Why we hide leads that might be selling the same thing

A question on the list was wrong

A software services firm we work with gets a daily list of leads from Mach. Each lead is graded against a set of checks before anyone sees it.

An ask for help must pass eight checks to be ready to contact. A company event, like a funding round, must pass seven.

One check asks whether the author sells the same kind of help. A firm that builds software for others is a rival, not a buyer.

Until October 21, a lead that failed only this check could still show as worth a look. It carried an orange question: are they selling it themselves?

On October 21, I looked at the live list and asked why we showed these at all. Nobody had a good answer, so the rule changed that day.

Your rivals use your buyers' words

A seller check matters most for a services firm. Its buyers and its rivals talk about the same things: apps, data, engineers and deadlines.

A web agency's page can read like a request for proposals. A consultancy hiring four engineers looks like a team short of hands. Both are selling.

On October 20, Mach judged 1,321 new posts. 452 read as a seller at 0.6 or more, and every one of them was dropped.

Another 242 sat in the middle, between 0.4 and 0.6. Only 9 of those were kept. That middle band is what this post is about.

Over a third of all judged posts read as clear sellers. If you search for buyers by their words, expect a share like it.

How the seller check reads a post

Mach reads each new post in up to two passes. The first asks a set of fixed questions of every post. One of them is whether the author is selling.

The second pass looks closer at the posts that survive the first. It asks the seller question again, this time against what the firm sells.

Why two passes? The first runs on every post, so it must be cheap. The second costs more, so it runs only where it matters. On October 20, 136 of 1,321 new posts got it.

Each pass gives a number from 0 to 1. Mach takes the higher of the two. Under 0.4, the check passes.

Clear sellers were always set aside. On October 20, 410 of the 1,234 dropped posts were dropped with seller as the reason.

The gap was the middle. A lead between 0.4 and 0.6 failed the seller check, but could still be kept as worth a look.

Being careful here is cheap. The whole judging bill on October 20 was 13 cents, both passes together. The costly part is a person's time.

16 leads sat in the middle

On October 20, 16 kept leads had failed the seller check. 7 came from the October 19 search and 9 from October 20.

None of them was ready to contact. Ready needs every check, so the seller check alone held them at worth a look.

Their higher seller reading ran from 0.40 to 0.53. 14 of the 16 read under 0.5. That is close to a coin toss, which is the problem.

13 of the 16 were hiring bursts. 10 of the 16 also failed a second check: Mach could not find who to contact.

And nobody acted on them. In two days, the firm's team added 5 leads from these searches to its accounts. None of the 5 was one of the 16.

A question on a list is not free. It asks your rep to research a likely rival, and most reps will rightly skip it.

What a seller post looks like

The first four examples below are paraphrased from the 16. We name no company and quote no one.

A request for proposals that was really a web agency's own page. It invited buyers to send their briefs. The words matched a buyer's ask, but the page was selling.

A founder working alone, looking for a partner to build with. The founder already builds tools for software agents. That reads more like a peer than a customer.

Firms hiring two to four engineers at once. A hiring burst often means a buyer needs help. But a services firm also hires engineers, to sell their time to clients.

A new product chief at an operations firm. The post spoke of growth through technology. The second pass could not rule out that the firm sells similar work.

The last example is made up, to show the reasoning step by step. A 120-person firm posts three engineering roles in one week. One of them is a delivery manager.

The first pass reads the ads alone. It sees a team growing fast and gives a seller reading of 0.3. On that reading, it looks like a buyer.

The second pass reads the ads against the firm's own offer. Two ads say the work is for client projects. The company page lists services and client case studies.

Now the seller reading is 0.5, and the higher of the two decides. The check fails. Nothing in the ads was false. The firm is hiring to sell engineers' time, as our firm does.

Seller tells you can check by hand

Mach reads these signs for the firm. If you qualify leads by hand, the same tells work. They are our working list, not findings from the data.

Run the seller test first, before need or timing. A rival can show real need, and it still will not buy from you.

Now: set aside, with a plain reason

Since October 21, a lead that fails the seller check is set aside. It never stays on the list as a question.

The set-aside view gives the reason in plain words: may be selling it themselves. A person can still open the lead and judge it.

The rule lives in grading itself, not in a filter on the screen. Every new search stores the reason with the lead.

Leads graded before the change are read the same way. No stored grade was rewritten by hand.

What we got wrong was having two bars for one question. Ready needed a seller reading under 0.4. Worth a look allowed up to 0.6.

We filled the band between with a question for a person. It looked careful. In practice, it moved our doubt onto a rep's morning.

The cost of the new rule is real. If one of the 16 was a buyer, the firm no longer sees it on the list.

We know such buyers exist. Our own test runs held one company as a real buyer. Yet its seller reading sat in the middle band every time. Under the new rule, it is set aside.

We accept that trade, and the reason is in the set-aside view. The lead is still there, with its reason, for anyone who knows better.

Pitching a rival wastes a morning, and it shows them how you sell. Across both days, 157 leads were kept. Without these 16, 141 remain.

One case stays open. A firm that sells software work can also buy it, to cover a skill it lacks. Those leads now wait in the set-aside view.

If people start opening set-aside sellers and adding them, the rule is too strict. That is the count to watch.

How to tell a real buyer from a competitor

Treat an unsure seller as set aside, not as a question. Questions on a list get skipped, or answered with a guess.

The check needs to know what you sell, in your own words. Mach keeps that in the Brain, its written record of your business.

Write it in one plain paragraph: the work, who does it and how you charge. That gives a seller check something to compare against.

Here is a made-up pair for Tarang Labs, a software team in Pune. "We connect bank systems with our own engineers, billed by the month" can be compared. "Your digital growth partner" cannot.

The first day's teardown showed sellers as the biggest kind of noise. This rule closes the gap in the middle.

One mistake to avoid: lowering the seller bar to keep more leads. A longer list with rivals on it is a worse list.