“We’re getting leads but our conversion rate is too low” — where they die

Getting leads but not converting is rarely one broken thing. Your conversion rate is eleven rates multiplied together, and the worst-looking link is usually not the one worth fixing: on 200 enquiries a month, closing a 67% gap in second-call close returns 0.74 deals, while closing a 37% gap in contact rate returns 1.69.

At a glance:

• The loss has a location. Six links multiply in series: contact rate, set rate, show rate, first-call close, second/third-call close, long-term follow-up. Five modifiers (speed to lead, sequence depth, objection handling, reschedule recovery, cancellation recovery) feed them.
• Check the drop is real first. At 20 attended calls, a 20% close rate has an exact 90% confidence interval of 7.1%–40.1%. Under roughly 100 events, a rate is a rumour, not a number.
• Rank by deals recovered, not by the size of the gap. A gap is measured in percent; its value is measured in deals, and every link sits on a different-sized pool.
• Tonight costs nothing: two CSV exports, five counts, four buckets, five phone calls.
• One link per measurement window. Change two and you learn nothing at the end of it.

Why am I getting leads but not converting them?

Because “conversion rate” is not a measurement, it is a product. Between an enquiry and a signed deal sit eleven measurable links, each with its own numerator and denominator, mapped with their formulas in the eleven levers between a click and a closed deal. Six sit in series and multiply. Five are modifiers and recovery loops that push volume back into a series link or change the odds inside one.

Because they multiply, the chain can shed a large share of its output with nothing visibly broken: five links each drifting 10% worse leaves you 59% of the deals and no obvious culprit. A conversion rate does not break — a link inside it does, and total deals, the one number you can already see, is the only one that cannot tell you where the loss happened.

Is my conversion rate really down, or is this just a quiet fortnight?

Half of what feels like a collapse is sampling noise. The NIST/SEMATECH e-Handbook of Statistical Methods works this example: twenty units sampled with four defective gives an observed proportion of 0.20 and an exact 90% confidence interval of (0.071, 0.400). Twenty attended sales calls with four closes is the identical calculation: your true close rate is between 7% and 40%, and last quarter’s 30% sits inside that.

We reproduced that example to check the method, then ran the same exact-binomial calculation at larger samples. Every row is an observed 20% close rate; only the call count changes.

The interval test: if last period’s rate falls inside this period’s confidence interval, you have not measured a drop. At 25 enquiries a month, two quiet weeks is noise and six is a trend. The most expensive thing a bad month causes is a decision taken on twenty events.

The next 24 hours: five numbers, one CSV, nothing to buy

1. Export two CSVs from your CRM — leads with created timestamps, activities with theirs. Use 60 days, not 30: it is the shortest window that gets most businesses enough events to clear the table above.
2. Count five numbers. Leads received; leads reached in a two-way conversation; appointments booked; appointments attended; deals closed. Each is a count, not an opinion. If one cannot be produced, that missing count is your first fix — an afternoon of CRM data hygiene, not a purchase. The denominator traps in the last two are in how to calculate a sales close rate.
3. Bucket last month’s lost enquiries into four piles: never contacted; contacted, never booked; booked, never showed; showed, never bought. Five minutes with a filter, and the biggest pile is your hypothesis, not your answer.
4. Phone the five most recent lost enquiries yourself. Free, unpleasant, and the only qualitative data here.
5. Do none of these four tonight: pause the ads, rewrite the offer, cut the price, let someone go. All four destroy the baseline you need tomorrow and none is reversible inside a week.

The death-point test: which of the eleven links is mine?

Take each link’s rate over the last 60 days and compare it with that link’s best 60-day period in your own history — your past controls for your offer, price and market; a published benchmark does not. Then apply the recoverable-deals rule: rank the links by current deals × (best rate ÷ current rate − 1), not by the size of the gap. The ranking at 200 enquiries a month:

Lead quality shows up here first: poor-fit leads do not answer, so quality lands on contact rate. The biggest percentage gap came fourth on the list that matters. A 67% improvement in second-call close returns 0.74 deals a month because it sits on a pool of 12 opportunities; a 37% improvement in contact rate returns 1.69 because it sits on all 200 enquiries. Rank by the gap and you will spend a quarter fixing the link with the smallest pool attached to it. With no clean history in your CRM, published bands are a stand-in of last resort — what counts as a good sales close rate traces each to a named publisher and denominator — but somebody else’s denominator points you at the wrong link.

Which single fix is worth most at my volume?

Here is the baseline those recovered-deal figures come from, so you can rebuild it with your own numbers.

