How we build the Desert Camel Report, and what we refuse to publish

We publish figures with their source, sample and date: outside figures as their publishers give them, and community figures only above the floors below. Where we do not have enough, the page says “Not enough data yet” and shows how many people we have.

• We never model or estimate a number we don’t have, and never fill a thin cell with a regression.
• We never publish a minimum, a maximum or a single person’s value.
• We never name an employer or show any figure for one company.
• We never put figures from different reports, years or samples on one axis.
• We never chart an outside source whose licence does not allow it; those stay in text until the publisher agrees.

• Outside sources, cited: the register below. Shown as “Sourced”, with “global” or “reported by” where that applies.
• Individual contributions from SDRs, AEs and RevOps people, screened. Shown as “From our community” with the number of people and companies.
• Company comp plans from sales leaders, one row per company, published as their own series and never pooled with individual rows.

Every published cell will show how many of its rows are verified from its first number. A “validated” label waits until we know how many people choose to verify.

Every published cell reports the median of all people and the median of company medians (one company, one vote). If the two differ by more than 15%, the method line says so. Each cell also lists how many rows were excluded, and why.

• Window. A cell uses 12 months of contributions. Recency is shown, not filtered: every cell prints its median as-of month and the share of rows from the last 6 months.
• Compensation. Monthly figures are annualised (× 12). Guaranteed cash is basic salary plus fixed allowances; OTE is guaranteed cash plus variable pay at 100% of target.
• Currency. Cells publish in the currency of the compensation (EGP, SAR, AED or USD). The small multiples and the pooled region convert to US dollars at one stated rate per edition: the official pegs for SAR and AED, and the Central Bank of Egypt’s official rate on the edition’s cut-off date for EGP, each named with its date.
• The pound. If the official EGP rate moved more than 10% inside a cell’s window, the cell says so and shows its US dollar value at the start and end rates.
• Nominal money. Every money figure is nominal, labelled with its year, and not adjusted for inflation.
• Gross and net. Egyptian base compensation is collected gross. An optional net figure is used only on the contributor’s own screen and never published.
• Percentiles. Linear interpolation (Hyndman-Fan type 7). Public labels read “Middle half of people: X to Y · typical: Z”.

No free filters and no per-company figures: combining filters lets anyone subtract their way to one person’s compensation. We publish these cuts only.

Contributions are not open yet. Nothing on the report collects compensation, quota or skills answers until they open. When they open, answers are sealed (encrypted) the moment they arrive, kept out of our CRM and never used for sales follow-up. Signals sign-ups are separate: we keep your name, email, the editions you pick and, if you give them, your role and market, plus the page you signed up on, in our contact list.

See our Privacy Policy.

• Skills are rated on a 5-level scale anchored to SFIA: Aware, Foundation, Practitioner, Advanced, Expert. Each level is written as behaviour.
• Target levels per role are RIVET’s proposal, not market data. The evidence behind them is thin and we say so; they change as the evidence grows.
• Gaps carry equal weight until our own count of regional job posts reaches its floor.
• A course is shown only when it fits the role, the level, the entry requirements and is available now. Every recommendation says it is RIVET’s own course. When no RIVET course teaches a skill, the check says so and offers a free first step.

Every outside source we considered, including the ones we rejected. A figure on a report page carries its source ID; if the ID is not here, the figure does not ship.

Opening every cell of the role-by-market grid as an early signal needs at least 90 usable contributions if they spread evenly, and 270 to make every cell a benchmark. The pooled region opens first.

Spotted a mistake or a source we missed? Tell us; we check it for the next edition.

Everything on the report’s pages is a row in an editorial sheet, and a row only reaches a page once a person at RIVET has approved it.

1. A routine proposes. A scheduled Claude routine reads public announcements, public posts and public salary guides and writes what it finds into the sheet as new rows: hiring moves, voices, compensation sources, skills and the Signals counts. Every proposed row carries a source link, the date the source carries, and the sample the source states.
2. Every proposed row is pending. A pending row is invisible. It is not on the page, not in a count, not in a chart.
3. A person approves or rejects. We read the source, check the figure against what the page would say, and set the row to approved or rejected. A rejected row stays in the sheet with its reason, so the same source is not proposed again.
4. The page is rebuilt. The approved rows are exported, the build script turns them into the page’s data file, and the page is generated from it. Only approved rows are built in, so the page cannot show a row nobody read.
5. The stamps are the data’s own. Each block says “Updated” and the newest date in that block. We never promise a day of the week or a time of day, because the data does not arrive on a timetable.

The first edition’s rows were researched and sourced in one pass and marked approved so they could ship together. Every one of them can still be rejected in the sheet, and the next build drops it.