AdvancedLesson 23 of 305 min read

Marketing Automation, CRM and Lead Scoring

Use a CRM and simple points to rank leads, set the next action for each band, and check the scores against what really happened, with a worked example of twelve leads.

In this lesson you will

  • Explain CRM and lead scoring
  • Build a simple points table
  • Sort leads into bands with actions
  • Test scores against outcomes

About 25 minutes

Who should Noor call first?

Noor now sells gift sets to small businesses, such as a hairdresser who wants candles for customers at Christmas. Twelve of them have left their details, opened emails or asked for a sample. She has a few hours a week to follow up, so she must choose. Marketing automation tools and CRM systems, which store contact details and records of every interaction, help with this. A common tool is lead scoring: give points for actions that suggest interest, take points away for signs of lost interest, and follow up first with the highest scores.

How scoring works

  • Points for interest. Opening an email is a small sign. Clicking is stronger. Viewing the pricing page is stronger still, and asking for a sample is the strongest.
  • Points off for lost interest. Unsubscribing, or having been quiet for a month.
  • Bands. Add the points, and sort leads into hot, warm and cold, with a different action for each.
  • Review. Compare the bands with what really happened, and change the points.

A score is a guess until you test it. The program below holds a made-up points table and twelve leads with the actions each took and whether they later bought. It scores every lead, puts each in a band, and then checks how many in each band bought. That check, comparing the prediction with the outcome, is the step most people skip.

POINTS = {"opened": 1, "clicked": 3, "viewed pricing": 5, "asked for sample": 10, "unsubscribed": -20, "quiet 30 days": -5}
# MADE-UP leads: actions, and whether they later bought
leads = [(["opened", "clicked", "viewed pricing"], True), (["opened"], False), (["opened", "clicked"], False),
         (["asked for sample", "clicked"], True), (["quiet 30 days"], False), (["opened", "clicked", "viewed pricing", "asked for sample"], True),
         (["opened", "unsubscribed"], False), (["viewed pricing"], True), (["opened", "clicked", "clicked"], False),
         (["viewed pricing", "clicked"], True), (["opened", "quiet 30 days"], False), (["clicked", "asked for sample"], False)]
def score(actions):
    return sum(POINTS[a] for a in actions)
def band(s):
    return "hot" if s >= 8 else "warm" if s >= 3 else "cold"
stats = {}
for actions, bought in leads:
    b = band(score(actions))
    t = stats.setdefault(b, [0, 0])
    t[0] += 1; t[1] += bought
for b in ("hot", "warm", "cold"):
    n, k = stats.get(b, (0, 0))
    print(f"{b}: {n} leads, {k} bought")
print("scores:", [score(a) for a, _ in leads])
print("hot buy rate:", round(100 * stats["hot"][1] / stats["hot"][0]), "%")
print("cold buy rate:", round(100 * stats["cold"][1] / stats["cold"][0]), "%")
print("lead who bought with cold score:", sum(1 for a, bgt in leads if bgt and band(score(a)) == "cold"))
hot: 5 leads, 4 bought
warm: 3 leads, 1 bought
cold: 4 leads, 0 bought
scores: [9, 1, 4, 13, -5, 19, -19, 5, 7, 8, -4, 13]
hot buy rate: 80 %
cold buy rate: 0 %
lead who bought with cold score: 0

The scores range from -19 to 19. The hot band has 5 leads and 4 bought, a buying rate of 80 per cent. The warm band has 3 leads and 1 bought, and the cold band has 4 leads and 0 bought. In this tiny sample the score ordered the leads well: hot leads bought most, and nobody in the cold band did. Notice one hot lead that did not buy. A score never promises a sale. It tells you where to spend the first hour. Notice, too, that with twelve leads one or two different outcomes would change the picture, so a real scoring system needs a lot more leads before you trust it.

What to do in each band

  • Hot. A personal message or call within a day, mentioning what they did.
  • Warm. A useful email, such as a guide or a case study, and another chance to ask for a sample.
  • Cold. Keep them on the list with low-effort content, and ask once whether they want to stay.
  • Unsubscribed. Stop. Honour the request quickly, as in lesson 6.
  1. Choose five to eight actions you can record, and give each points.
  2. Set bands, and decide the action for each in advance.
  3. Keep notes on who bought, and compare the bands every few months.
  4. Adjust points that do not predict well, such as opens, which email programs can inflate.
  5. Keep it human. A score ranks people, so use it to be helpful, never to pressure.
Common mistake: counting opens heavily

Over-trusting opens. As lesson 6 explained, open counts can be inflated by email software, so a lead can look warmer than they are. Give more weight to clicks, replies and visits to important pages.

Respect people’s data

Contact records are personal data. Keep only what you need, keep them secure, record how consent was given, and delete data you no longer need. Rules differ by country, and this course is not legal advice.

List five actions that show someone is close to buying your product. Give each points. Score three real or imagined leads, and decide your first action for each.

Practice for this lesson

Check what you learned

Five quick questions. Pick an answer and you will see straight away whether it is right and why.

What is a CRM?

  1. A type of advert
  2. A system that stores contacts and their interactions
  3. A search tool
  4. A logo

It records who each contact is and what has happened.

What is lead scoring?

  1. Counting leads
  2. Deleting leads
  3. Buying leads
  4. Giving points for actions that suggest interest

Points rank leads so you follow up the likeliest first.

How did the hot leads do in the sample?

  1. None bought
  2. Most bought
  3. All bought
  4. They were not tested

Four of five hot leads bought.

Why compare bands with real outcomes?

  1. It is required
  2. It shortens lists
  3. It deletes leads
  4. A score is a guess until tested

Checking predictions against results improves the points.

Why give opens less weight?

  1. Email software can inflate them
  2. They are illegal
  3. They are rare
  4. They cost money

Pre-loading can make opens unreliable.

Summary

  • A CRM records contacts and interactions.
  • Lead scoring ranks leads by actions that suggest interest.
  • Check scores against real outcomes and adjust.
  • Handle contact data with care and respect opt-outs.

Next lesson

Lesson 24 covers advanced SEO and content at scale.

Questions

How many leads do I need to trust a score?

Many more than twelve. With small numbers, one or two outcomes change the picture, so treat early results as hints and keep checking.

Do I need paid software?

Not to start. A spreadsheet with points per action is enough to learn the idea. Tools help once volume grows.

Are the leads real?

No. They are made-up sample leads and outcomes to show the method.

Progress is saved in this browser.