§ About

Built by the person who kept getting handed the invoice.

Product marketer. GTM engineer. For years I tried to answer two questions — why we win, why we lose — from the calls we already recorded. The market wanted six figures and six months. So I built the thing that made it a Tuesday afternoon instead.

discera · foundersolo & shipping
A
Ahmet Nuri Ozcelik
Founder · PMM + GTM Engineer
  • BackgroundProduct marketing
  • ThenGTM engineering
  • ObsessionWin/loss from real calls
  • NowBuilding Discera
LinkedIn
§ What win/loss used to cost
$40–60K/ year

Billed like a strategy consulting engagement.

3–6months

Of implementation before the first report shipped.

1PDF / qtr

Static, thin-sample, already stale on arrival.

— and the raw material was already ours.

§ The background

A marketer who learned to build.

The trade
Product marketing

Positioning, messaging, competitive intel, launches — all of it dies without real evidence of why buyers move.

The turn
GTM engineering

Wiring the revenue stack, writing the scripts, automating the work everyone else did by hand. I stopped waiting for tools and started building them.

When large language models got good, the first thing I reached for wasn’t a chatbot. It was the pile of sales calls nobody had time to read.

The two questions I could never answer fast enough

Why we won. Why we lost.The answers were always in the calls. Getting them out was brutal.
§ Doing it by hand

The work everyone wants and nobody does.

01
Open the calls one by one

Scrub recording after recording for the moment the deal turned. A dozen calls in, you’ve lost the thread of the first one.

02
Beg reps for context

Chase AEs for the ‘real’ reason a deal died. Everyone remembers it differently, and memory is kind to itself.

03
Hand-build the deck

Paste quotes, tag themes by hand, and by the time it’s ‘done’ it’s a quarter old and the market has moved.

I did this more times than I want to admit. Slow, un-scalable, stale on delivery. So I looked at what the market was selling.

§ The old way vs Discera

Same question. One of these is a rip-off.

Dimension
Managed program
Discera
Price
$40–60K / year
From $29 / month
Time to answer
3–6 months
Minutes
Coverage
A hand-picked sample
Every call in scope
Freshness
A quarterly PDF
On a schedule, always current
The data
Interviews you commission
The calls you already own
Access
Seats you don't need
Read-only, no extra seats
§ The other trap

Gong is built for reps. You wanted the insights.

Already have the recordings? Great. But Gong is priced per seat for salespeople who live in it all day — not for the people who need the patterns across the calls.

Built for
Reps running deals

Per-seat licenses for people who live inside a sales tool all day.

You are
PMM · CI · RevOps · Founder

You need the read across every call — not a seat to run deals.

So I solved it. Discera connects to Gong read-only and turns the calls you already record into the insight layer on top — no extra rep seats, no salesperson prices just to read what your buyers said. You bring the calls; Discera does the reading.

§ How it became Discera

I built the agents I wished I could buy — then realized everyone needed them.

01
Scripts for myself

AI agents that read our calls the way I would — pull the transcript, find the objection, name the competitor, diagnose the loss — across hundreds of calls at once.

02
It beat the agency

Minutes of runtime produced a sharper, broader, more current read than the six-figure program did in a quarter. It read every call and never got tired.

03
Everyone has this pile

Same recordings. Same unanswered questions. Same six-figure quote in a folder. The agents I built for one team were the missing product for all of them.

Discera is the answer I wanted: ask any question in plain English, run it across every call, on a schedule, for a fraction of the price — and keep the evidence you already own.

§ Security posture

Your calls are sensitive. We treat them that way.

Win/loss data is some of the most confidential material a company has. Trust is earned with architecture, not adjectives.

Read the full security page
Encrypted end to end

TLS 1.2+ in transit; AES-256 at rest. Gong and HubSpot credentials Fernet-encrypted, keys in Secret Manager.

Never trains models

Under our terms with Anthropic and OpenAI, your call data is never used to train any model.

Workspace-isolated

Every report, credential, and job scoped to your workspace, enforced at the query level. Read-only everywhere.

SOC 2 in progress

A formal program is underway on Google Cloud in a private VPC with no persistent servers.

Still a founder you can email directly.

A hard question about win/loss or competition — or a bone to pick with something on this page — I’d genuinely like to hear it.