Return contact rate alone to 52% and every downstream volume scales by 1.368: 4.57 deals becomes 6.26. Return all five links to their own best and the chain produces 10.60 deals — 2.32× at 5.30% enquiry-to-deal, on the same enquiries and ad spend. The five individual gains add to 8.92 deals, not 10.60, because series links multiply rather than add (why small conversion gains compound). Two honest limits: 2.32× is what the arithmetic permits, not what a quarter delivers, and with five links moving at once a result tells you the chain improved and nothing about which change did it.

The next seven days: change one link, not five

Days one and two, build the chain table from the two CSVs. Day three, rank it by recoverable deals. Day four, change the one link with both the most recoverable deals and enough monthly events to clear the interval table; where those are two different links, take the one with the events, because the other is unmeasurable at your volume. Day five, set the review date: one sales cycle plus the events the interval table requires. Days six and seven, instrument the three timestamps that made this week slow — enquiry received, first two-way contact, appointment booked. Changing two links in one window is how a conversion project produces six months of work and no attributable result.

When low conversion is not a conversion problem

• The offer or the price moved. If contact, set and show rates are all inside their historical intervals and only the close rate fell, the chain is intact and something outside it changed. No follow-up cadence fixes that.
• The lead source changed under you. Contact rate collapses first when a supplier resells the same enquiry or widens its targeting. That is a supplier conversation, and your timestamps are the evidence for it.
• It is a contract or cash problem in a conversion costume. Disputing what a lead vendor agreed to deliver is a lawyer’s question; deals closing while the money does not arrive is collections and an accountant.
• Someone has complained about consent. That goes to the regulator — the ACMA in Australia, the FTC and FCC in the United States — and outranks every number here.

What it costs to run this chain yourself, every month

The method above is complete and you can run it without us. The first build takes half a day if your CRM already stamps enquiry-received, first-contact and booked times, closer to a week if you must reconstruct it from call and SMS logs; upkeep is about two hours a month. The skill it needs is not sales skill — it is somebody willing to read a confidence interval and refuse to act on a twenty-event sample.

What breaks at volume is the recovery end: reschedule recovery, cancellation recovery and the 90-day-plus follow-up pool all need contact attempts within minutes, at hours nobody is rostered for. In our own client work, the largest single-link gains we see are on speed to lead, roughly 3× for a business still running 2020 operations rather than 2026 AI-driven ones, with roughly 2× each from doubling contact rate and from doubling set rate. Those three do not multiply: 3 × 2 × 2 = 12× is a double count, because fixing speed to lead is part of how contact rate improves and contact rate is part of how set rate improves. Our published outcome figure is an average 7× sales lift; the methodology page states the method behind it and discloses a median closer to 4×.

LeadsNow runs this eleven-link chain as the operating report on its own campaigns — 50,769+ AI-booked sales appointments since 2017 — and the two links it carries for clients are contact rate and set rate, worked by AI appointment setting at the hours a rostered team does not cover.

Questions people ask when their leads stop converting

How many sales calls do I need before my close rate means anything?

Enough that the confidence interval is narrower than the change you care about. The NIST/SEMATECH e-Handbook of Statistical Methods, section 7.2.4.1 works the example: twenty units sampled with four defective gives an observed proportion of 0.20 and an exact 90% confidence interval of (0.071, 0.400). Twenty attended calls with four closes is the same calculation, so your true close rate is between 7% and 40%. At 100 attended calls the interval narrows to 13.7%–27.7%; at 400, to 16.8%–23.6%.

Should I fix speed to lead first?

Often, because it has the best external evidence of any link, but it is a modifier rather than a series link: it works by raising contact rate, so it only pays if contact rate is your death point. Oldroyd, McElheran and Elkington, in Harvard Business Review, March 2011, audited 2,241 US companies with a test web lead: 23% never responded at all, and the average response among those answering within 30 days was 42 hours. A separate study of 1.25 million leads across 29 B2C and 13 B2B US companies found firms contacting within an hour were nearly seven times as likely to qualify the lead as those contacting an hour later — where qualifying meant a meaningful conversation with a decision maker, not a sale. The data is from 2011.

My conversion rate dropped but nothing changed on my side. What happened?

Check whether enquiry volume moved at the same time. If volume rose and deal count stayed flat, your rate fell without anything breaking — you bought more of a cheaper lead, and the right comparison is deals per month, not percent. If volume was flat and deals fell, run the interval test, then the death-point test.

What is a normal enquiry-to-deal rate?

There is no portable number, because every published close rate carries a different denominator: deals over all leads, over qualified opportunities, or over issued proposals are three different fractions, and a business can report all three truthfully in the same month. A rate without its denominator is not a number. The figures we could trace to a named publisher and method, each with its denominator attached, are collected in our close-rate benchmarks guide.