Audit and redesign a lead-routing setup — Revenue Operations Lead take-home

Revenue Operations Lead45 minutes

Context

Fernwood Cloud is an invented B2B software company with 90 employees, 6 SDRs, 8 account executives (4 mid-market, 3 enterprise, 1 strategic), and one customer success team. Leads come from web forms, a webinar platform, a product trial registration, and uploaded purchased lists. Sales leadership says leads are "getting lost" and AEs complain about junk. You have been asked to diagnose the routing and propose a replacement. Nothing is changed in any live system.

Materials

Current rules, in the order they run: (1) If country is not US or Canada, assign to the next SDR in round robin. (2) If employees is 1000 or more, assign to the enterprise AE queue. (3) If the email domain matches an existing customer, assign to the account owner. (4) If lead source is "webinar", assign to SDR round robin. (5) Everything else goes to the mid-market AE queue. Stated SLA: "respond fast." Last month, 1,240 leads: 410 web form, 380 webinar, 290 trial, 160 list upload. Median time to first touch: 31 hours. 22 percent of leads untouched after 7 days. A 30-lead sample of last month's routed leads is attached as routed_leads.csv (columns: lead_id, source, country, employees, email_domain, existing_customer, company, account_type, assigned_to, rule_fired, hours_to_first_touch; "none" means untouched after 7 days). Misroute examples the team called out: (a, lead L-1007) a 3,000-employee customer's new employee starts a trial and goes to an SDR in Germany. (b, L-1008) A webinar attendee at a 1,200-person company goes to the enterprise queue before the webinar rule runs, and the AE ignores it as "just a webinar lead". (c, L-1009) A list-upload lead at a competitor is assigned to a mid-market AE. (d, L-1010 and L-1011) Two leads from the same company arrive an hour apart and go to two different SDRs. (e, the Enterprise-AE-queue rows) The enterprise queue has one AE on leave and nobody is covering it. The team has not reviewed the other rows, and the sample holds more misroutes than the five they listed. Capacity: SDRs average 60 leads per week each; AEs want no more than 25 new leads per week.

Data

Download materials.csv (30 rows, 11 columns)

Columns: lead_id, source, country, employees, email_domain, existing_customer, company, account_type, assigned_to, rule_fired, hours_to_first_touch.

The table scrolls sideways, so every column is there.

lead_id	source	country	employees	email_domain	existing_customer	company	account_type	assigned_to	rule_fired	hours_to_first_touch
L-1001	web-form	US	120	brightdock.example	no	Brightdock	prospect	MidMarket-AE-queue	5	4
L-1002	web-form	US	450	cobaltmill.example	no	Cobalt Mill	prospect	MidMarket-AE-queue	5	22
L-1003	trial	US	85	pinewave.example	no	Pinewave	prospect	MidMarket-AE-queue	5	3
L-1004	webinar	US	300	trellisworks.example	no	Trellis Works	prospect	SDR-1	4	70
L-1005	list-upload	US	60	quarryhq.example	no	Quarry	prospect	MidMarket-AE-queue	5	none
L-1006	trial	CA	210	maplegrid.example	no	Maplegrid	prospect	MidMarket-AE-queue	5	9
L-1007	trial	DE	3000	ostwindgroup.example	yes	Ostwind Group	customer	SDR-2	1	41
L-1008	webinar	US	1200	granitelane.example	no	Granite Lane	prospect	Enterprise-AE-queue	2	none
L-1009	list-upload	US	800	rivalstack.example	no	Rivalstack	competitor	MidMarket-AE-queue	5	18
L-1010	web-form	DE	180	kellerbau.example	no	Keller Bau	prospect	SDR-3	1	52
L-1011	web-form	DE	180	kellerbau.example	no	Keller Bau	prospect	SDR-4	1	none
L-1012	web-form	US	2400	tidalmedical.example	no	Tidal Medical	prospect	Enterprise-AE-queue	2	none
L-1013	trial	US	35	sparkloop.example	no	Sparkloop	prospect	MidMarket-AE-queue	5	6
L-1014	webinar	US	90	gmail.example	no	(personal email)	unknown	SDR-5	4	none
L-1015	web-form	GB	600	harrowtech.example	yes	Harrow Tech	customer	SDR-6	1	27
L-1016	list-upload	US	150	opalfreight.example	no	Opal Freight	prospect	MidMarket-AE-queue	5	none
L-1017	web-form	US	1500	summitcare.example	yes	Summit Care	customer	Enterprise-AE-queue	2	2
L-1018	trial	US	500	lumenroast.example	no	Lumen Roast	prospect	MidMarket-AE-queue	5	14
L-1019	webinar	US	2000	bayfieldbank.example	no	Bayfield Bank	prospect	Enterprise-AE-queue	2	none
L-1020	list-upload	US	40	tinyforge.example	no	Tinyforge	prospect	MidMarket-AE-queue	5	none
L-1021	web-form	US	320	redfinchlabs.example	no	Redfinch Labs	prospect	MidMarket-AE-queue	5	8
L-1022	trial	US	320	redfinchlabs.example	no	Redfinch Labs	prospect	MidMarket-AE-queue	5	31
L-1023	web-form	MX	270	solmarfoods.example	no	Solmar Foods	prospect	SDR-1	1	60
L-1024	webinar	US	75	nimbusdesk.example	no	Nimbus Desk	prospect	SDR-2	4	29
L-1025	list-upload	US	1100	rivalstack.example	no	Rivalstack	competitor	Enterprise-AE-queue	2	none
L-1026	web-form	US	700	cedarpoint.example	yes	Cedar Point Co	customer	AE-Luis	3	5
L-1027	trial	US	25	gemstonehq.example	no	Gemstone	prospect	MidMarket-AE-queue	5	none
L-1028	web-form	CA	4500	northernlight.ca	no	Northern Light	prospect	Enterprise-AE-queue	2	44
L-1029	webinar	US	2300	ironbridgeco.example	no	Ironbridge	prospect	Enterprise-AE-queue	2	none
L-1030	trial	US	140	foxglovebio.example	no	Foxglove Bio	prospect	MidMarket-AE-queue	5	12

Showing the first 25 of 30 rows. Download the CSV for all rows.

