Monitor and optimise automations

Building automation isn't the end, it's the beginning. Automations need ongoing optimisation just like campaigns and pages. The systematic approach: track performance metrics, identify failures and bottlenecks, test improvements, and connect automations into complete systems that compound effectiveness.

This chapter shows you how to monitor automation performance systematically, identify and fix failures quickly, A/B test workflows to improve conversion rates, and compound automations into complete growth engines.

Don't just track 'workflow ran'. Track outcomes and efficiency.

Throughput metrics: Records processed per day, week, month. Trends show if volume is growing and whether automation is keeping up. Example: lead assignment workflow processed 180 leads this week versus 200 last week. Why the drop? (Could be normal seasonality, could be form broken, could be workflow failing silently.)

Success rate: Percentage of records that complete successfully versus fail or error. Target: >95% success rate for stable workflows. If dropping below 95%, investigate.

Example: Welcome email workflow runs 200 times, 190 succeed, 10 fail with 'email address invalid'. 95% success rate, acceptable. But if 40 fail (80% success), something's wrong (maybe email validation on form is broken).

Execution time: How long workflow takes to complete. Matters for time-sensitive automations. Example: lead assignment should complete within 5 minutes. If taking 2 hours, reps aren't following up fast enough.

Goal completion rate: Did the automation achieve its goal? Lead assignment: what percentage of assigned leads get followed up within 2 hours? Welcome sequence: what percentage of nurtured leads become MQLs? These outcome metrics matter more than process metrics.

Create an automation dashboard: list of all workflows, records processed this month, success rate, average execution time, goal completion rate (where applicable), last failure date, last maintenance date. Review monthly. Flag any workflow with <90% success rate or execution time >2× normal.

When automation fails, diagnose root cause and fix it permanently, don't just patch the symptom.

Common failure types: Data issues (missing required field, invalid format), logic errors (condition doesn't handle edge case), integration failures (API down, rate limit hit, authentication expired), timing issues (workflow triggered too early before data ready), volume overwhelm (too many records processed simultaneously).

For each failure, ask: why did this happen? (immediate cause) → why did the immediate cause happen? (root cause) → how do we prevent it permanently? (solution).

Example failure: Lead assignment workflow fails with 'cannot read property company_size'. Why? Company size field is empty for this lead. Why is it empty? Form doesn't require it (allows submission with blank). How do we prevent? Option 1: Make field required on form. Option 2: Add condition in workflow checking if company size exists, if not assign to default queue. Choose option 2 (better UX to not block form submission).

Failure patterns: After 10-20 failures, patterns emerge. 60% of lead assignment failures are missing company size. 30% are invalid industry values. 10% are miscellaneous. Fix the top 2 causes (90% of failures) first.

Document fixes in workflow notes: 'Added empty company size handling 2024-03. Previously failed 12 times/month, now fails <1 time/month.' This helps when troubleshooting future issues and shows improvement over time.

Preventive maintenance: Don't wait for failures. Quarterly, review high-volume workflows. Check if logic still correct. Update for changed business rules. Add validation for new edge cases. Test with current sample data (what worked 6 months ago might not work now if forms or processes changed).

Just like ad creative and landing pages, workflows can be tested and optimised.

Test timing: Send welcome email immediately versus wait 1 hour versus wait 1 day. Which gets better open rates and conversion to next stage? Split traffic 33/33/33, run for 4 weeks, measure. Example result: immediate send gets 45% open rate, 8% convert to demo. 1-hour delay gets 52% open rate, 12% convert to demo (winner). 1-day delay gets 38% open, 6% convert. Implement 1-hour delay for all.

Test content: Welcome email variant A (focuses on product features) versus variant B (focuses on customer success). Split traffic 50/50 for 2 weeks. Measure reply rate, demo booking rate, progression to MQL. Implement winner.

Test logic: Lead assignment rule A (assign based on company size only) versus rule B (assign based on company size + industry). Measure: time to first contact, opportunity conversion rate, win rate. If rule B leads convert 30% better despite taking slightly longer to assign, it's the winner.

Test sequence length: Nurture sequence with 3 emails versus 5 emails versus 7 emails. More touches might improve conversion, or might annoy people. Test. Measure: MQL rate, unsubscribe rate, negative replies. Find optimal length.

Use same discipline as landing page testing: change one variable at a time, run for sufficient duration (minimum 2-4 weeks for email workflows), reach statistical significance (95% confidence), segment results if workflows serve multiple audiences.

Track tests in experiment log (same log as landing page tests). Document hypothesis, test design, results, learnings. Apply winning patterns across similar workflows.

Individual workflows create value. Connected workflows create compound value.

Example compound system for lead management: Workflow 1: Form submission triggers lead creation and assignment (assigns to right rep based on criteria). Workflow 2: Assignment triggers welcome email sequence (personalised based on segment). Workflow 3: Email engagement triggers scoring (opened email +5, clicked link +10). Workflow 4: Score threshold triggers MQL status change. Workflow 5: MQL status change triggers sales notification and task creation. Workflow 6: Task creation triggers follow-up reminder sequence (if not completed in 2 hours, remind, if not completed in 4 hours, escalate).

Six workflows working together create complete lead management system. Each workflow does one thing well. Together they create seamless experience from form to follow-up.

Designing compound systems: Start with desired outcome (lead converts to opportunity within 48 hours). Work backwards: what needs to happen for that outcome? (rep needs to follow up quickly) → what triggers rep follow-up? (task creation and reminders) → what triggers task creation? (MQL qualification) → what triggers MQL? (engagement scoring) → what triggers engagement? (email opens and clicks) → what triggers email sequence? (lead assignment) → what triggers assignment? (form submission).

Now you have the workflow chain. Build each piece. Test individually. Then test the complete system end-to-end.

Optimising compound systems: Bottleneck analysis. Where do records get stuck? If 100 leads enter system but only 20 become MQLs, where are the other 80 dropping out? (Maybe: 30 never open welcome email - email problem. 40 open but don't engage - content problem. 10 engage but scoring threshold too high - logic problem.) Fix bottlenecks sequentially, biggest first.

Compound improvement: Improving workflow 3 (email engagement) by 10% improves workflow 4, 5, and 6 outcomes by 10% because they depend on workflow 3. This is compound effect. Small improvement in early workflow amplifies through the system.

Track automation performance metrics: throughput (records processed), success rate (>95% target), execution time (should be consistent), goal completion rate (outcome metrics). Build dashboard showing health of all workflows.

Identify and fix failures systematically. Ask why failures happen, find root cause, prevent permanently. Look for failure patterns (80% of failures from 20% of causes). Document fixes. Perform quarterly preventive maintenance.

A/B test workflows like landing pages. Test timing, content, logic, sequence length. Change one variable at a time, run for sufficient duration, reach significance, implement winners. Track tests in experiment log.

Compound automations into complete systems. Connect workflows so output of one triggers next. Design systems by working backwards from desired outcome. Optimise bottlenecks where records get stuck. Small improvements compound through dependent workflows.

With automation systematic, you've completed the demand generation playbook series. Apply these frameworks to build compound growth across your four growth engines.