start from an accounting export of software spend, because it catches the tools nobody thinks of. Add the admin list of your single sign-on if you have one, then fill one row per tool.

Tool category	Job it does	Owner	Active users	Data in and out	Verdict
CRM	Source of truth for deals	Head of sales	9 of 14 seats	In: forms. Out: email tool	Keep
Product analytics	Behavior after signup	Product	4 of 20 seats	In: app events	Keep, rescope
Enrichment tool A	Company data on leads	Nobody	0 in 3 months	None	Cancel
Form tool	Landing page forms	Marketing	2 of 2	Out: CRM, manual export	Fix the sync

Each business object needs one source of truth: the account, the deal, the invoice, the user. When two tools claim the same object, write down which one wins and who fixes the conflict.

Check that each question has a tool that answers it

Listing tools tells you what you pay for. Coverage tells you whether you can run the business. A product analytics tool is built to follow people after signup, so asking it where a qualified lead came from will give you a confident answer of poor quality.

Question	Tool that should answer	How to test it
Which channel brought this qualified lead?	Acquisition analytics joined to the CRM	Compare 20 recent leads with their real history
Which actions predict retention?	Product analytics	Run it for one cohort, by account
What does this account pay and when did it change?	Billing tool	Reconcile the total with accounting
Why did we lose this deal?	CRM	Share of lost deals with a usable reason
Who is at risk this month?	Product data pushed to customer success	Ask the team where they see it

Then run the five-customer trace. Take five recent customers and follow each one by hand, from first visit to first invoice, across every tool. Note where the identifier changes, where a value is typed manually and where the trail stops. This takes an afternoon, and it finds more than an architecture diagram. Anything broken goes to the data reliability work.

Audit automation, and who watches it

List every recurring task in growth, sales ops and customer success, and every running automation, with the time it takes and what happens when it breaks.

Task	Hours per month	Today	Failure visible to	Verdict
Sync form leads to CRM	0	Automation, no alert	Nobody until sales complains	Add an alert and an owner
Weekly reporting	12	Manual, three exports	The person doing it	Automate
Lead qualification	20	Manual	Sales	Automate the first pass, human review on the rest
Reply to a strategic customer	6	Manual	Customer	Keep human

Sort each line into automate, assist with AI, keep human or stop doing it. Stopping is the most underused option. The rule for AI work is the one that matters on Derrick: agents take much of the recurring work under written rules, and people review what matters. An automation without an owner or a failure alert is a future incident with a delay on it.

Look at publishing as well. If a new landing page requires a developer ticket, the marketing team’s pace is set by the product backlog. On Derrick, moving the site to code on Astro and Cloudflare Pages made publishing and changing thousands of pages a matter of code, and that kind of decision shows up directly in your pace figures.

Map ownership and incentives

Get the organization chart, the current objectives and the bonus criteria. Then fill one row per funnel stage.

Metric	Single owner	Can act alone on	Depends on	Last action shipped
Qualified leads	Head of marketing	Content, ads	Sales definition of qualified	2 weeks ago
Activation rate	No one		Product, customer success	Unknown
Net revenue retention	Customer success	Outreach	Product, billing	1 month ago

Empty cells are the finding. Activation very often sits between marketing, which stops at signup, and sales or customer success, which start at the deal. Compare the objectives too: a marketing team paid on lead volume and a sales team paid on revenue will argue over lead quality forever, and each will be right by their own numbers.

An audit can say that activation has no owner, or that a skill is missing. Whom to hire, move or let go depends on context you don’t have from the outside, so I leave that decision to the founders.

Put numbers on the pace of execution

Export tasks from the tool the team uses, keeping creation date, ship date and any written hypothesis or result. If you have a table growth_tasks:

SELECT
  DATE_TRUNC('month', shipped_at) AS month,
  COUNT(*) AS shipped,
  PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY shipped_at - created_at) AS median_lead_time,
  COUNT(*) FILTER (WHERE hypothesis IS NOT NULL AND result IS NOT NULL) AS with_learning
FROM growth_tasks
WHERE shipped_at IS NOT NULL
GROUP BY 1
ORDER BY 1;

Say the query shows 6 actions shipped a month, 34 days of median lead time and only 1 action with a written result. The finding is that the team ships but doesn’t learn, and that anything needing a developer waits in line. Judge the figures against your own history, then compute real capacity: people × days available per month × the share of time free of recurring work. With three people, 20 days a month and 25% free, the team has 15 person-days a month.

Then check the rituals: a weekly growth meeting with a decision at the end, a single backlog, a monthly review of results. A ritual that produces no decision can go.

Score the machine and deliver

Dimension	1 looks like	5 looks like
Stack coverage	Key questions answered by export and guesswork	Every question has a tool and a tested answer
Data flow	Manual copies, identifiers lost along the way	One account traceable from first touch to invoice
Automation	Silent failures	Owner and alert on each flow
Ownership	Metrics without an owner	One name per metric, objectives aligned
Pace	No measure of what ships	Lead time known, each action leaves a result

Write each problem as problem, consequence, action: “Activation has no owner, so no action on it has shipped in six months. Name one person and give that person a fixed share of a developer’s time.”

Your deliverable holds the stack table with its verdicts, the coverage grid, the ownership matrix, the pace baseline, the capacity figure, and the short list of fixes. The capacity figure goes straight into the 6-month roadmap.

Your next step: pull last year’s software spend from accounting, fill the verdict column, then pick five recent customers and start the trace.