Task

Diagnose first, then design. Use a doc, a table, or a diagram, and explain your decisions as you go.

Walk through each of the five misroute examples and name the rule or gap that caused it, then scan the remaining rows and list any further misroutes you find.

Group the causes into a short list of root problems.

Propose a routing design: the order of rules, what each segment gets (sales-ready, nurture, disqualify, customer, competitor), who owns it, and the SLA for each with a reason.

Decide what happens when the owner is absent or over capacity.

Choose three metrics that would tell you in a month whether it works, and what you would do if each moved the wrong way.

Say what information you would still need to verify before shipping, and how you would roll out without breaking live routing.

Stay inside 45 minutes.

Deliverables

A diagnosis table: each misroute (the five listed plus any you found in the sample), the cause, and the root problem it belongs to
A routing spec: rule order, segments, owner, SLA, and the reason for each SLA
An absence and capacity policy for owners
Three success metrics with a stated action if each moves the wrong way
A list of facts to verify before shipping and a rollout plan
A one-paragraph summary a sales leader could read in a minute

Rubric

Dimension	What good looks like	Evidence to look for	Levels, weak to excellent
Routing decisions	Defines segments, ordered rules, owners, SLAs with reasons, and a disqualify and competitor path; handles ties and absence.	A spec with rule order justified; SLAs that differ by segment; fallback owner; duplicate-company handling.	Weak One flat list of rules with no segments, owners, or SLAs. Adequate Segments named but one SLA for all; no handling of ties, absence, or duplicates. Good Segments, ordered rules, owners, and an SLA per segment; reasons thin; absence handled loosely. Very good Justified rule order, SLAs that differ by segment, a fallback owner, and same-company grouping. Excellent Complete spec with disqualify, competitor, and customer paths, explicit tie-breaking, capacity-aware SLAs, and an exceptions queue.
Systems thinking	Finds root causes behind the five examples instead of patching each one, and identifies where a human reviews exceptions.	Five examples collapsed into a few root problems; an exceptions queue; capacity limits used in the design.	Weak Patches each example separately; no root causes. Adequate Groups some examples but misses that rule order and missing account context drive most of them. Good Names two or three root problems; exceptions handling mentioned briefly. Very good Collapses the examples into a few root problems and designs an exceptions queue sized to capacity. Excellent Finds root causes across the sample, ties each to a design decision, and predicts which metric will show the fix.
Data hygiene	Notices what the rules cannot know without clean data (account match, company size, customer status) and what must be verified.	Named dependencies on account matching; a check on the company-size source; no assumption stated as fact.	Weak Treats the data as clean and the rules as sufficient. Adequate Notices a data gap but does not say what to verify. Good Names dependencies such as account matching and company size, without checks. Very good Names the dependencies, proposes checks on the size source and account match, and avoids guesses stated as fact. Excellent Quantifies the data gaps from the sample, proposes specific verification queries, and sequences rollout around them.
Clear write-up	A leader can read the summary and the spec and know what changes and why.	Short summary at the top; consistent terms; trade-offs stated.	Weak Summary missing or unreadable by a sales leader. Adequate Summary restates the problem without the change. Good Summary says what changes; trade-offs missing. Very good A sales leader can read the summary and spec and know what changes, why, and what it costs. Excellent One-minute summary with decision, trade-offs, and risks; terms consistent through the spec.
Working with AI	Uses an assistant to stress-test the spec against the examples, and verifies its suggestions.	Examples replayed against the new rules; a correction when the assistant produced a conflict.	Weak Pastes assistant suggestions without replaying any lead. Adequate Replays one example; accepts conflicts between rules. Good Replays the five listed examples and fixes the obvious conflicts. Very good Replays the five examples and several others against the new rules and corrects an assistant error. Excellent Builds a repeatable replay of the whole sample, catches rule conflicts, and documents what changed per iteration.

Why unaided AI alone does not pass this

The rules, volumes, and a 30-row routed sample are provided, so the work is judgment on that data: tracing specific rule-order failures (L-1008 is a rule-order bug, L-1010 and L-1011 are a company-level grouping gap) and spotting misroutes the team did not list, such as a personal-email webinar lead, a customer on a UK domain sent to an SDR, and enterprise leads sitting untouched. SLAs must be sized against the stated capacity of 60 leads per SDR and 25 per AE per week. A reviewer can verify the work by replaying the examples against the new design and by asking the candidate to walk through two decisions, such as how they ordered the rules and why.

Notes for the reviewer

Unlisted misroutes planted in the sample: L-1015 (a customer on a UK domain goes to an SDR through the country rule), L-1014 (a personal-email webinar lead), L-1021 and L-1022 (same company, a form and a trial, split across queues), L-1025 (a second competitor lead), and L-1028 (a 4,500-person Canadian company in the enterprise queue, untouched for days, like the other Enterprise-AE-queue rows). Check that the candidate replays the examples against the new design. Strong answers separate lead-to-account matching from routing. To verify, ask the candidate to walk through two decisions, such as how they ordered the rules and why. Not observed is informative, not punitive.

Adapt this in ChatGPT